# Extended AI-readable corpus — gaganmalik.io Last-Updated: 2026-08-14 (build; regenerated on deploy). Short index: [llms.txt](https://www.gaganmalik.io/llms.txt) ## Tiered files (optional smaller fetches) - [Case studies only](https://www.gaganmalik.io/llms-stories.txt) - [Newsroom only](https://www.gaganmalik.io/llms-newsroom.txt) - [Lab prototypes only](https://www.gaganmalik.io/llms-labs.txt) - [ai.txt](https://www.gaganmalik.io/ai.txt) - [Entity JSON](https://www.gaganmalik.io/api/knowledge/summary) --- ## About Gagan Malik (canonical profile) Gagan Malik: Product and Design Executive. Board advisor to CEOs and CXOs. Product and Design Executive. Board advisor to CEOs and CXOs. Former Wipro Partner, Chicago Booth MBA · 100M+ users · $3B+ revenue influenced · consultant to Fortune 100 global brands. Product and Design Executive. Board advisor to CEOs and CXOs. Former Wipro Partner, Chicago Booth MBA · 100M+ users · $3B+ revenue influenced · consultant to Fortune 100 global brands. I lead product and design programmes for scaling companies: building teams, shipping measurable outcomes, and advising boards on design-led growth across technology, retail, banking, and insurance. For CEOs and CXOs who need boardroom-level product and design counsel from someone who has operated at scale. Former Wipro Partner, Chicago Booth MBA · 100M+ users · $3B+ revenue influenced · consultant to Fortune 100 global brands ### How I design - **Frame with the job and the metric**: Every engagement starts with the user job to be done, the business constraint, and the success metric we will use to judge the work before anyone opens Figma. - **Explore three options, kill with evidence**: I sketch real alternatives, run the cheapest test that can falsify them, and record what we rejected and why. Stakeholder taste is input, not a verdict. - **Tie each pattern to a measured outcome**: Shipped UI links to a research signal or experiment result. Programme scale sits in the footnotes; the case study shows which component moved which number. ### How I lead - **Ship fast, iterate faster**: I don't wait for perfection. I validate with real users, kill projects when data says so, and move at startup velocity even in enterprise contexts. - **Design drives revenue, not just delight**: Every design decision ties back to business outcomes. I speak boardroom language while maintaining design craft. - **Lead teams of 10+, influence C-suite**: I'm a P&L owner, not just a design lead. Proven at 0→1 and 1→100M scale. ### Currently - Product and Design Executive, gaganmalik.io — [gaganmalik.io](https://www.gaganmalik.io/) ### In the past - Founder & CEO, Presto — [Presto](https://askpresto.com/) - Co-founder, Permitech (Techstars 2023) — [Techstars Portfolio](https://www.techstars.com/portfolio?q=permitech) - Partner, Wipro — [Wipro](https://www.wipro.com/) - Lead UX Designer, John Lewis & Partners — [John Lewis](https://www.johnlewis.com/) - Product Manager, Aviva — [Aviva](https://www.aviva.com/) ### Education - MBA, The University of Chicago Booth School of Business — [Chicago Booth](https://www.chicagobooth.edu/) - MSc, Technology Entrepreneurship, UCL — [UCL](https://www.ucl.ac.uk/) - Entrepreneurship Summer School, London Business School — [LBS](https://www.london.edu/faculty-and-research/strategy-and-entrepreneurship/entrepreneurship-summer-school) - BSc Honours, Computer Science, University of Greenwich — [Greenwich](https://www.gre.ac.uk/) ### Top case studies - **Lloyds Banking Group** (Design Director): Saved £15M annually in call center costs Led 10-person design team with P&L ownership Delivered 75% faster resolution times for 5M+ customers Negotiated C-suite buy-in across 4 brands Killed £2M initiative 6 months in when user research invalidated core assumptions, saving 12 months and £5M+ - **Google AdSense** (Design Lead): Unlocked millions in revenue optimization Email campaign driving AI Auto Ads adoption for 2M publishers Increased onboarding completion rates Accelerated feature adoption at scale - **EE UK Telecom** (Navigation Redesign): Solved 54% user navigation failure rate Navigation redesign for £2.8B telecommunications site Streamlined to 3-section mega nav structure Achieved 100% alignment with user mental models ### What I'm open to - **CEO / CXO advisory**: Product design counsel for leaders who need operator experience at board level - **Board advisory positions**: Design strategy, product vision, growth acceleration - **CXO / VP Design roles**: Strategic leadership without full-time commitment - **Strategic consulting engagements**: High-impact projects with measurable ROI ### Testimonials - "Consistent focus on the user truly changed how we think about solutions." — James Riggs, Apple - "Rare ability to balance business, design and technology." — Sumit Bhardwaj, Aviva --- ## About page FAQs ## How do I work with Gagan? Gagan works as a Product and Design Executive and board advisor to CEOs and CXOs, plus strategic consultant and design leadership for scaling companies. Engagements start with a discovery call. Reach out via email or LinkedIn, or use the Ask page for quick questions. ## What does Gagan offer? Board-level product and design counsel, CXO advisory, and strategic consulting for CEOs and scaling companies from £10M to £100M+. Capped retainer plans (Starter, Pro, Max) add ongoing design leadership on tap. Focus: design-led growth, 0→1 and 1→100M scale, measurable outcomes. ## Does Gagan still do hands-on design work? Yes. Board counsel and design leadership include hands-on framing, critique, and shipping when that is what moves the metric. Pure slide decks without product craft are not the engagement. ## How can I get in touch? Email getintouch@gaganmalik.io, connect on LinkedIn, or use the Ask page for quick answers about UX, product and design, or working together. All contact options are listed below. ## How much does it cost? Plans range from Starter (£4,750) to Max (£14,750) depending on scope and engagement length. See the Pricing page for full details. --- ## Pricing FAQs (English) ## Pricing FAQs ### What's included in each plan? Starter, Pro, and Max include regular check-ins, priority design leadership requests, Slack or channel access, and a shared task board. Pro adds stakeholder workshops, ongoing research, and design system guidance. Max adds strategy workshops, a dedicated design lead, and priority support. Capacity is capped — see Compare plans for workstreams per tier. ### Is there a free trial? There's no free trial. You can book a discovery call to discuss your needs, or start with the one-day Quick-wins consultation (£975 ex VAT) to get immediate improvements before committing to a subscription. ### How do I get started? Book a discovery call to align on scope, or choose a plan above and click Get started to complete checkout securely via Stripe. --- ## Case studies ### Simplifying sales reporting (Apple) [View case study](https://www.gaganmalik.io/en/success-stories/apple) [Retrieval brief — apple-sales-reporting-ai] Category: technology Thesis: A unified mobile-first sales analytics app replaced eight fragmented tools and gave field and leadership one source of truth with offline trust. Answers: What did Gagan do at Apple? | How did Gagan unify Apple sales reporting? | Why does mobile-first analytics matter for field sales organisations? Implication: Consolidate fragmented reporting before adding AI; design offline and last-sync states where the job is airports and shops. Proof: Eight apps replaced by one dashboard; 78% of opens outside office hours (pilot markets); Weekly reconciliation meetings fell directionally per territory; 150K users in sales organisation Concepts: enterprise analytics, mobile-first, single source of truth, field sales **The account executive is in a retail store and leadership wants the number now.** Eight apps meant eight versions of the truth: hardware revenue in one tab, services attach in another, channel mix somewhere else. Field teams reconciled spreadsheets in airport lounges while partners waited. The opportunity was one live picture everyone could trust, with answers in the app instead of attachment chains. **The Opportunity** — Eight tools, no shared truth for performance questions Apple's global sales organisation ran on a patchwork of tools. Teams across major regions jumped eight apps for basic performance questions: hardware revenue, services attach, channel mix, quarterly targets. Figures did not always match, field work slowed, and nobody had one live view that tied hardware, services, and channel together. The job-to-be-done was clear: give account executives, leadership, and partners one source of numbers they could trust for demand conversations and quarterly plans. I led discovery that included field interviews across regions, shadowing sales calls, and mapping the reconciliation work people did weekly. Success meant fewer debates over which spreadsheet counted and more time on execution. **The Solution** — One mobile-first analytics app: same numbers for field, leadership, and partners I designed a mobile-first sales analytics app so fragmented data lived in one place for field teams. Three options were considered. A web dashboard that required VPN and a laptop: rejected, because account executives work from airports and retail floors, not desks. A read-only mobile view of existing tools: rejected, because it would perpetuate the eight-app problem. A unified mobile-first app with offline support and role-based views: what I shipped. I owned the key UI decisions: dashboards that drew the same numbers for every product line, region, and channel; analytics layers that supported quarterly planning with drill-down by territory and SKU; channel views that let account executives see distributors, carriers, and retail side by side; reporting that replaced hand-built spreadsheets with up-to-date figures for spotting gaps. I integrated the app with Apple's existing CRM and designed offline modes with explicit last-sync states so account executives could trust cached territory numbers without VPN at airports and retail floors. That pattern later explained why most opens happened away from desks. I validated the design through field pilots in three markets before global rollout. **The Impact** — One live view replaced eight tabs and cut weekly debates over the numbers One app replaced the habit of checking eight systems and disagreeing on the figure. Account executives saw the same real-time picture as leadership, so demand conversations and channel plans started from shared data. Partner briefings matched what executives saw without reconciliation drag. Less time matching spreadsheets meant more time on planning and execution; field plans could stay aligned with quarterly goals because everyone started from the same screen. Because I unified hardware, services, and channel into one mobile dashboard with offline cache and last-sync states, account executives checked one screen instead of rebuilding spreadsheets before partner calls, explaining why 78% of app opens happened outside office hours and weekly reconciliation meetings per territory fell by roughly half (directional, pilot markets). Takeaways: - One source of truth removes most of the reconciliation work that slowed decisions. - Mobile-first with offline coverage matters when your job is airports and shops, not desks. - When leadership and the field read the same dashboard, plans stop arguing with each other. --- ### Better onboarding and activation (Google) [View case study](https://www.gaganmalik.io/en/success-stories/google) [Retrieval brief — google-product-activation] Category: business Thesis: Segmented activation email by stall point and CMS beat generic welcome blasts for publisher onboarding. Answers: What did Gagan do for Google publisher activation? | How does Gagan approach product activation and onboarding? Implication: Segment onboarding by behaviour and stall point; show the expected outcome in the first screen of each email. Concepts: activation, onboarding, email journeys, publishers **The snippet sits in your clipboard and the revenue line stays flat.** Auto Ads could find inventory site owners missed, but most publishers never finished setup. One tag, scattered docs, and a black-box story about ML meant capable tools sat idle while manual placements felt safer. The job was not to explain the algorithm louder. It was to meet each publisher where they stalled and get them to the first impression. **The Opportunity** — Two million publishers, one snippet, and a 58% cliff before the first ad shows Auto Ads used machine learning to place ads and surface inventory owners overlooked, yet adoption lagged behind the product's depth. The integration was a single snippet, but it read opaque if you did not ship code every day. Manual placements stayed familiar; the ML path sounded like a gamble you could not verify from a kitchen table on a Sunday night. My JTBD research surfaced three blockers: unclear value (why trust ML over hand-placed units?), technical confusion (where does the snippet actually go on WordPress versus a custom CMS?), and verification anxiety (how do I know it is working?). Support tickets showed 'Auto Ads not showing revenue' as the top activation issue. Publisher surveys found 68% did not understand the difference between Auto Ads and manual placements. Funnel analysis showed 58% drop-off between copying the snippet and seeing the first ad impression. Constraints were strict: no in-product UI changes, email-only intervention. Messages had to work across WordPress, custom CMS builds, and static sites. Success metric upfront: activation above 35% within 30 days of signup. **The Solution** — Three emails for three stalls: explain, paste, prove I closed the gap with segmented email so each publisher saw copy matched to where they were stuck, not a generic welcome blast. A single email to all signups could not serve technical and non-technical owners equally; platform diversity across WordPress and custom stacks ruled out an in-product tutorial overlay. I shipped a behaviour-led email journey segmented by signup source and stall point. Each email had a fixed layout hierarchy: subject line named the stall ('Paste your snippet in 3 steps'), hero screenshot showed the expected outcome (revenue line moving), numbered steps with CMS-specific screenshots, and a single primary CTA ('Verify in reporting') so publishers knew what to do next without opening docs. I wrote the first email to explain what Auto Ads actually did, with side-by-side examples of ML placements next to manual ones. I wrote the second, sent to publishers who read the docs but never pasted, to walk through code placement with screenshots for common CMS platforms. I wrote the third, sent to those who pasted but saw no impressions, to cover reporting verification and troubleshooting. I wrote follow-ups that carried publisher proof (for example Sarah increasing RPM 28% without touching placements) and deeper technical paths for owners ready for mobile anchor ads, AMP, and cross-device setups. I validated before scale. A/B tests on subject lines, send timing, and technical depth ran against 50K publisher cohorts before full rollout. **The Impact** — From 22% to 41% activation, with 35% fewer setup tickets Activation climbed from 22% to 41% within 30 days of signup after the segmented programme launched, clearing the 35% target I set upfront: segment 2 (paste instructions) recovered the largest stall cohort, and segment 3 (verify impressions) converted publishers who had pasted but never saw revenue move. Support tickets tied to Auto Ads setup dropped 35% quarter-over-quarter because screenshots replaced abstract ML copy at each stall point. Segmenting email by stall point with CMS-specific paste steps and reporting verification screenshots lifted activation from 22% to 41% and cut setup tickets 35% quarter-over-quarter. The same segmentation patterns later applied to matched content and in-feed ads with similar lift. Takeaways: - Segmented email meets publishers at the stall point, not at signup vanity. - 22→41% activation and −35% setup tickets followed stall-mapped email layout. - Peer revenue proof beats feature lists when trust is the real blocker. --- ### Orderpad UX audit (research → design spec) (Just Eat) [View case study](https://www.gaganmalik.io/en/success-stories/just-eat) [Retrieval brief — improving-restaurant-order-tracking] Category: design Thesis: Five-brand heuristic research with live kitchen observation replaces opinion with fundable backlog priorities for partner UX. Answers: What did Gagan do at Just Eat? | How does Gagan run competitive UX audits for marketplaces? Implication: Score competitors on one rubric during live service; attach survey percentages finance can sequence. Proof: Orderpad ranked last in five-brand roll-up; 82% wanted Partner Centre access from Orderpad; 56% wanted driver visibility on busy nights Concepts: UX research, heuristic evaluation, marketplace, B2B **Saturday night: three tablets, one counter, and nobody sure which order the rider is collecting.** Restaurant teams run Just Eat's Orderpad beside rival apps on the same surface. When a phone line, a rider app, and multiple aggregators compete for attention, staff reach for the tablet that makes the next step obvious. Just Eat asked for a side-by-side review against Deliveroo, Uber Eats, Grubhub, and SkipTheDishes to see where friction appeared first. **The Opportunity** — Growth rode on counters where Orderpad lost the five-brand comparison Just Eat's partner base depended on small kitchens tolerating chaos. Orderpad could accept orders, but it often failed to show where each ticket sat in the flow, when food needed to be ready for handoff, or what tapping 'On Its Way' committed to downstream. Teams worked around those gaps; the cost showed up as late deliveries, support calls, and poor reviews. My JTBD research identified three frustrations: unclear order status (prep versus ready versus collected), no direct access to Partner Centre tasks (hours, radius, pricing) during service, and weak driver visibility compared to rivals. I led discovery that included in-restaurant observations during live shifts (Friday and Saturday nights, lunch rushes), partner surveys via Qualaroo in-product pop-ups, and a Nielsen heuristic framework [NN/G heuristics](https://www.nngroup.com/articles/ten-usability-heuristics/)::pill applied across five brands. Constraints: audit only, no implementation commitment. Recommendations had to separate quick wins from structural changes requiring policy or contract shifts. **Out of scope for this engagement:** Orderpad pixel changes, A/B tests, or production UI. The deliverable was research, prioritisation, and a design spec for backlog funding. Success metric upfront: one scoring sheet all stakeholders could point to, with partner survey percentages finance could fund against. **The Solution** — Same rubric across five brands: walkthrough, survey, live shift I structured the audit around six Nielsen-based review areas: system status visibility, fit for kitchen work, control and speed, recognition over recall, layout clarity, and error recovery. The same rubric ran across Just Eat, Deliveroo, Uber Eats, Grubhub, and SkipTheDishes so results were directly comparable. Three audit scopes were considered. Product walkthrough only: rejected, because it missed real operating conditions. Partner interviews only: rejected, qualitative feedback was abundant but unfunded. Combined product analysis, surveys with percentages, and live observation: what I shipped. I paired product walkthroughs at realistic order volumes with competitor benchmarking on the same six heuristics. Deliveroo and Uber Eats showed live driver maps and explicit prep→ready→collected timelines; Grubhub and SkipTheDishes surfaced Partner Centre shortcuts from the order screen. Those were the patterns Orderpad lacked on status visibility and in-flow partner tasks. Qualaroo in-product surveys captured importance ratings for specific features. In-restaurant observation during live service covered noise, rush periods, and multilingual teams. The output was one scoring sheet showing Just Eat ranked last across the five brands, theme briefs with screenshots on status visibility, Partner Centre access, and driver tracking, and a prioritised readout separating quick fixes (clearer button labels) from business-level decisions (opening Partner Centre API to Orderpad required cross-functional alignment). I specified a recommended status rail — prep, ready, collected, driver en route — as the design target when Orderpad could not show where a ticket sat during live service. That spec became the bridge from audit to backlog: finance could size 82% Partner Centre demand and 56% driver-visibility demand against a concrete UI pattern, not a theme label alone. Field validation was the point. Kitchen visits during real service caught what desk walkthroughs missed: which status gaps cost seconds when the rider was already at the door. **The Impact** — Orderpad ranked last, with 82% and 56% survey counts finance could fund Across Deliveroo, Grubhub, Uber Eats, SkipTheDishes, and Just Eat, the combined heuristic roll-up put Orderpad last. That was uncomfortable in the room, but it replaced opinion with one curve everyone could point to: the gap ran across status, fit to kitchen reality, control, and recovery, not a single bad icon. In-product surveys showed 82% of partners rated reaching Partner Centre from Orderpad as important. 56% wanted a clearer view of drivers on busy nights, while optional driver tooling sat at almost no adoption next to rivals who ship live tracking by default. Those numbers turned 'they keep asking' into backlog pressure with a size on it. Nothing in an audit ships pixels by itself. What shipped was clarity: ranked themes, a cut between work that could ride normal release cadence and work that needed brand or policy concessions, and constraints spelled out so roadmap bets matched what ops could open and what contracts allowed. Impact for this engagement is measured in **research outcomes**, not conversion: a funded backlog sequenced by partner survey weight, a single five-brand curve stakeholders could point to in prioritisation, and a status-rail design spec ready for implementation when policy allowed Partner Centre access from Orderpad. Because the recommended status rail named prep→ready→collected→driver en route explicitly, stakeholders could size backlog against expected kitchen behaviour (fewer 'where is this order?' moments during rush) rather than a vague 'status' theme. That clarity is why 82% Partner Centre demand and 56% driver-visibility demand converted to sequenced roadmap bets. Because I scoped the engagement to research and design spec only (no Orderpad pixels or A/B), the deliverable stayed honest: funded backlog and a concrete UI target, not implied conversion lift. Takeaways: - A shared score across five brands beats slide battles when every team has a different favourite fix. - 82% Partner Centre and 56% driver visibility — survey percentages finance could sequence. - Audit scope was research and spec only; no Orderpad pixels or A/B shipped in this engagement. --- ### Gamified car insurance (Aviva) [View case study](https://www.gaganmalik.io/en/success-stories/aviva) [Retrieval brief — aviva-gamified-car-insurance] Category: business Thesis: Gamification in insurance works when badges and scores tie to real premium outcomes, not engagement for its own sake. Answers: What did Gagan do at Aviva? | How did Gagan design gamified car insurance at Aviva? Concepts: insurance, gamification, safe driving, premium value **The premium quote still prices you by postcode, not by how you drove home last night.** UK motor cover leaned on coarse segments: safer drivers paid like risky ones because price was not tied to trips on the road. Aviva saw a chance to differentiate with behaviour-based pricing and engagement: reward safer driving and make insurance feel personal instead of punitive. **The Opportunity** — Demographic buckets masked real driving risk on the road UK car insurance still leaned on broad demographic buckets: safer drivers often paid like risky ones because price was not tied to trips on the road. Most people think they drive better than average, so flat stereotypes felt unfair and missed real risk. Aviva needed a signal from the vehicle, not only the postcode on the form. My JTBD research showed two core needs: young drivers wanted a fair chance to prove safe behaviour and lower premiums; families wanted transparency on what behaviours actually affected price. I led discovery that included ride-alongs with telematics pilot users, surveys with 2,000+ drivers, and competitive teardowns of black-box hardware programmes. Constraints: app had to work on iOS and Android without specialist hardware; scoring needed to feel fair and explainable; the app had to connect directly to Aviva's quoting and policy systems, passing a driver's score to the quote engine at conversion without manual re-entry. Success metric: at least 30% of users qualifying for discounts within 200 miles, with positive ROI within two years. **The Solution** — UK-first smartphone telematics: score the trip, quote from behaviour I led design for the UK's first smartphone-based telematics app, pricing cover from how people actually drive. Three options were considered. Black-box hardware installed in vehicles: rejected for installation friction and cost. OBD-II dongles: rejected for limited vehicle compatibility. Smartphone GPS and sensors: what I shipped. Any driver with a smartphone could participate without hardware installation or vehicle compatibility limits. I owned the key UI decisions: the app used GPS to log acceleration, braking, cornering, speed, and phone use over 200 miles. Drivers got a 1 to 10 score, quick feedback after each trip, optional sharing with family, and a direct path into a quote. I designed trip breakdowns to show star ratings for braking, acceleration, cornering, and phone use so drivers saw where behaviour improved, not just a single headline score. I designed a badge system that recognised milestones (first 200 miles, 30-day safe streak, night driving mastery). I validated through beta testing with 5,000 drivers across urban and rural routes; A/B tested badge visibility (visible badges increased 200-mile completion by 18%); tested score explanation copy with non-technical users to ensure fairness perception. **The Impact** — Downloads above plan, measurable savings, brand lift on invention Aviva Drive passed 1M downloads in about six weeks, well above the annual plan, picked up Apple's 'App of the Week', and held around 4.5/5 in the store. Roughly four in ten users scored 7.1 or above and saved about £101 on average; most people in the programme qualified for some discount. Because I designed trip breakdowns with star ratings per behaviour and visible badges for milestone progress, drivers corrected braking within three trips and 18% more completed 200 miles when badges were visible in beta, feeding discount qualification at scale and turning the business case positive inside two years. Perception of Aviva as inventive moved up by 33 points, so telematics read as a practical product bet rather than a short campaign. Takeaways: - Behaviour-based pricing rewards actual performance instead of demographic assumptions. - Visible badges and per-trip star ratings drove 18% more 200-mile completions in beta. - Smartphone telematics can reach positive ROI within two years at scale. --- ### From Hold to Handled (Lloyds Banking Group) [View case study](https://www.gaganmalik.io/en/success-stories/lloyds-bank) [Retrieval brief — lloyds-from-hold-to-handled] Category: business Thesis: Self-serve insurance with progressive disclosure and one design system across four brands cut contact-centre handle time and shifted volume to digital. Answers: What did Gagan do at Lloyds Banking Group? | How did Gagan reduce insurance call centre handle time at Lloyds? | What is progressive disclosure in regulated self-serve flows? Implication: Invest in one cross-brand pattern and rule-based flows before bespoke per-brand UX; measure digital self-serve share and resolution time. Proof: ~75% resolution time reduction on targeted call types; 60% digital self-serve within six months; 80%+ amendment completion across four brands; 26M+ users served Concepts: banking, insurance, self-serve, design system, contact centre, progressive disclosure **Twenty minutes on hold to change an excess while the kettle boils twice.** Lloyds Banking Group's home insurance contact centres carried amendments, cancellations, and policy queries that ate most of the day. Customers wanted cover changed without the wait. Agents wanted policy truth without tab-hopping across four brands. The opportunity was one pattern that put accurate policy actions in front of staff and customers at once. **The Opportunity** — Three call types, 250 FTEs, and resolution stretching past fifteen minutes The UK's largest retail bank faced acute pressure in home insurance contact centres: high volume, rising cost, frustrated customers. Three call types (amendments, cancellations, policy or cover queries) consumed nearly 75% of total handling time and over 250 FTEs. Average resolution ran from about six minutes to well over fifteen, leaving little slack for good service. My JTBD research identified three needs: customers wanted to amend cover (add contents, change excess) without 20-minute waits; agents needed faster access to policy details during calls; operations needed consistent data capture across four group brands. I led discovery that included contact centre shadowing across Halifax, Lloyds, Bank of Scotland, and MBNA, call recording analysis on repeat questions, and stakeholder mapping showing each brand on different back-end systems. Constraints: no changes to core policy admin systems; flows had to work for all four brands with minimal forking; regulatory compliance required specific disclosures at amendment. Success metric: handling time reduction above 60% for target call types, with digital self-serve crossing 50% within six months. **The Solution** — One design system, rule-based flows, volume shifted to digital As design director I led strategy and team coordination and owned hands-on design of progressive disclosure on amendment flows, asking for information only when the customer's path required it. Bespoke flows per brand would have put Halifax, Lloyds, Bank of Scotland, and MBNA on diverging maintenance paths with inconsistent experiences. Generic CRM forms with agent scripting would have perpetuated hold times. I built one design system with rule-based flows for self-serve and agent use across all four brands. I aligned the team on shared components, validation patterns, and error states so updates propagated across all four brands without rework. I designed amendment flows to open with one question ('Do you have your policy number?'), then branched to collect only relevant fields. I carried clear next steps in mobile-first layouts, disabled primary actions until forms were complete, and plain-language summaries before submission. Self-serve for policy queries reduced phone volume without pushing complexity onto customers. Agents stayed free for fraud, claims disputes, and vulnerable customer support. I prototyped quickly and tested with 200+ people across all four brand customer bases before rollout. Rule-based routing meant the first payload matched how dispatch and servicing were sequenced, not a generic CRM shape agents had to rewrite. Validation started in the field. Pilot launched with Halifax first (highest volume), instrumented handling time and completion rates, then rolled to Lloyds, Bank of Scotland, and MBNA sequentially. **The Impact** — 75% cut in resolution time, 60% digital self-serve in six months Lloyds moved the heaviest insurance jobs into digital self-serve: the same routes that had consumed three-quarters of handle time. Average resolution time fell by about 75% for targeted call types. Digital self-serve crossed 60% within six months, exceeding the 50% target. Because I designed progressive disclosure on amendment flows (one starter question, then only relevant fields), customers completed amendments in under four minutes on mobile instead of fifteen on phone, and agent satisfaction improved as teams handled judgement-heavy work instead of repetitive data entry. Completion rates on amendment flows stayed above 80%. Directionally, the programme freed capacity equivalent to roughly 60 FTE across targeted call types as volume shifted to digital self-serve. Takeaways: - 75% resolution time cut and 60% digital self-serve within six months on targeted call types. - Progressive disclosure kept amendment completion above 80% across four brands. - One design system across four brands cut rework without flattening brand identity. --- ### Better product discovery (EE) [View case study](https://www.gaganmalik.io/en/success-stories/ee) [Retrieval brief — ee-global-navigation-redesign] Category: design Thesis: Tree-tested information architecture with three clear buckets beat relabelling eight menus for telecom task success. Answers: What did Gagan do at EE? | How did Gagan redesign EE global navigation? | When should a CEO fund tree testing over stakeholder opinion? Implication: Run tree tests on high-frequency tasks before IA debates; structure nav around jobs (Shop, My EE, Help) not org charts. Proof: 200+ tree-test participants; 20 high-frequency tasks validated; Three-bucket IA (Shop, My EE, Help) Related essays: pmf-is-not-a-funnel Concepts: telecom, information architecture, tree testing, navigation **The shopper needs the coverage checker and the site offers eight menus with overlapping labels.** EE's navigation buried key tasks under ambiguous names: coverage, upgrade paths, and account help often meant three wrong clicks or a call to support. Tree testing showed where the structure failed; task success scores gave us the mandate to regroup. **The Opportunity** — Shoppers lost in eight top-level menus with overlapping labels EE ran eight top-level navigation items with unclear boundaries: 'Why EE', 'Network', 'Help', and five product categories competed for attention. Customers hunting for coverage, upgrade eligibility, or account help often clicked three menus before finding the right path or abandoned to call support. My JTBD research identified three core missions: shopping for devices and plans, managing existing accounts (My EE), and getting help and network status. I led discovery that included clickstream analysis showing high bounce rates from nav, support ticket tagging revealing 'could not find coverage checker' as a top issue, and stakeholder workshops mapping business priorities against customer intent. Constraints: rebrand was launching in parallel so nav had to accommodate new product naming; back-end content management required minimal restructuring. Success metric: task completion rates above 70% for top ten customer journeys. **The Solution** — Tree testing first, then three buckets: Shop, My EE, Help I ran tree testing with 200+ participants across twenty high-frequency tasks (finding coverage checker, comparing pay monthly plans, checking network status, understanding upgrade eligibility). Three options were considered. Keep eight menus with relabelling: rejected, because tree tests showed structural confusion, not only labelling issues. Mega-menu with all options visible: rejected for mobile complexity. Three clear buckets (Shop, My EE, Help) with supporting sub-navigation: what tested highest for task success. I owned the key IA decisions: 'Shop' carried all acquisition paths (phones, SIM-only, broadband, business); 'My EE' grouped account management, bills, and upgrades for existing customers; 'Help' contained network status, coverage checker, contact options, and troubleshooting. I moved coverage checker from buried under 'Why EE / Network' to visible under 'Help', I validated through card sorting with 150 participants. Sub-navigation used consistent patterns: product type, then contract type, then device or add-ons. Validation approach: tree testing showed task success improvement from 52% to 78% for coverage tasks, 61% to 82% for upgrade paths, and 48% to 74% for network status checks. A/B tested the redesign in prototype with 300 users before production rollout. **The Impact** — Task success up 26 points average; coverage and help calls down Tree testing and task success scores gave us clear evidence: the three-bucket structure improved completion rates across all twenty tested tasks. Coverage checker usage increased 34% in the first quarter after launch because I moved it from buried under Why EE / Network to Help › Coverage checker: the same IA change that raised task success on coverage tasks from 52% to 78% in tree tests. Support contacts related to 'cannot find information online' dropped 22% in the same period. Because I moved high-intent tasks into Help where shoppers already look, average clicks to task completion fell from 4.2 to 2.8 across measured journeys, cutting support noise without over-promising on acquisition (Shop conversion improved a modest 3–4 points consistently across mobile and desktop). Takeaways: - Tree testing and card sorts pay off when navigation carries most of the discovery job. - Moving coverage checker to Help drove +34% usage and −22% 'cannot find' contacts. - Three buckets beat sprawl when shoppers need My EE, Shop, and Help without guessing. --- ### Faster checkout (John Lewis) [View case study](https://www.gaganmalik.io/en/success-stories/john-lewis) [Retrieval brief — john-lewis-faster-checkout] Category: business Thesis: Discovery and checkout UX with defensible attribution tied filters, Quick View, and mobile checkout to measurable incremental revenue. Answers: What did Gagan do at John Lewis? | How did design drive revenue at John Lewis? | How does Gagan attribute product discovery improvements to P&L? Implication: Split attribution across discovery and checkout workstreams; instrument time-to-basket and completion, not vanity engagement. Proof: £8.2M incremental revenue (attributed year measured); 40/45/15 discovery attribution split; +9pp mobile checkout completion; NPS +8 on ease of finding products Concepts: retail, eCommerce, discovery, checkout, attribution **Four hundred thousand products and the right washer still hides on page six.** John Lewis shoppers loved the range across home, fashion, and electronics, but discovery dragged and checkout added friction on the device most visits used. The brief was speed without shrink: keep the catalogue that defines the brand, lose the dead ends that sent people elsewhere. **The Opportunity** — A £15B catalogue where loyalty met filter fatigue John Lewis runs more than 400,000 SKUs online. Loyal shoppers still stalled in category depth: filters failed to surface the right stock quickly, and Quick View stopped short of a confident buy. More than half of visits were on phones; beside Amazon and sharp DTC brands, small friction meant measurable revenue left on the table. My JTBD research identified three pain points: slow filtering on deep pages (smart TVs returning 200+ results with unclear differentiation), Quick View needing multiple clicks for size, delivery, and stock, and checkout asking logged-in customers for details the brand already held. Session recordings showed filter abandonment patterns. Exit surveys flagged 'too hard to compare products' as a top frustration. Usability testing with 80+ participants across mobile and desktop confirmed the same story. Constraints: stock and fulfilment systems could not change, UI-only intervention. Brand guidelines required premium look-and-feel throughout. Success metric: measurable lift in discovery speed and checkout conversion. **The Solution** — Filters with memory, Quick View that finishes the job, PDPs that front-load trust I ran three workstreams in parallel: filtering with live counts and saved combinations, Quick View with size, delivery, and add to basket in the overlay, and product page hierarchy that surfaced key specs earlier. Each workstream had rejected paths. Filters: faceted search with no memory versus saved combinations; I shipped saved combinations after testing showed 35% faster repeat browsing. Quick View: lightweight preview versus full overlay with checkout actions; I shipped the full overlay after A/B tests showed 12% higher add-to-basket from listings. PDP: side-by-side comparison tool versus clearer single-product hierarchy; I shipped clearer hierarchy because comparison required back-end changes I could not take. I designed filters to show live counts before apply. I built saved combinations so returning shoppers resume from 'washers under £400, A++ rated, delivery this week' without rebuilding the query. I designed Quick View to carry size selector, delivery estimate, stock status, and add to basket so confident buyers never left the listing. Product pages moved delivery, returns, and spec tables above the fold on mobile. **Checkout workstream (parallel):** I rejected a bare guest-only path that ignored logged-in shoppers who already held delivery and payment details. I shipped a three-step mobile checkout with logged-in prefill, inline validation on address and payment errors, and a visible order summary before pay, so the second JTBD (checkout speed) had its own rejected alternatives, not only discovery fixes. I validated each stream separately. A/B tests ran at 50K+ sessions per workstream before combining. Instrumented funnels attributed revenue lift to specific changes. **The Impact** — 35% faster discovery, £8.2M incremental revenue, attribution you could defend Filtering and Quick View cut time-to-product by about 35%, measured as median clicks from category entry to add-to-basket. Listing and product page work lifted engagement and basket size. Attribution split cleanly: improved filters contributed approximately 40% of the discovery speed gain, Quick View 45%, clearer PDP hierarchy 15%. The programme landed around £8.2M incremental revenue in the year I measured: discovery work (filters + Quick View) drove the majority via faster time-to-basket. Because I shipped logged-in prefill and inline error recovery on a three-step mobile checkout, mobile checkout completion rose 9 percentage points in the same instrumentation window; that represents the checkout workstream contribution alongside discovery attribution above. Post-launch surveys showed NPS improvement of 8 points for 'ease of finding products'. Takeaways: - Live counts and saved filter combinations cut restart friction on ranges too large to browse blind. - £8.2M incremental revenue with defensible 40/45/15 discovery attribution split. - Checkout prefill and inline errors added +9pp mobile completion in the same window. --- ### Predictable spend forecasting (Vodafone) [View case study](https://www.gaganmalik.io/en/success-stories/vodafone) [Retrieval brief — vodafone-mobile-connectivity] Category: technology Thesis: Plain-language IoT spend forecasts, named cap owners, and docs that match UI vocabulary reduced bill shock and integration friction. Answers: What did Gagan do at Vodafone? | How did Gagan design IoT spend management at Vodafone? | How should finance and ops read IoT cost forecasts? Implication: Forecast in finance language with fleet/site breakdowns; align API and UI vocabulary before scaling integrations. Proof: 58% drop in unexpected spend escalations in two quarters; Cap integration onboarding 4h to 45m; 50M+ users on platform Concepts: IoT, forecasting, developer experience, bill shock **The invoice lands on a Tuesday and three teams chase the same spike.** On Vodafone's IoT platform, cost stayed opaque until billing closed. Finance wanted a forecast. Ops wanted attribution. Developers wanted guardrails in production. By then the period was history. The job was earlier signal: plain-language forecasts, caps with a next step, alerts operators could act on, and docs that matched the UI word for word. **The Opportunity** — Operational spend, rear-view visibility Spikes came from rollouts, misconfigured devices, roaming, firmware shifts, and fleet usage, not one-off bugs. After billing closed you could explain what happened. You still could not reliably forecast the bill, set guardrails early, or catch drift in time to act. Bill shock eroded trust and finance slowed expansion sign-offs. Ops burned time on spikes without clean attribution or a fast path to root cause. When setup and docs were hard to use, automations never shipped and good controls stayed slideware. My JTBD research with 30+ IoT programme owners across automotive, utilities, and logistics identified three needs: finance wanted forecasts grouped by fleet, site, or cost centre, not only API IDs; ops wanted alerts that named actionable slices ('Site 4 roaming overage', not 'threshold exceeded'); developers wanted docs using the same vocabulary as the UI so integrations shipped without round-trips. Billing spike post-mortems and developer onboarding observation confirmed the gaps. Constraints: monthly billing cycles could not change; forecasting had to handle weak signals without false precision; caps needed clear ownership and audit trails. Success upfront: reduce 'billing surprise' escalations by 50% within two quarters. **The Solution** — Plain-language forecasts, caps with ownership, docs that match the API I designed forecasting to answer 'What will we likely spend this period?' in language finance teams already use. Three directions were on the table. Exact predictions with confidence intervals: rejected, because false precision eroded trust when reality diverged. Historical average only: rejected, because it missed rollout and configuration changes. Plain-language forecast ('Likely £12K to £15K this month based on current usage and planned rollouts') with breakdowns by fleet and site: what I shipped. I structured breakdowns to follow how teams already group fleets and sites, not API schema forced onto reporting. Weak signals were labelled clearly ('based on three days of data' versus 'forecast') instead of presented as exact. I mirrored real ownership in caps: warn before a hard stop, show what next when a threshold crossed (pause devices, request increase, review usage), and keep a simple log of who changed what. Anomalies stayed short and actionable: 'Site 4 roaming overage: 240 devices connected to non-home networks', with links to the device list and usage breakdown teams already use. I aligned developer docs vocabulary with the UI: 'spending cap' in both places, not 'Create a spending threshold' in the API and 'Set a cap' on screen. Quickstarts and copy-paste examples cut integration friction. I validated with eight IoT programme teams across verticals, A/B tested forecast phrasing with finance users (plain language versus confidence intervals), and ran developer onboarding sessions. Time to first successful cap integration dropped from 4 hours to 45 minutes. **The Impact** — Steer while the period is still open Teams shifted from 'what happened last month' to what's coming, where caps apply, and what changed. Finance could plan earlier. Ops could act before close. Programme owners could grow fleets with fewer invoice surprises. Plain-language forecasts grouped by fleet and site, cap warnings with named owners, and docs aligned to UI vocabulary ('spending cap' in both places) drove billing escalations tagged 'unexpected spend' down 58% in two quarters, clearing the 50% target. Developer onboarding for cap integration fell from 4 hours to 45 minutes. Takeaways: - 58% drop in 'unexpected spend' escalations — plain forecasts and named cap owners. - Docs↔UI vocabulary match cut cap onboarding from 4h to 45m. - Forecast copy must match how finance groups fleets and sites, not only engineering IDs. --- ### Service requests made faster (Welsh Water) [View case study](https://www.gaganmalik.io/en/success-stories/welsh-water) [Retrieval brief — welsh-water-trusted-field-operations] Category: design Thesis: Structured digital intake with branching triage cut callbacks and routed emergencies without burying them behind billing queues. Answers: What did Gagan do at Welsh Water? | How did Gagan improve Welsh Water service request intake? Implication: Separate civic-emergency intake from billing in UX and routing; capture location and severity on screen one. Proof: Callbacks for incomplete tickets 40% to 11%; 3M+ people served Concepts: utilities, service design, triage, digital intake **Water is pooling on the pavement and the hold music has been playing for nineteen minutes.** Welsh Water serves around three million people [Welsh Water](https://corporate.dwrcymru.com/en/about-us)::pill across most of Wales and adjoining areas. A burst on a main does not wait for a billing query to finish. Neither should the person reporting it. **The Opportunity** — Three million served, one queue for meter reads and mains bursts Welsh Water already operated at full civic scale: storms, bursts on large mains, and heavy capital work were normal operating noise for a supplier serving around three million people [Welsh Water](https://corporate.dwrcymru.com/en/about-us)::pill. Contact centres carried that load alongside everyday customer work. My JTBD research surfaced three triage problems. Most contact was billing and accounts: it deserves a straight answer but does not need a crew. Leaks and supply hits showed up less often in totals but felt urgent when water crossed the pavement or the shower dropped to a trickle. Both reasons landed in the same phone queue. Contact centre call analysis showed about 20 minutes average wait during busy periods: tolerable for a meter question, a poor match when water is leaving the pipe. Field dispatch shadowing revealed 40% of tickets required callbacks to fill location or severity gaps. Customer interviews echoed the same frustration: 'I reported a leak and waited on hold.' Constraints shaped the programme. Self-serve meant guided digital journeys plus robotic process automation into back-office and work-management systems (this predated reliable LLMs). Automation had to integrate with existing trunk repair, sewer inspection, and reinstatement dispatch. Emergencies could not hide behind complex forms. Success upfront: move high-stakes reporting (leak, burst, no water) onto digital intake that captured clean, structured reports automation and dispatch could trust. **The Solution** — Guided intake and RPA, not a chatbot pretending to understand I ran a design sprint with product, operations, and content to agree what to build first: guided flows that walk someone through reporting a leak or another service request, capturing location, severity, anything unsafe on site, and photos when useful, with plain-language next steps on screen. Three options were considered. Generic webform with free-text description: rejected, because it perpetuated ambiguity and callback loops. Chatbot-style conversation: rejected, because the programme predated reliable LLMs and customers needed fast triage, not conversation. Guided journey with branching questions and structured payload: what I shipped. I scoped flows so residents could finish a high-stakes request without starting on the phone. Branching removed ambiguity before dispatch: 'Is water coming from the ground or a building?' routed mains bursts differently from internal leaks; 'Is the leak on a pavement, road, or private property?' determined crew type; 'Is anyone at risk?' (traffic, electrical hazard) triggered emergency escalation. Location capture used postcode search, map pin, or what3words (Welsh Water coverage includes rural areas where postcodes are coarse). Photos were optional but encouraged for severity assessment. Behind the scenes I designed a structured payload (location coordinates, severity tier, risk flags, photo metadata, property type) that rules-based routing and RPA could pass into trunk repair, sewer inspection, and reinstatement queues without agents rewriting fields. The first payload matched how those job types were sequenced in dispatch, not a generic CRM shape. Plain-language copy paired with field-realistic job data so automation had something reliable to act on. I stopped at MVPs on purpose: enough to prototype and user-test, without treating one workshop as a production launch. Forty residents tested prototypes against realistic scenarios ('You see water on the pavement outside your house, report it'). I observed completion rates and confusion points, iterated copy and question order, then piloted with one regional contact centre before national rollout. **The Impact** — Callbacks down from forty per cent to eleven, crews dispatched twenty-eight per cent faster The aim was practical: less ambiguity when a request arrives, fewer callbacks to fill gaps, and dispatch seeing what crews need on the first screen. Agents kept doing what only people should do (vulnerable customer support, complex billing disputes). Automation took predictable steps that had been burning minutes and mis-keys. Because I designed branching intake with structured location, severity, and risk fields on first submit, callback rates on digital leak reports dropped from 40% to 11% and average time from report to crew dispatch shortened 28% for digitally submitted requests. Satisfaction scores and capital programmes shift with regulation and weather; the durable bet was narrower: when intake is structured, people spend less time on hold for emergencies that should never have waited behind billing volume. Takeaways: - Callbacks on digital leak reports 40→11% after branching intake replaced free-text ambiguity. - 28% faster report-to-dispatch when structured payload matched crew job types. - High-stakes reporting needs its own front door, not a longer queue behind billing. --- ### One workspace for team collaboration (Presto) [View case study](https://www.gaganmalik.io/en/success-stories/presto) [Retrieval brief — presto-automated-monetisation] Category: technology Thesis: Workspace AI needs one object graph across boards, docs, and assistant so status, knowledge, and answers share the same truth. Answers: What did Gagan build at Presto? | How should workspace AI share context with Kanban and docs? | Why do assistant pilots fail when chat is separate from deliverables? Implication: Unify projects, tasks, and citations before scaling agent features; measure shared-truth perception across surfaces. Proof: 78% of teams reported board, Spaces, and assistant agreed on state; 42% drop in context-switching in beta Related essays: against-one-click-coding, search-finds-pages-chat-finds-answers Concepts: workspace AI, agent orchestration, object model, SaaS **The assistant opens with 'I can't see that workspace' while the board already shows the sprint.** Presto Platform (askpresto.com) shipped Kanban boards, rich-text Spaces, and an AI assistant with Ask, Plan, and Create. In practice they behaved like three tools on one login: status on the board, knowledge in Spaces, answers in chat that never shared the same objects. Stand-ups became reconciliation. Reporting became screenshots. The job was one governed workspace where context rode with the work. **The Opportunity** — Three surfaces, three truths, one tired programme lead Teams tracked delivery on the board, wrote specs and notes in Spaces or a wiki, and opened generic chat when they needed help. Context did not carry across. Assistant threads were ephemeral. Reporting pulled from the board while decisions lived in Spaces drafts managers stitched together by hand. My JTBD research across 40+ delivery, product, and engineering teams surfaced three integration gaps: board assignments and Spaces pages referenced the same projects but shared no object model; assistant runs disconnected from deliverables; reporting told one story while the knowledge base told another. Workflow shadowing during sprint planning and reviews showed the cost in repetition and mistrust. Competitive analysis of Notion, Linear, Asana, and ClickUp set the bar for what 'integrated' had to mean. Constraints were non-negotiable: role-based access (viewer, editor, admin); assistant responses that cited workspace sources; no breaking existing workspaces mid-rollout. Success upfront: more than 70% of teams reporting that board, Spaces, and assistant agreed on project state. **The Solution** — One object graph for boards, Spaces, reports, and the assistant I designed end to end on a shared shadcn Maia system with role-based access. Keeping separate surfaces with manual links would have perpetuated context switching. Embedding docs inside board cards would have trapped knowledge inside board hierarchy. I built a shared workspace object model where boards, Spaces, and assistant all referenced the same projects, tasks, and people. I structured workspaces and boards to carry assignments, due dates, priorities, and lanes so status stayed on the board, not in spreadsheets. I built Spaces with Notion-style rich text for knowledge and drafts; I designed Create mode and the draft drawer to land AI output on real pages, not ephemeral chat replies. I designed the assistant with Ask (quick questions), Plan (multi-step reasoning), and Create (draft generation), with mentions (@person, @project, @space), workflow templates such as weekly status summary, and models curated per workspace. I designed reporting with day, week, and month views: completed tasks, blocked items, assignment distribution, all from the same board data teams already updated. I put home cards and the composer on the first screen: weekly actions, integration prompts, recent work visible at a glance. I validated in the field, not in a deck. Twenty teams ran a three-month beta. Instrumented context-switching frequency dropped 42% against the prior separate-tools baseline. Alignment perception ('board and Spaces agree on status') climbed from 38% to 76%. **The Impact** — One project name, one status, one place to trust the assistant Teams stopped treating the board, knowledge base, and chat as three competing truths. Fewer reviews opened with a recap nobody had time for. Fewer assistant replies started with a request for context the workspace already held. Managers read movement in reporting instead of stitching a story from side channels and screenshots. Create mode with citations to board tasks and existing pages moved assistant output from discard to ship. Seventy-eight per cent of teams reported that board, Spaces, and assistant shared the same truth, clearing the 70% target and underpinning beta retention and paid-seat expansion conversations (directional; specific revenue figures under NDA). Takeaways: - 78% of teams reported shared truth across board, Spaces, and assistant — above 70% target. - Create mode with citations turned assistant output from discard to ship. - Reporting earns trust when it pulls from the same assignments the board already tracks. --- ### Expert travel guides (Telegraph Travel) [View case study](https://www.gaganmalik.io/en/success-stories/the-telegraph) [Retrieval brief — telegraph-travel-waitrose] Category: design Thesis: Editorial authority wins in travel apps through trip-shaped itineraries, offline use, and expert picks—not scraped listings at scale. Answers: What did Gagan do at Telegraph Travel? | How do media brands compete with Google in travel products? Proof: 130K app downloads first year; 47% second-trip return within six months; 73% rated editorial quality above free alternatives Concepts: media, travel, offline, editorial product **The gate closes in twenty minutes and your signal dies.** Travellers do not reach for another list; they reach for judgement they can trust. Lonely Planet earned that over decades. Google-class trip tools earned it on convenience and scale. The Telegraph Travel Guides app had to earn shelf space beside both: Telegraph judgement, packed for the pocket, readable when roaming is expensive and the next decision cannot wait. **The Opportunity** — Shelf space beside Lonely Planet authority and Google-scale convenience The competitive bar was never 'another travel app.' Readers already held Lonely Planet up for authority and Google-class surfaces for map-first convenience. My job was to show why Telegraph belonged on the home screen when someone lands tired, abroad, and has thirty minutes to decide what to see. My JTBD research surfaced three non-negotiables: trusted picks (shortlisted hotels, honest walk routes, food writing with a point of view, not scraped filler), usable in transit (offline reading, maps that work without signal, quick decisions on a small screen), and a credible reason to choose Telegraph over brands they already know. I ran traveller interviews in airports and hotels at real decision moments, tore down Lonely Planet, Google Maps Explore, Tripadvisor, and Airbnb Experiences, and surveyed Telegraph Travel readers: 68% wanted expert picks in app form. Constraints were tight. Content had to come from Telegraph Travel journalism, not scraped listings. The app had to work offline. Monetisation could not trash the brand with intrusive ads. Success upfront: more than 40% of active users returning for a second trip within six months. **The Solution** — Twenty-nine deep guides, not a database pretending to be editorial I put Telegraph Travel judgement into the product structure itself: authored picks, trip-shaped itineraries, and maps you could trust without signal. Three directions were on the table. User-generated content with Telegraph curation: rejected, because it dilutes the authority readers pay for. A full destination database with Telegraph reviews layered on: rejected, scope and maintenance would bury the desk. Twenty-nine destinations [Telegraph 2015](https://www.telegraph.co.uk/travel/news/Telegraph-Travel-Guides-app/)::pill at launch with deep Telegraph coverage per city: what I shipped. I paired content and UX craft in the same room. I structured destinations and days out from Telegraph journalism: expert hotel picks, restaurant shortlists, walk routes, sights, each with timing, transport, and booking links where it helped. I shaped itineraries around how people actually move ('One day in Rome: ancient and modern', 'Weekend in Barcelona: Gaudí trail'), not abstract bucket lists. I designed maps and saved guides to work offline so navigation did not depend on roaming. I kept copy Telegraph Travel throughout: honest, opinionated, written for readers who already know a city has more than one cathedral. I made positioning deliberate: Telegraph authority and tested routes against Lonely Planet breadth; Telegraph curation against Google scale where everything is listed and nothing is chosen. I led app store messaging with expert-led, offline, no ads. I validated in the field, not in a lab. Sixty travellers used prototypes during real trips while I watched in-situ behaviour. I A/B tested itinerary formats (step-by-step versus overview-first). Instrumentation showed 82% saving at least one guide for offline use, which confirmed offline was not a nice-to-have. **The Impact** — Judgement that survives the second trip The Guides line turned reputation into a planning ritual travellers could pack. Downloads reached 130K [Telegraph 2015](https://www.telegraph.co.uk/travel/news/Telegraph-Travel-Guides-app/)::pill in the first year post-launch. Telegraph Travel reached 1.1M readers across web and app surfaces. Itineraries structured as trip-shaped day plans with offline save drove 47% of active users back for a second trip within six months, clearing the 40% target. Seventy-three per cent rated editorial quality above free alternatives; readers who saved offline itineraries were more likely to reopen the app for a second destination, linking the structure to repeat use rather than one-off downloads. Takeaways: - 47% returned for a second trip within six months — offline itineraries drove repeat use. - 130K downloads year one; 29 deep destinations beat a scraped database. - Against Lonely Planet and Google, editorial IP has to live in structure and copy. --- ### Loyalty for food lovers (Waitrose) [View case study](https://www.gaganmalik.io/en/success-stories/waitrose) [Retrieval brief — waitrose-loyalty-food-lovers] Category: business Thesis: Premium loyalty competes on food identity and transparent personalised value, not points dashboards that commoditise the brand. Answers: What did Gagan do at Waitrose My Waitrose? | How should premium retailers design loyalty without Clubcard-style points? Proof: 4.6M active My Waitrose members; Offer clarity 42% to 68% after personalised offer cards; +12% repeat visits among active members vs non-members Concepts: loyalty, retail, personalisation, premium brand **The card is in your hand and the offer on screen makes no sense.** Four and a half million people scan My Waitrose every week for coffee, treats, and the feeling that Waitrose knows what they cook. The Partnership had scale on its side. The experience still had to earn the next visit before someone gave up at the till. **The Opportunity** — Seven per cent more members on a food brand that cannot win on points alone John Lewis Partnership's 2024/25 annual report puts the scale in plain sight: My Waitrose grew 7% to 4.6 million active members [JLP report 2025](https://www.johnlewispartnership.co.uk/~/media/Files/J/john-lewis/corp/documents/john-lewis-plc-ara-24-25-signed.pdf)::pill while more customers shopped with the Partnership's brands. Waitrose sales reached £8.0 billion [JLP report 2025](https://www.johnlewispartnership.co.uk/~/media/Files/J/john-lewis/corp/documents/john-lewis-plc-ara-24-25-signed.pdf)::pill. The business invested £150 million in lower prices [JLP report 2025](https://www.johnlewispartnership.co.uk/~/media/Files/J/john-lewis/corp/documents/john-lewis-plc-ara-24-25-signed.pdf)::pill since 2023, relaunched Waitrose No.1, refurbished stores, and pushed on-demand grocery sales up 110% [JLP report 2025](https://www.johnlewispartnership.co.uk/~/media/Files/J/john-lewis/corp/documents/john-lewis-plc-ara-24-25-signed.pdf)::pill. The experience bet sat underneath those headlines. My JTBD research surfaced three shopper missions: everyday value (free coffee, treats, savings people grasp before the till), food discovery (recipes, seasonal ranges, counters, member events), and transparent personalisation (offers tied to purchase history without opaque targeting). Member surveys showed 73% valued free coffee and treats but only 42% understood how personalised offers worked. Exit interviews named 'too complicated to use offers' as a top barrier. Service design mapping traced the friction across online, app, and in-store touchpoints. Constraints were non-negotiable: offers had to work across every channel without Partners learning a second script; personalisation had to comply with data privacy and feel explainable; the value story had to stay premium. Success upfront: active member growth above 5% annually, with offer redemption rates improving. **The Solution** — Free club for food lovers, not another points race My Waitrose works when membership connects three loops, not when it behaves like a bolt-on scheme. Three directions were on the table. Points-based loyalty like Tesco Clubcard: rejected, because it commoditises the brand and competes on scale Waitrose cannot win. A points dashboard UI: rejected, because it trained shoppers to chase balances instead of food identity. Premium subscription with an upfront fee like Amazon Prime: rejected, because it adds a barrier at entry and does not match Partnership values. Free membership with everyday rewards [My Waitrose](https://www.waitrose.com/ecom/mywaitrose/become-a-member)::pill, food identity, and transparent personalised value: what I shipped. I designed everyday reward to keep the reasons to scan visible: free tea or coffee, treats with a hot drink, selected savings (20% off wine, seasonal produce), and offers people understand before they reach the till. I wrote expiry, usage ('Scan your card at checkout'), and terms in plain language so Partners and customers resolved questions without manager calls. I turned membership into a food-lover club: recipe inspiration tied to seasonal ranges, counter highlights (fishmonger picks, cheese pairings), Waitrose No.1 discovery, member tastings, cooking classes, supplier stories, and competitions. Communications led with food ('This week's seasonal star: British asparagus'), not generic retail ('Save 15% on vegetables'). I made personalised value and relevance legible. Offers explained why they appeared ('Based on your purchases: 10% off organic chicken this week'), when they expired (countdown in app and email), and how to use them (online code, scan in store, automatic at till). Customers could see offer history and opt out of categories. I grounded validation: service design blueprints mapped touchpoints across coffee run, top-up shop, weekly shop, and entertaining missions; prototypes tested offer clarity with 120 members across age groups; pilot stores instrumented redemption and satisfaction before national rollout. **The Impact** — 4.6 million members, clearer offers, twelve per cent more repeat visits Post-launch member surveys showed offer clarity rising from 42% to 68% ('I understand why I received this offer') after the personalised offer card pattern shipped with explicit why/expiry/how copy. Repeat visit frequency increased 12% among active members versus non-members in the same measurement window: the loyalty-attributable behaviour shift I can defend separately from headline Partnership P&L. Offer cards with explicit why, expiry, and how-to-use copy moved offer clarity from 42% to 68% in member surveys. Repeat visits rose 12% among active members against non-members in the same window. My Waitrose reached 4.6 million active members [JLP report 2025](https://www.johnlewispartnership.co.uk/~/media/Files/J/john-lewis/corp/documents/john-lewis-plc-ara-24-25-signed.pdf)::pill, up 7%, within broader Partnership growth (Waitrose sales £8.0B, operating profit +£122M: context, not loyalty-only attribution). Takeaways: - Offer clarity 42→68% after personalised offer cards with why/expiry/how copy. - +12% repeat visits among active members — loyalty-attributable, separate from P&L. - Rejected points-dashboard UI; food identity and transparent value beat balance chasing. --- ### Better team collaboration (Eutelsat OneWeb) [View case study](https://www.gaganmalik.io/en/success-stories/eutelsat-oneweb) [Retrieval brief — eutelsat-oneweb-engineering-docs-workflow] Category: design Thesis: Template-led engineering intake, named review owners, and live document state cut version drift when launch windows do not wait. Answers: What did Gagan do at Eutelsat OneWeb? | How does design ops help satellite engineering documentation? Proof: Pack completeness 61% to 89% with template intake; 38% fewer review round trips in pilot Concepts: design operations, documentation workflow, satellite, review routing **Launch windows do not wait while two disciplines cite different revisions.** Satellite programmes pull together spacecraft, ground sites, suppliers, and regulators. When iteration speeds up, parallel changes multiply. Signatures wait on the wrong pack. Programme leads pull status from mail instead of one clear picture. The job was collaboration that matched how satellite work actually happens: visible owners, live state, audit trails reviewers could trust. **The Opportunity** — Strong engineering, fuzzy version truth Eutelsat OneWeb's published fleet counts run to [650 satellites](https://www.eutelsat.com/system/files/2025-10/DOC_Group_URD-Eutelsat-2024-25_EN_301025.pdf)::pill in low Earth orbit, [70 teleports](https://www.eutelsat.com/satellite-network/geo-leo-multi-orbit-satellite-network)::pill on the ground, and [510 engineers](https://www.eutelsat.com/system/files/2025-10/DOC_Group_URD-Eutelsat-2024-25_EN_301025.pdf)::pill in the workforce figures Eutelsat reports for FY 2024–25. The painful pattern was familiar: two disciplines referencing different revisions, signatures waiting on the wrong pack, programme leads chasing answers across threads instead of reading one status view. My JTBD research identified three workflow gaps: new specs and change packs started from blank folders with no standard structure; review routing was opaque (who owns the next step?); document state scattered across email, shared drives, and PLM with no single source of truth. Workflow shadowing across three satellite programmes, stakeholder interviews with systems engineers, programme managers, and quality leads, and an audit of rejection reasons at review gates showed 62% were 'incomplete pack' or 'wrong revision cited'. Constraints: existing PLM and document management systems were not changing (workflow had to integrate, not replace); specialists could not be forced through a big-bang tool swap during launch windows; regulatory compliance required specific review checkpoints. Success upfront: reduce review round-trips at integration and acceptance gates by at least 30%. **The Solution** — Templates, routing, and live status beside the repositories teams already used I shaped automation around real programme rhythms: structured starting points, routing that respected roles and boundaries, review checkpoints only where policy required them. Three directions were on the table. New centralised document repository to replace existing PLM and shared drives: rejected, because migration mid-programme was unrealistic and specialists resisted tool swaps. Email-based workflow with manual tracking: rejected, because it perpetuated the opacity problem. Lightweight automation layer integrating with existing repositories and surfacing metadata and hand-offs: what I shipped. I designed template-led intake with named sections (scope, interfaces, test evidence, sign-off checklist), required fields that blocked submission until complete, and standard revision naming. The routing view showed live states (Draft, In review, Changes requested, Approved) with a named owner label at each gate (author, discipline lead, systems engineer, quality reviewer, programme approval). Notifications stayed short and pointed to the live artefact ('Spec 4201 Rev C ready for your review' with direct link). Hooks into existing repositories so status reflected what people actually filed: when a document moved to the PLM system of record, workflow metadata updated automatically. I proved each flow with a pilot programme first (one satellite build stream), instrumented review round-trips and time-to-gate-approval, then widened stream by stream. Pilot showed 38% reduction in review round-trips and 22% faster time through integration gate; scaled to five additional programmes over six months. **The Impact** — Less reconciliation, more engineering judgement Teams spent less time debating which file counted and more time on the technical decision itself. Template-led intake with required sections and named owners at each review gate raised pack completeness from 61% to 89%. That shift cut review round-trips 38% in pilot and dropped gate cycle time 34% across programmes using the automation. Programme leads could scan document state at a glance instead of chasing answers across threads. Post-pilot surveys showed 82% of engineering leads reporting 'workflow reduced reconciliation overhead'. The lasting shift was cultural: paperwork treated as part of the mission rhythm, not an admin sideshow after the real work was done. Takeaways: - Pack completeness 61→89% after template intake with required fields and named gate owners. - 38% fewer review round-trips in pilot; 34% faster gate cycle time at scale. - Automation sticks when it sits next to existing repositories instead of inventing a new source of truth overnight. --- --- ## Newsroom (writings & updates) ### The Right to Wake You Has to Be Earned [Read more](https://www.gaganmalik.io/en/newsroom/the-right-to-wake-you-has-to-be-earned) [Retrieval brief — the-right-to-wake-you-has-to-be-earned] Category: design Thesis: Respecting silent mode is the default of good product design; a life-safety override is legitimate only while it stays scarce, geographically bounded, and about danger to life. Answers: When is it legitimate for a product or a state to override Do Not Disturb or silent mode? | Are UK Emergency Alerts push notifications, and why do they bypass silent mode? | How do UK Emergency Alerts sit in citizen-service UX next to GOV.UK Notify and flood warnings? | How should product teams think about alert fatigue versus a genuine emergency interrupt? | What did the UK wildfire Emergency Alert on 14 August 2026 actually say, and who was it for? | How does Gagan Malik connect notification design to attention control? Implication: Do not train people to swipe government or product sirens. Keep forced interrupts rare, bounded, and tied to danger to life. Growth re-engagement copy should never borrow that licence. Proof: Author lock-screen photograph, EE iPhone, Friday 14 August 2026, 19:02, UK Severe Alert card; same alert again around 22:40 (Gagan Malik); GOV.UK About Emergency Alerts (mast broadcast, silent override, about ten seconds, life-threatening use); Apple Support, Government Alerts on iPhone; Apple Support, Government Alerts on Apple Watch; BBC reporting on Stourbridge fires and Jane Baxter evacuation; The Independent, 14 August 2026, Louise Haigh on the most extensive active-incident use; Plymouth February 2024 first live UK alert (Wikipedia, Emergency Alerts United Kingdom); Cabinet Office 19 March 2023 launch: modelled on US, Canada, Japan, Netherlands; The Guardian, 21 April 2023: Covid SMS took more than 48 hours, cell broadcast chosen; GOV.UK Notify: transactional email, text, and letters to users who gave a number; GOV.UK, Get flood warnings by text, phone or email (Environment Agency, opt-in); Government Design Principles (GDS): start with user needs, do less, understand context; The Atlantic, Zoë Schlanger, July 2024: Wireless Emergency Alerts and alert fatigue Related essays: design-is-attention-control-not-aesthetics, waiting-is-a-design-surface, why-attention-is-the-most-consumed-drug, someone-designed-that-button Concepts: emergency alerts, silent mode, notification design, alert fatigue, attention control, UK Emergency Alerts By Gagan Malik Friday 14 August 2026. It came twice. First at 19:02: the lock screen still doing weather, thirty-two degrees, mostly sunny, a yellow triangle the only colour on a grey card. The phone treated it like a call I had not agreed to take. Around 22:40 it came again. Same triangle. Same government. Same override of a silence I had already chosen. Author photograph: UK Severe Alert on an EE iPhone lock screen, Friday 14 August 2026, 19:02. I design products that are supposed to honour that refusal. Silent mode is a preference stated in hardware. Respecting it is the default of good design, not a courtesy. The card that overrode it was a Severe Alert issued by the UK Government to England and Wales. Very high wildfire risk. No disposable barbecues, firepits, garden incinerators, or fireworks. Call 999. Search gov.uk/alerts. I have already argued that design decides where attention lands, not how the screen looks, in [Design Is Attention Control, Not Aesthetics](https://www.gaganmalik.io/en/newsroom/design-is-attention-control-not-aesthetics). This was that decision with the volume forced up. The right to wake you has to be earned. The override stays legitimate only while it is scarce, bounded, and about danger to life. #### Silent Mode Exists Because Almost Every Interrupt Is a Lie Most badges on your lock screen are not trying to keep you alive. They are trying to get the thumb back. I have been paid to write the second kind. Re-engage lapsed users. A/B the copy. Put the winner live. The person who was asleep did not. That work sells itself as care. It is extraction wearing a red dot. Silent mode exists because we trained people to need it. The honest product default is to stay outside the room until you are invited. Growth teams hate that sentence because invitation does not graph. I still believe it. I have also cashed the cheque that ignored it. #### This Was Not a Push Notification The UK system is a broadcast from the masts around you. Government pages are blunt about what that means. They do not need your number. They do not need you to have location services on. Your phone may siren even if it is on silent. It may vibrate, and it may read the message out, for about ten seconds. [gov-uk-alerts](https://www.gov.uk/alerts) Apple's iPhone note matches: an alarm-like sound for government alerts, on by default. There is a switch for sound in Silent mode, and another if you want the class off. [apple-gov-alerts-iphone](https://support.apple.com/en-gb/102516) The Watch follows the phone. Turn the class on for iPhone and it is on for Watch. A cellular Watch can still get the siren when the phone is not in the room. [apple-gov-alerts-watch](https://support.apple.com/en-gb/111817) So the override is policy. You can leave. The designers ask you not to. Honouring silent mode is still the correct product. A life-safety siren is the only interrupt that should be allowed to lose that argument on purpose. Someone near a mast may need those ten seconds more than an undisturbed evening. #### This Is the Loud End of Citizen UX Most of government does not get to wake you. This card does. Citizen services already have a quiet pattern. GOV.UK Notify sends email, text, and letters because a service has your number: an MOT, a court date, an NHS appointment. You invited that knock. Silent mode still holds. [gov-uk-notify](https://www.notifications.service.gov.uk/) Flood warnings are the same family. You sign up. The Environment Agency texts or calls a place you said was yours. [gov-uk-flood-warnings](https://www.gov.uk/sign-up-for-flood-warnings) The Government Design Principles sit under both: start with user needs, do less, understand context. A person asleep is the context. Government should only do what only government can do. [gov-uk-design-principles](https://www.gov.uk/guidance/government-design-principles) Emergency Alerts are the other pattern. Cell broadcast. No number. Silent override. When the Cabinet Office switched the system on in March 2023, it named the countries it copied: the United States, Canada, Japan, and the Netherlands. Flood, fire, severe weather. [gov-uk-alerts-launch](https://www.gov.uk/government/news/launch-of-life-saving-public-emergency-alerts) Covid SMS proved the quiet pattern cannot scale to a country in danger. The networks queued texts by number. It took more than forty-eight hours, and scammers copied the tone. [guardian-alerts-years](https://www.theguardian.com/technology/2023/apr/21/how-the-uk-emergency-alert-system-test-has-been-years-in-the-making) America's Wireless Emergency Alerts show the failure mode. If the siren is used for road closures and statewide Amber noise, people treat the next climate warning as spam. [atlantic-climate-alerts](https://www.theatlantic.com/science/archive/2024/07/climate-push-alert-emergency-warning/678936/) Friday's triangle is only legitimate if it stays the loud exception in a stack that is otherwise polite. #### Nineteen Houses Were Already Gone By the time the card landed on Friday evening, Thursday had already taken a street. In Stourbridge, West Midlands Fire Service said nineteen houses were fully destroyed and eighteen more damaged. A grass fire near the golf club had run into homes. [bbc-west-midlands-fires](https://www.bbc.com/news/articles/cp8xr2n9560o) Jane Baxter's house backs onto that club. She told BBC Radio 4 that police knocked. The fire was changing direction. Grab what you need. She left with the cat, to her son's. [bbc-west-midlands-fires](https://www.bbc.com/news/articles/cp8xr2n9560o) That is the human the tally never has to name. I kept the heat, the weather widgets, and an evening still intact at 19:01. She kept a cat and a son's spare room. The Independent put the send at around 7pm across England and Wales. It also noted more than a thousand wildfires in the country this year. [independent-emergency-alert](https://www.independent.co.uk/news/uk/politics/emergency-alert-wildfires-uk-wales-warning-b3033354.html) Those thousand are a tally. Jane is the cost the tally is for. #### The Card Said the Next Thing to Do The card used verbs you can do with a match or a tap. I am not going to pretend the yellow triangle was a coded severity ladder. I only know what the photograph shows: it was the one mark that was not grey. Above it, the widgets still said thirty-two and mostly sunny. Heat sitting over a fire warning is the weather the warning is about. #### A Horn That Never Stops Becomes Weather A car horn is built to violate the street. Used once, your shoulders jump before you know why. Used all afternoon in a jammed junction, the same sound turns into climate. You stop flinching. You start hating the person who still believes the noise means something. The yellow triangle on my lock screen was trying to be the first horn, not the fiftieth. The body is the instrument here. If that triangle becomes weekly weather, it will arrive to a person who already swiped. #### The Charge That This Is Just a National Nanny I cannot start a wildfire from a third-floor kitchen. A national interrupt about disposable barbecues, firepits, and fireworks will land on millions of people who were never going to light one. Once a government may wake a country for a list of chores, the next list will be softer and the siren will be the same. Alert fatigue is what happens when every ping claims to be the fire. Rarity is a promise. Promises get spent. If you wanted a clean ethics rule, it would be this: no product, and no state, ever overrides the silence a person already chose. Hold that. Then look at what the send was actually for. First Secretary Louise Haigh said the government had issued an Emergency Alert to England and Wales because of continued wildfire risk. She called it the most extensive use of the UK system in response to an active incident. [independent-emergency-alert](https://www.independent.co.uk/news/uk/politics/emergency-alert-wildfires-uk-wales-warning-b3033354.html) The system's own pages limit the job to life-threatening emergencies: wildfires, severe flooding, extreme storms. [gov-uk-alerts](https://www.gov.uk/alerts) In February 2024 it told people in Plymouth to move because of an unexploded wartime bomb. [wikipedia-uk-emergency-alerts](https://en.wikipedia.org/wiki/Emergency_Alerts_(United_Kingdom)) Since the first national test in April 2023, a government education note counted five real alerts during major storms when lives were at risk. [gov-uk-education-hub-alerts](https://educationhub.blog.gov.uk/2025/09/what-you-need-to-know-about-the-emergency-alerts-test-on-7-september/) That is still a short list. Jane Baxter did not need a barbecue memo. She needed a knock. Friday's card was for the next person with a match in a field, and for the crews walking toward it. The nanny charge stays true as a warning about a later summer. #### The Licence Is the Rarity Silent mode is still the right default, and this siren is the exception only while it stays scarce. Spend that sound on anything that is not danger to life, and the next card is spam. Jane Baxter left with a cat because someone knocked; the rest of us got the triangle twice so the next street does not need that knock. --- ### A Trial Frame Is Not a Prescription [Read more](https://www.gaganmalik.io/en/newsroom/a-trial-frame-is-not-a-prescription) [Retrieval brief — a-trial-frame-is-not-a-prescription] Category: design Thesis: A trial frame lets an optometrist swap lenses in front of a patient's eyes before anyone signs a prescription; a locked product vision shipped before a team has tried the alternative lenses is the same category error, dressed as conviction instead of diagnosis. Answers: When has a product team actually tested a vision, versus tried it once and treated it as final? | What is the difference between a design prototype and a locked product spec? | Does "disagree and commit" justify skipping user testing before shipping a vision? | How much user testing counts as a genuine test before a team commits to a direction? | Why do polished mockups get treated as final decisions before anyone has used them? | What does optometry's trial frame teach product teams about testing before committing? Implication: Treat the first shipped vision as a trial frame, not a prescription. Test with a small number of real users (five is a reasonable minimum, per Nielsen and Landauer) before locking a direction, and build in a scheduled retest the way optometry mandates a re-examination, rather than shipping a first guess and defending it indefinitely. Proof: Jake Knapp / Google Ventures Design Sprint methodology (gv.com/sprint); Jakob Nielsen and Thomas Landauer, five-user usability testing model (Nielsen Norman Group, 2000); Jeff Bezos, 2016 letter to Amazon shareholders ("disagree and commit"); College of Optometrists, UK guidance on re-examination intervals Related essays: familiarity-is-a-floor-not-a-ceiling, show-the-job-not-the-feature, waiting-is-a-design-surface, there-is-perfection-in-imperfection, optimise-is-maximise-with-better-manners Concepts: product vision, design prototyping, usability testing, disagree-and-commit, design sprint, trial frame, premature commitment By Gagan Malik An optometrist has a trick for making you doubt your own eyes. She lowers a metal frame in front of your face and drops a loose lens into the front slot. Then she asks a question with no correct answer: better, or worse? You squint at the chart. She swaps the lens for its neighbour, half a dioptre stronger. Better, or worse? You are not being tested once. You are being triangulated, one guess at a time, towards a number neither of you knew when you sat down. That metal frame is called a trial frame. It holds nothing permanent: the lenses come out as easily as they went in, and nobody signs anything until the chart stops changing. A prescription is the opposite kind of object. It is a decision, dated and printed, that says this combination of glass is correct until further notice. Most product teams skip the trial frame entirely. They put one lens in front of the eye, call it the vision, and print it as a roadmap. Then they defend it in every meeting that follows, as though the chart had already stopped changing. #### Why Locking Early Feels Like Leadership A founder who says here is the vision gets funded. A founder who says we are still trying three lenses gets asked when the deck will be ready. Investors, engineering leads, and marketing calendars all want the same thing: a fixed target they can build a schedule against. A locked vision is cheaper to plan around than an honest one. So the room rewards conviction over correction long before anyone has tried a second lens. That pressure has a shape in the tools themselves. A rough sketch invites debate. A polished mockup, built in the brand's real typeface with real copy in every field, reads like a decision that has already been made. Once that file lands in a board deck, reopening it costs someone's credibility, not just an afternoon. The file was built to be tested. The room treats it as signed the moment it is shared. #### Prototypes Are Meant to Be Thrown Away Jake Knapp designed the opposite discipline at Google Ventures. He called it the [Design Sprint](https://www.gv.com/sprint/): four days building a fake good enough to put in front of five real customers on day five. The whole approach runs on one rule: a prototype is disposable, built just enough to learn and never more. A team that has not yet watched someone use the thing has no business defending it. Here is the tension, held at full strength. The sprint's own facade can calcify the moment it survives Friday's test with a shrug instead of a verdict. Ambiguous feedback gets read as permission. The facade gets promoted to final design overnight. The team that spent four days building something disposable spends the next four months defending it, as though Friday had settled the matter for good. #### What Farah Watched Get Signed Off Farah ran her first and only round of user testing on a checkout redesign on a Thursday afternoon. Three of five users hesitated at the same step, unsure whether their new address had actually saved. She wrote it up that evening. The stakeholder deck, built from the same screens two days earlier, had already gone to the board on Tuesday. It carried a sign-off line under the mockup that nobody had thought to leave blank. Her finding was filed as an item for version two, a version the roadmap never funded. I have done the same thing to my own testers when the room needed to move. I heard the hesitation, logged it politely, and shipped the lens I already had in the frame. Nobody lied in that meeting. Everyone had simply agreed, without saying so, that Thursday's test was theatre and Tuesday's deck was the diagnosis. #### Nobody Frames the Trial Lenses Nobody frames the trial lenses. They go back into their case within the hour: no ceremony, no signature. They are ready to be tried again on the next patient who sits in that chair. What eventually gets framed is the finished pair: engraved with a date, handed over as though the question were now closed. Even that closed answer has an expiry. UK guidance from the [College of Optometrists](https://www.college-optometrists.org/clinical-guidance/guidance/knowledge,-skills-and-performance/the-routine-eye-examination/frequency-of-eye-examinations) tells practitioners not to go longer than two years before re-examining an adult. The eye that was measured correctly in one sitting drifts, and a lens that was right then is not guaranteed to be right now. A shipped product vision rarely carries that clause. Nobody schedules the retest. The lens that was right at launch stays on the face for years after the eye it was measured for has changed. #### Give the Objection Its Full Weight Give the objection its full weight. Jeff Bezos built disagree and commit for exactly this problem. In a [2016 letter to Amazon shareholders](https://www.aboutamazon.com/news/company-news/2016-letter-to-shareholders), he described greenlighting a show he personally doubted. By that point in a genuine disagreement, he wrote, nobody can know the answer for sure. Further argument only spends time nobody gets back. A team that tries a hundred lenses and never signs a prescription has not escaped the mistake this essay describes. It has built a slower version of the same one. Read the letter again and the difference holds. Bezos committed after a genuine disagreement had already been aired, not before anyone had looked through a second lens. Jakob Nielsen wrote the familiarity law that gets misquoted in design reviews, the one I examined in [Familiarity Is a Floor, Not a Ceiling](https://www.gaganmalik.io/en/newsroom/familiarity-is-a-floor-not-a-ceiling). He also found, with Thomas Landauer, that testing a design on five real users catches roughly eighty-five per cent of its usability problems. [Nielsen Norman Group](https://www.nngroup.com/articles/why-you-only-need-to-test-with-5-users/) still publishes that finding as guidance. That is a number small enough to fit in one Thursday afternoon. Commit after that Thursday and you are disagreeing and committing. Commit before it and you are writing a prescription off the first lens you happened to try. #### The Lens Nobody Swapped A trial frame lets you find out which lens is correct before anyone signs anything. A prescription is what you write after that finding, not a substitute for having looked. Farah's three hesitant users are still squinting at the first lens nobody let her swap. --- ### Where Research Dies. [Read more](https://www.gaganmalik.io/en/newsroom/where-research-dies) [Retrieval brief — where-research-dies] Category: design Thesis: A research finding only survives past the readout when it becomes a constraint a later production stage must pay to satisfy, the way Titan Eye+ and Ogilvy India's Eye Test Menu required scouting and sizing every highway dhaba wall; a finding left as a message anyone downstream can ignore for free dies between the readout and the build. Answers: Why do most user research findings never change what actually gets built? | What made the Titan Eye+ Eye Test Menu campaign work as a road-safety intervention rather than another awareness poster? | What is the practical difference between a research finding that becomes a message and one that becomes a constraint? | How can a leader tell, before approving a research readout, whether a finding will survive into production? | Why did India's public eyecare awareness campaigns and its commercial-licence eye test both fail to get drivers to act on poor vision? | Is the Eye Test Menu campaign's claim that its menus functioned as a real vision test independently verified? Implication: Before signing off a research readout, ask what specific thing downstream cannot proceed unless the finding is satisfied, not what message will be put in front of the user. If nothing has to change shape, cost more, or get measured with a tape, treat the finding as decoration. Proof: Titan Eye+ and Ogilvy India, Eye Test Menu, One Show 2025 Bronze Pencil, D&AD 2025 Wood Pencil, Kyoorius 2025 Blue Elephant; VisionSpring with Queen's University Belfast, 9,857 commercial drivers screened, Optometry Times August 2025; ThePrint, 1 June 2022, on pre-existing prevalence and licence-renewal eye checks Related essays: someone-designed-that-button, the-unemployment-rate-is-fine-that-is-the-trap, there-is-perfection-in-imperfection, the-desk-she-could-not-delete Concepts: user research, behavioural design, findings versus constraints, research operations, road safety, awareness campaigns, production constraints, India By Gagan Malik At Ganesh Kripa dhaba, off the Mumbai highway, the old menus are gone. Tables cleared. Walls cleared. The only way to order food now is a board printed in type that shrinks as it descends: chai you can read from the car park, lines near the bottom that force a man to stand up and walk closer. For a lot of the drivers who stop here, somewhere on that descent the dishes stop being dishes and turn into a smudge. [One Show](https://www.oneclub.org/awards/theoneshow/-award/58448/eye-test-menu/) Eye Test Menu: usability notes. Menu art: Ogilvy / Titan Eye+. Annotations: author. That object won a Bronze Pencil at the [One Show](https://www.oneclub.org/awards/theoneshow/-award/58448/eye-test-menu/) and a Wood Pencil at [D&AD](https://www.dandad.org/awards/professional/2025/239672/eye-test-menu/), and Titan says it took two Bronze Lions at Cannes. [Titan](https://x.com/TitanCompanyLtd/status/1935605237227856081) The prizes are for the idea, and the idea is the least useful thing about it. What is worth taking is duller. A finding about how these men think about their own bodies had to survive five handovers: strategy, optometry, design, a location scout, a print run. It came out at the far end intact, in the form of a measurement. Almost nothing survives that trip. #### Everybody already knew the drivers could not see The vision statistic was not the discovery. ThePrint reported it in June 2022, more than two years before any of this went up on a wall. One manufacturer's eyecare camp at Delhi airport had found one in three drivers needing urgent correction. Another found half its drivers needed glasses, with five per cent showing the onset of cataract. In the same piece, Anshu Taneja of VisionSpring named the real barrier. It is not affordability. A basic pair of prescription spectacles costs three or four hundred rupees. Drivers go without because many of them feel glasses are a sign of weakness. [ThePrint](https://theprint.in/opinion/dashboard/indian-drivers-are-experiencing-a-myopia-epidemic-large-numbers-just-cant-see-properly/978244/) So the problem was documented and the fix was cheap. The law was already on the case as well. Renewing a commercial licence in India requires a doctor to certify eye health, which that same reporting calls an eyewash. Two instruments were running at national scale, and between them they reached almost everyone and moved almost nobody. Awareness gets chosen as often as it does because it costs nothing to anyone in the chain. A poster, a press day, a line on a renewal form. Nobody downstream has to change what they are building, and nobody has to take a bigger number to finance. #### The finding was not about eyes Read the campaign's own account and the useful finding is sitting there, phrased almost as an aside. One in three drivers has compromised vision. The same drivers do not consider impaired vision serious, find clinical communication intimidating, and put an eye test near the bottom of a list headed by a day's income. [One Show](https://www.oneclub.org/awards/theoneshow/-award/58448/eye-test-menu/) Hold both halves at full strength and they pull against each other. The more urgently you need someone to act, the less a message can do for you. A message asks a person to notice something, and the thing you need noticed is what their own perception is busy covering up. What was done with that is the whole case. Nobody wrote a better message. Ogilvy's Bengaluru and Mumbai teams went to the dhabas instead, where drivers already stop to eat and talk and get an hour off the road. [One Show](https://www.oneclub.org/awards/theoneshow/-award/58448/eye-test-menu/) The old menus came off the tables and the walls, and an eye test went up in their place. [Ogilvy](https://www.ogilvy.com/ap/work/eye-test-menu) Then it was made the only way to order food. Nobody had to be talked into a test, because nobody was taking a test. They were ordering dinner, and finding out. #### Somebody had to measure the wall This is the part I would put in front of anyone who runs research for a living. Dhabas have no standard menu: each serves its own regional dishes, listed in the language of that state. So the menus were built one at a time, per dhaba and per language, and the layout still had to work as a real eye test. That meant scouting every location beforehand. Someone had to settle how large the menu should be and where on the wall it should sit. The type sizes had to test something at the distance a person actually reads from. Optometrists were consulted on the design and were there on the day to check the results. [One Show](https://www.oneclub.org/awards/theoneshow/-award/58448/eye-test-menu/) That work has names on it: Agastya Manmohan is credited with the art direction and Timeslider with running the camps on the ground. [Kyoorius](https://kyooriuscreative2025.awardsengine.com/?action=ows:entries.details&e=182119&project_year=2025) Both the scouting and the test's validity rest on the campaign's own account. No optometrist outside it has confirmed the menu works as a clinical test rather than a prompt to look closer. Nor has Timeslider independently confirmed how the sizing was actually done. That is what a finding looks like when it survives. A message costs nothing downstream and can be ignored by everyone who touches it after you. A constraint has to be satisfied, so somebody's Tuesday gets harder and somebody's budget gets bigger. That is exactly why it stays in the work. I cannot prove the general rule from one campaign and I am not going to pretend otherwise. But I can tell you which of my own findings have died, and not one of them ever made anybody measure anything. #### I have killed a finding this way I grew up in Delhi, in North India, with poor eyesight from my teens. Admitting it was treated as a sign of weakness. Household remedies came first; a proper examination came late, if it came at all. That was lived cultural experience, not a finding I had to be talked into. I have also been on the other side of the table. I have run UX field studies, sat in the interviews myself, and packed uncomfortable findings into detailed decks that everyone praised. Then nothing in the product changed. The room had agreed. The slides had landed. The build went on without them. #### The strongest case against this, properly stated Let me state the strongest case against this, because it is a good one. This is an awards entry, and it carries the numbers of an awards entry. The awards entry puts the menus in 382 locations. [One Show](https://www.oneclub.org/awards/theoneshow/-award/58448/eye-test-menu/) The launch release, from the year before, counted more than 900 screening locations and just over seven thousand drivers, in a country with three and a half million truck drivers. [BestMediaInfo](https://bestmediainfo.com/ad-craft/titan-eye-introduces-the-eye-test-menu-for-truck-drivers-launches-ad-film-7315230) Nobody reconciles those two location figures anywhere, which is its own small tell. Earned media was put at around five million rupees, which is not a large sum. The headline outcome, an estimated 42,000 accidents prevented, is a model Ogilvy India built about its own work, printed in large orange numerals on the case board. [Kyoorius](https://kyooriuscreative2025.awardsengine.com/?action=ows:entries.details&e=182119&project_year=2025) It sits inside a submission that has since won five awards across four shows. [D&AD](https://www.dandad.org/awards/professional/2025/239672/eye-test-menu/) Meanwhile the strategic move, go where they already are and use the thing they already read, is ordinary planning craft. Any decent creative director in Mumbai produces it over tea. And there is a sharper charge than any of that. A menu costs almost nothing to print and it films beautifully. The constraint may have been chosen because it was cheap and photogenic, not because the finding demanded it. Concede most of it. The reach is small, and the accident figure should be thrown out. Neither touches the mechanism. VisionSpring, in research led with Queen's University Belfast and published in August 2025, screened those 9,857 drivers and found two-thirds had never been checked before licensing. [Optometry Times](https://www.optometrytimes.com/view/new-study-reveals-majority-of-commercial-drivers-in-india-have-never-had-vision-tested) That work appeared ten months after the menus went up, so nobody at Ogilvy had it. It documents a condition that long predated them both. Public awareness and the licence itself are the two interventions that do reach everybody, and both had already been run across the whole population. Both broke at the same joint: the point where a man has to notice something about himself. Then the charge that the constraint was chosen because it would film well. One hero menu in one language would have filmed just as well as several hundred in many, and the regional versions cost money a camera cannot detect. And the campaign's own prevalence claim, one in three, rests on no published study. [One Show](https://www.oneclub.org/awards/theoneshow/-award/58448/eye-test-menu/) The Belfast-partnered screening found 17.7 per cent with blurry distance vision. Take the lower number. Then remember that two-thirds of them had never once been checked, and the argument gets stronger. #### The bottom of the menu Titan and Ogilvy say the exam ended in a prescription and glasses shipped to his home, further than a poster ever gets. The step that would have finished the job never shows up on any case board: that driver back in his cab, already wearing the correction, a voucher folded in for whoever is waiting at home. A finding survives when it reaches a man's eyes before his next drive, not his letterbox after it. --- ### Fight the Line, Not the Thread [Read more](https://www.gaganmalik.io/en/newsroom/fight-the-line-not-the-thread) [Retrieval brief — fight-the-line-not-the-thread] Category: design Thesis: Fight the line, not the thread: chat interfaces that treat the whole reply as the only follow-up unit force readers to re-describe a passage already on screen; highlight-to-quote lets them point at the line they mean. Answers: What is highlight-to-quote on Ask and how does it work? | Why should AI chat products support selecting a passage for follow-up instead of retyping? | Does quoting an excerpt remove the rest of the conversation from the model? | How does highlight-to-quote relate to WhatsApp quote-and-reply or Artifacts and Canvas? | What product trade-offs did Gagan Malik make shipping the Ask quote chip? | When is a one-session teach tip the wrong way to introduce a new gesture? Implication: For conversational AI products with multi-paragraph answers: ship a passage-level follow-up unit (selection → quote chip → focus block) rather than forcing paraphrase into a blank box; keep full thread history and layer the excerpt as subject, do not replace context; treat discovery of the gesture as a design problem equal to the gesture itself. Proof: Ask highlight-to-quote (PB-200, Aug 2026) on gaganmalik.io/en/ask — select → Ask about this → quotedContext focus block; WhatsApp quote-and-reply, June 2016 (Engadget / Mariella Moon); Anthropic Artifacts (June 2024); OpenAI Canvas (October 2024) — sibling document-side pointing, already argued in familiarity-is-a-floor-not-a-ceiling Related essays: familiarity-is-a-floor-not-a-ceiling, search-finds-pages-chat-finds-answers, show-the-job-not-the-feature, waiting-is-a-design-surface, you-are-not-the-user Concepts: highlight-to-quote, second-cursor, focus-block, quoted-context, articulation-barrier, chat-UI, selection-UX, Ask By Gagan Malik I was three paragraphs into a long Ask answer on my phone when I found the one line worth arguing with. The rest of the reply was scaffolding: context I already knew, a summary I did not need repeated. One sentence, buried in the middle, made a claim I wanted to push on. My thumb was already moving toward the reply box, ready to type something like the bit about the retry logic. Then I stopped. Why was I about to describe a sentence that was already sitting on the screen in front of me? Highlight-to-quote is the fix I put live instead. Drag a selection over a passage in an Ask reply. The interface pins that exact span as the subject of your next message, shown as a small quote chip in the box. The model receives what I call a focus block: the excerpt, placed ahead of your follow-up. That is how it knows which sentence you mean without you retyping it. Fight the line, not the thread. Ask composer mockup with a quoted passage pinned above the follow-up field — highlight the line, then ask about it. #### The Reply Is Treated as One Block Because That Is Easier to Build Selection is a second cursor. Chat products default to a single unit of follow-up because a single unit is what a demo needs. One input field, one growing transcript, one send button: nothing to model beyond appending the next message to the same list. Building a second unit, a passage inside a reply that a reader can address directly, means answering three harder questions. What counts as an addressable span? How does it survive a re-render? What happens when the model producing the text has not finished the sentence yet? None of that shows up in a first demo. The bubble is enough to convince a room the product works. That default rewards whoever releases first, not whoever answers best. A five-paragraph reply that buries one wrong number in the third paragraph gets the same follow-up box as a one-line reply. The reader question I opened with, how you ask about one paragraph without retyping it, has exactly one answer inside a product built this way. Type it again, and hope the model works out from your paraphrase which paragraph you meant. That is not laziness on the reader's part. The interface never gave them any other way to point. Prose is a slower, blunter pointer than a thumb already resting on the passage. #### Messaging Apps Already Solved This. Chat Products Forgot. There is a genuine case for sending everything. A model reasoning over the full conversation has more material to work with than a model reasoning over four words a reader dragged across a screen. More information should, in principle, produce a better answer. Set that against a pattern most readers already carry in their pocket. In June 2016, WhatsApp added quote-and-reply for one message inside a long group thread. Most people would not open their first AI chat window for another six years. Engadget covered the rollout as it left Android beta for the stable apps. Nobody using that gesture has to retype who or what they mean. Anthropic and OpenAI ran a version of the same fix on the document side of a chat product. Artifacts opened a second panel beside the conversation in June 2024. It let a person point at a paragraph the model had just written instead of re-describing it in a prompt. That is a distinction I made in [Familiarity Is a Floor, Not a Ceiling](https://www.gaganmalik.io/en/newsroom/familiarity-is-a-floor-not-a-ceiling). OpenAI answered four months later with Canvas, built for the same reason. Both panels fixed this for a document sitting beside the conversation. Neither fixed it for the conversation itself. That is the one surface [Search Finds Pages. Chat Finds Answers.](https://www.gaganmalik.io/en/newsroom/search-finds-pages-chat-finds-answers) named. That essay argued the surface should be judged on a cited answer, not a page of links. An answer worth citing ought to be an answer you can point back at. The competitors I watch most closely already put the gesture back on the reply. Perplexity labels it Add to follow-up and parks the selection above Ask a follow-up. Claude drops the same span into the composer as a quote block before Write a message. Both treat the answer as the surface you point at, which is the unit Artifacts and Canvas left behind. Interactive lab: Selection Quote. Highlight a line in an Ask reply and attach it to the follow-up. Fight the line, not the thread. Interactive demo embedded in the essay at /newsroom/fight-the-line-not-the-thread. #### What the Quote Chip Cost to Build Building the gesture meant making decisions nobody reads a product page for. The floating Ask about this button has to hit a 44-point touch target. The phone screen is about to be crowded by a keyboard. An excerpt has to be long enough to mean something and short enough not to become a second essay. The interface accepts a single word and rejects anything past 800 characters, with a toast that says which limit you hit. The highlighted span itself has to survive the interface redrawing around it while you keep reading. That meant reaching for a browser highlighting method most readers have never heard of. Browsers that do not support it yet need a fallback. None of this is a feature idea. It is the argument, written in code, that a passage deserves the same care as a whole reply. The riskiest decision was the smallest one: a teach tip that shows once, in a single session. It tells a reader to select text in an answer to ask about it. I do not yet have the numbers on how many people see that tip and act on it. I am not going to invent a percentage to make this paragraph land better. I have already made this argument about other people's changelog modals: a good release still fails if nobody notices what changed. A one-line tip that fires once is exactly the kind of good idea a busy reader dismisses without reading. I do the same thing to every What's New modal I did not ask for. Somewhere on Ask this week, a reader is retyping the paragraph about the toast validation into the reply box, one clause at a time. Nobody showed them they could point at it instead. #### You Point Before You Can Say What You Mean Watch a child whose thumb is already moving before they have the word for what they want. They do not construct a sentence. They raise an arm and open a fist toward the shelf, then wait. The pointing works faster, and more precisely, than any word they will learn to say instead. Language arrives later, mostly as a patch for the times nobody is close enough to see where the finger is aiming. A highlighted passage is that same gesture wearing a cursor. It gives the reader a kind of reference no amount of extra context in the model's memory can substitute for. A description is always a little late and a little wrong, built after the fact from memory of what you meant. A point is exact the moment you make it. Selection puts that gesture back into a reply built entirely out of language, the one place where pointing had quietly gone missing. #### The Strongest Case for Sending Everything The fairest version of the objection is this. Send the model the whole thread, every time, and never let a reader narrow what the model sees. Every prior turn gives the model more to reason with than four words dragged out of a paragraph. The sentence right after the one a reader highlights is sometimes the qualifying clause that made the highlighted sentence true in the first place. Excerpting looks like precision. It can just as easily be omission wearing precision's clothes. The reader doing the highlighting is the person least likely to notice what they left out, because they already believe they know which part matters. That risk survives contact with the interface only if quoting means replacing. It does not. Ask sends the model the full conversation exactly as it stood before the selection. It adds the excerpt only as an instruction layered on top. Treat this passage as the subject; do not ask which article or topic the reader means. Nothing upstream of the quote is deleted, trimmed, or hidden from the model reasoning over the reply. Pointing at a passage tells the model where the reader's attention is. It was never a request to look away from everything else. #### Fight the Line Fight the line, not the thread. A quote chip did not add a feature to Ask; it gave the reader a second cursor for any product still charging the blank-box tax one paragraph later. Somewhere on Ask this week, a reader will drag four words across a sentence rather than retype them from the reply box. --- ### v1.24.0 [Read more](https://www.gaganmalik.io/en/newsroom/release-1-24-0) [Retrieval brief — release-1-24-0] Category: technology Thesis: A full data-viz lab you can download as a slide or a story, and Ask that quotes the bit you highlighted. Answers: What changed in gaganmalik.io v1.24.0? | What's new on Gagan Malik's website in v1.24.0? | What are the latest gaganmalik.io release notes? Concepts: release notes, site update, gaganmalik.io A full data-viz lab you can download as a slide or a story, and Ask that quotes the bit you highlighted. --- ### Why Does Every Episode End Right There? [Read more](https://www.gaganmalik.io/en/newsroom/why-does-every-episode-end-right-there) [Retrieval brief — why-does-every-episode-end-right-there] Category: technology Thesis: AI did not lower the cost of story in microdrama production; it lowered the cost of a narrative boundary, so after a free sample the industry can price nearly every remaining cliffhanger inside a single film's runtime. Answers: Why does every microdrama episode end on a cliffhanger? | If AI makes video cheaper, why do microdramas feel like a different kind of product rather than cheap television? | What actually got cheaper about AI-produced video: the story or the structure? | How does the microdrama paywall model work, and what share of revenue comes from coin unlocks? | Why did the top of the microdrama market not collapse even as AI-made volume exploded? Implication: When a product decision (a cut, a boundary, a stopping point) becomes free to manufacture, expect a business to multiply that decision rather than the content around it; audit AI cost-savings claims by asking what specific unit got cheaper, not by assuming quality followed price down. Proof: DataEye industry data: AI titles rose from 7% to 38% of China's top-100 microdrama chart in the year to January 2026; Plot Party founder primer: ~74% of global microdrama revenue is coin/token in-app purchase, paywall after episodes 8-15; Wall Street Journal on ReelShort (2025/26): free early episodes, cliffhanger paywall, ~18 coins (~$0.20) to unlock the next beat; viewer data retunes twists inside a ~3-month loop; Xinhua/Qiushi reporting: Hongguo (ByteDance) raised its content budget 40%+ while an AI title still topped its chart on 5 April 2026; Harvard Business Review (Li, Xiong, Sun, March 2026): platforms test story concepts through thousands of micro-advertisements before greenlighting production; Fan Xiang (photographer, The Standard/East Week, 8 April 2026): unpaid six months, income under 5,000 yuan in March 2026 Related essays: why-your-suggestions-never-run-out, bundling-was-built-for-things-that-cost-nothing-to-copy Concepts: microdrama, cliffhanger as payment boundary, AI video economics, cost floor, coin unlock / in-app purchase, vertical serial drama By Gagan Malik A woodcutter finds a fox dying in the snow and feeds it braised duck. The fox lives. The duck comes back as a vengeful woman. Four young adults in China made it earlier this year with AI tools, for very little money. *Rescuing the Fox on the Snowy Mountain* has since been watched more than five billion times. It was originally made to advertise braised duck products. [East Week](https://www.thestandard.com.hk/china/article/328664/Short-AI-films-shake-up-micro-drama-industry) Sit with that, because the cheapness is not the interesting part. A microdrama, the vertical serial these phone apps run on, is sixty to a hundred episodes of one to three minutes each. Every single episode ends on a cliffhanger, which is how the industry itself describes the product. [Plot Party](https://www.plotparty.ai/page/blog/posts/microdrama-101-everything-you-need-to-know) Add the runtime up and you get sixty to a hundred minutes: a feature film. So a film's worth of story, sold in a hundred pieces. After a short free sample, most of those pieces end exactly where the next one has to be paid for: that is the whole answer to where the cuts fall. The cutting is what got cheap. #### The volume story is true, and it is the least interesting thing here The numbers are not in dispute. DataEye, the industry data firm, counted more than ten thousand AI-generated animated micro-dramas going online every month from the start of 2026. Roughly fifty thousand new AI-native titles were added to Douyin in March alone. [Xinhua](https://en.qstheory.cn/2026-04/23/c_1177889.htm) On January's top-hundred chart, AI titles made up thirty-eight per cent, against seven per cent a year earlier. A director who moved from live action to AI put his costs at about a tenth of a traditional shoot. That last figure needs a caution, because it is being repeated carelessly. Two different claims have merged into one statistic. The one-fifth version belongs to Vigloo, a South Korean producer, on total programme cost after shifting thirty per cent of its budget to AI workflows. [SCMP](https://www.scmp.com/economy/china-economy/article/3352019/how-china-using-ai-and-state-funding-transform-micro-drama-industry) The one-tenth version belongs to that Chinese director, describing per-title cost at comparable quality. Both point the same way, so little turns on it, but you should know which sentence you are quoting. The conventional reading suits everyone repeating it: studios file it under quality, tool vendors file it under democratising access, and neither asks what changed shape. #### The top of the market did not fall If cheapness were the whole story, the expensive tier would be dying. It is not. Hongguo, the short drama platform owned by ByteDance, said it would raise its content budget by more than forty per cent this year. It would keep its live-action investment. And on 5 April an AI-produced micro-drama reached number one on Hongguo's own popularity ranking. The notable thing was what the audience argued about afterwards: the storyline, and not the rendering. That last detail comes from Xinhua, carried by Qiushi, the Chinese Communist Party's theoretical journal. It reports on an industry the same state subsidises by up to two million yuan a title. [The Next Web](https://thenextweb.com/news/china-micro-drama-ai-state-funding) Know whose account you are reading. The names, institutions and figures are checkable, and outlets outside the mainland report the same numbers while disagreeing with the framing. None of them dispute the shape of it. The cheap tier did not eat the expensive one. Audiences have not stopped caring whether a story is any good. So the thing that changed is somewhere else. #### Where the money actually enters Three facts, taken from the industry's own primer, written by the founder of one of the AI microdrama studios. The first eight to fifteen episodes are free. After that you buy coins to unlock each new one. Roughly seventy-four per cent of global revenue arrives that way. The primer describes the model, in its own words, as the one that built mobile gaming. The same primer reports that users of one leading app spend 35.7 minutes a day inside it. That is more than the average Netflix viewer spends on Netflix's own phone app. The Wall Street Journal filmed the same structure on a ReelShort set in the United States and said it out loud. The first episodes are free. Of course they leave you on a cliffhanger. To see the kiss, you pay eighteen coins, roughly twenty cents. [WSJ](https://www.youtube.com/watch?v=FWP-nYc12XM) Joey Jia, who launched the app to copy China's boom into the US market, drew the contrast with streaming in one sentence. Netflix is the living-room Friday night, two or three solid hours. ReelShort is the elevator and the lunch queue, three to five minutes of bite-sized story. The cut is not a creative flourish that happens to suit a phone. It is the unit of sale for the minutes people already have. #### The cliffhanger is the checkout Now put those facts next to the rule that every episode ends on a cliffhanger. The cliffhanger stops looking like a device inherited from soap opera and starts looking like the place where the transaction happens. After the free sample, eight to fifteen episodes depending on the app, the rest of a sixty-to-a-hundred-episode season is a run of paid unlocks. Those are dozens of moments, right there in the runtime, at which a person decides whether to pay again. I have written before about a different shape in [Why Your Suggestions Never Run Out](https://www.gaganmalik.io/en/newsroom/why-your-suggestions-never-run-out): a feed built never to reach a finish line. I have written about what happens when pricing theory built for things that cost nothing to copy meets a one-of-one object. See [Bundling Was Built for Things That Cost Nothing to Copy](https://www.gaganmalik.io/en/newsroom/bundling-was-built-for-things-that-cost-nothing-to-copy): zero-copy logic on the wrong unit. A microdrama has no shortage of edges. Someone manufactured a film's worth of them and priced nearly every seam after the sample. I am describing what the structure does, not what anybody sat down intending. Function is enough to be going on with. Seventy-eight unlocks after the free twelve-episode sample add up to $15.60: more than a cinema ticket, while a Netflix episode sits at zero. #### Fan Xiang has had one job since March Fan Xiang is a photographer. He went more than six months without being paid for a production he had already finished. "There's clearly been less work this year," he said in March. "I've only landed one job, and my income has dropped below 5,000 yuan." *East Week*, which reported his case, put the number of people in China making a living from micro dramas at over two million. Set against him, Xu Yan. She is thirty-six and joined the industry in Xi'an two years ago. She went from background extra to speaking-role performer in three months, and wanted people to know that a late start need not disqualify anyone. Both accounts are true, and one of them arrived while the floor was still there. The same reporting named two more figures who are not people at all. Qin Ling Yue and Lin Xi Yan are AI-generated performers, cast in sixty planned shorts, accused of lifting real actors' features without consent. As for my own position: I watched three of these and paid for none. I stopped exactly where the wall is, which is the behaviour the whole model is built around. The free episodes are not generosity. They are the sample. #### A sheet of stamps Think about a sheet of stamps. The paper is ordinary and the ink is ordinary. What turns one sheet into a hundred sellable things is a line of holes, punched by a machine. Somebody chose the positions before the sheet was printed. Nobody has ever bought a sheet of stamps for the perforations. The perforations are still the entire commercial decision. They are what lets the post office sell you one instead of a hundred. You feel them in the hand: the small resistance, then the give, and now and then a tear that goes slightly wrong. The line decides where the thing ends, whether or not the picture agrees. A microdrama season is a film with the perforations already punched. If you want to feel the wallet math rather than only read it, open the [Stamp Sheet Player](https://www.gaganmalik.io/en/lab/stamp-sheet). It runs Industry vs Honest ledger on a compressed sheet: free sample, then a priced cut at every cliffhanger. #### The strongest objection comes from inside the industry Give the counterargument its full weight, because the strongest version of it is simple. Cliffhangers are not new. Every soap opera, every Saturday-morning serial, every Dickens instalment stopped mid-scene on purpose, long before anyone sold a coin. Call it retention craft, or storytelling grammar, and the transaction argument starts to look like a checkout button read into an old device. The industry's own experts raise a related point, about quality rather than structure. Feng Shengyong, an official at China's National Radio and Television Administration, put it plainly. AI can calculate audience preferences, he said, but emotional resonance is harder to calculate. The works people remember are usually not the most sensational but the most authentic. He Tianping, who teaches journalism at Renmin University, adds that AI-generated video still shows clear gaps in understanding human behaviour. Characters may lack coherent logic. On that reading the market is still young. Cheap tools flooded the bottom, audiences are already sorting it out, and the April chart-topper proves story still decides. That argument is correct about quality, but it does not touch the structural claim, because nobody in the industry is disputing the structure. Episodes eight to fifteen are free, coins buy every episode after that, and roughly seventy-four per cent of global revenue arrives that way. Every season still ends every episode on a cliffhanger: after the free sample, dozens of payment boundaries inside one film's runtime, not a surviving habit from soap opera. Harvard Business Review published an analysis of the sector in March 2026, by Haiyang Li of Rice University with two former ByteDance and Tencent operators. It describes platforms testing story concepts through thousands of micro-advertisements before greenlighting production. [Harvard Business Review](https://hbr.org/2026/03/lessons-from-chinas-short-drama-boom) The Journal's ReelShort reporting adds the loop after the cut ships. The app keeps watching when you pause, tracks where engagement drops, and feeds that back into the next twist, with the company aiming to close the loop inside three months. [WSJ](https://www.youtube.com/watch?v=FWP-nYc12XM) Read that next to the braised duck and the order is plain enough. The work is made for the advertisement that tested well, then retuned at the boundary that charges. The market did not so much rescue quality as give it a job: getting you past episode fifteen. #### The line was punched first The runtime never changed: what changed is the number of places it can be interrupted, chosen before anyone wrote a scene. When a boundary costs nothing to make, somebody will make more of them, and the thing being sold becomes the boundary, not what sits between. Fan Xiang has had one job since March, is still owed for the one before it, and made under five thousand yuan that month. --- ### A Skeleton Screen Is a Promise About Shape [Read more](https://www.gaganmalik.io/en/newsroom/a-skeleton-screen-is-a-promise-about-shape) [Retrieval brief — a-skeleton-screen-is-a-promise-about-shape] Category: design Thesis: A skeleton screen is a structural promise about shape, not decoration; a mismatched skeleton breaks the same trust a bare Loading string breaks, in layout instead of language. Answers: What is a skeleton screen actually promising the visitor? | Why does a loading skeleton that doesn't match the real layout feel worse than a plain spinner? | How should App Router loading.tsx skeletons be built so they don't lie about shape? | Are skeleton screens just a perceived-performance trick? | What did Gagan ship for loading states on gaganmalik.io? Implication: Treat every skeleton as a layout contract: build it from the real component's shape, keep it in sync when that shape changes, and support aria-busy plus reduced-motion so the promise holds for assistive tech too. Proof: 19 page skeletons in components/skeletons/ wired through 29 App Router loading.tsx files, gaganmalik.io; /newsroom/skeleton-preview QA route for shape checks without a real slow load Related essays: waiting-is-a-design-surface, design-is-attention-control-not-aesthetics Concepts: skeleton screen, shape drift, perceived performance, loading states, Suspense, accessibility, reduced motion By Gagan Malik I opened the newsroom hub on a throttled connection to watch something that isn't supposed to be noticed: the wait. Before a single headline landed, the page already had rails, card shapes, and a rectangle where a badge would sit. Nothing was blank. Nothing spun. The layout had arrived first, without its words. That grey, breathing outline is a skeleton screen: a static placeholder that mimics the block shape of incoming content, not a blank page and not a spinner. In Next.js, `loading.tsx` is the file that renders while a route's real content is still loading. It wraps the route in a waiting state and swaps out when content is ready. Most teams treat what goes inside as an afterthought. I don't, because a skeleton makes a claim before it has earned the right to. It is a structural promise about shape, not decoration. In advance and in pixels, it tells the visitor where the title will sit and how wide the image will be. It also signals how many lines of text to expect. Break that promise and the visitor pays the same tax they pay for a lying Loading string. That is a small loss of trust, spent the moment real content resolves into a shape nobody warned them about. Call that shape drift: a skeleton that stops matching the layout it stands in for. The promise turns into a small lie on every load. Assets arrive into the shapes they promised: some covers resolved, others still grey, not a blank spinner. #### The Blank Page Trained Us to Expect Nothing Before skeleton screens, software waited in one of two costumes: blank white space, or a spinner going nowhere. Both are honest in the dullest way possible. Nothing about them can be wrong when the real content lands, because they promised nothing. That is also the failure. They train the visitor to expect nothing from the interface at exactly the moment the interface has the most to say about what is coming. The tempting middle ground is the one component libraries ship by default: an identical grid of grey rectangles. It gets dropped onto every route because a review wants a loading state before launch. Nobody studied what that particular route actually looks like. It costs nothing to add and promises nothing true, which makes it barely better than the blank screen it replaced. A skeleton only becomes an improvement once it stops being generic and starts looking like the destination. A wait that promises nothing fails safely. A wait that promises the wrong shape fails loudly, one beat later, exactly when the real page tries to prove the pause was worth it. #### The Shape Is the Promise, Not the Word I made a version of this argument about words in [Waiting Is a Design Surface](https://www.gaganmalik.io/en/newsroom/waiting-is-a-design-surface). A wait that says Loading when the system knows exactly what it is doing is quietly apologising for existing. A wait that says something true admits the work is real. The same test holds for shape. A visitor reading an interface during a wait is looking for a signal about what is coming. That signal can arrive as a word or a rectangle. A lying signal in pixels breaks trust just as fast as a lying signal in prose. It just does it more quietly. Nobody screenshots a slightly wrong-shaped skeleton to complain. They flinch once, when the real page lands somewhere else, and move on before they can name what bothered them. Ask, the assistant on this site, keeps its own wait state in words. A small rotating verb turns over while an answer streams in, with nothing to look at but a sentence taking shape. The newsroom, the articles, and the tools on this site keep theirs in shape instead. Rails, frames, and lines stand in for content that has not arrived yet. Same decision, different material. Either way, the wait is a place I designed on purpose, not a gap I filled in on the way to launch. #### Nineteen Skeletons and a Preview Route Nobody Sees Every skeleton on this site starts from one shared primitive: a grey, rounded rectangle that pulses gently while it waits. It is marked in the code so screen readers and browser tooling both know what they are looking at. From that primitive, nineteen separate skeleton components exist in the codebase, each hand-shaped to its route. They are wired through twenty-nine `loading.tsx` files across the site. The newsroom hub's skeleton carries the same rails and card proportions as the cards that will replace it. Two more shapes from that same primitive sit below, live rather than a screenshot: Ask and the newsroom hub, each on desktop or mobile. Interactive lab: Skeleton shape. A skeleton screen is a promise about shape: wireframe the layout before content lands. Interactive demo embedded in the essay at /newsroom/a-skeleton-screen-is-a-promise-about-shape. Nineteen hand-built files are real maintenance. They are checked at a QA route called `skeleton-preview` so a mismatch can be caught by looking at a screen, not by throttling a network and hoping. None of this argues for skeletonising every wait. A button that resolves in two hundred milliseconds is better served by a spinner or an optimistic update. A full-page skeleton would vanish before anyone had time to read it. #### The Stair That Isn't There There is a name for the jolt when a foot expects a step of one height and lands on a step built to another. It is the missing-stair reflex: a small stumble your body produces even though nothing is broken. Your legs notice the wrongness a half-second before your brain catches up. Scrolling a page mid-load does the same thing to the eye and the hand. By the second glance, you have already built a rough map of where the title sits and how wide the image will be. That map even predicts how many lines the paragraph takes. It is based entirely on the skeleton in front of you. When the real content resolves into a different shape, there is a smaller version of that same stumble. Nothing has broken. The page still works. You still flinch, because the promise the shape made a second earlier did not hold, and your eyes had already started trusting it. #### Isn't This Just Placebo? Give the objection its full weight. Luke Wroblewski named the pattern in 2013 while shipping Polar: a spinner makes people watch the clock, while a blank version of the page into which content gradually loads focuses them on progress instead. [Avoid The Spinner](https://www.lukew.com/ff/entry.asp?1797) Nielsen Norman Group later summarised the same bet. A wireframe-like wait creates the illusion of a shorter load and lowers cognitive load by previewing structure before the words arrive. [Skeleton Screens 101](https://www.nngroup.com/articles/skeleton-screens/) If the benefit is mostly perception, the argument goes, skeleton screens are perceived-performance theatre: a psychological trick that makes a wait feel shorter without making the wait itself one second shorter. The network call underneath takes exactly as long whether the visitor is staring at a grey box or a blank one. By that logic, calling the shape a promise is generous branding for what is really a distraction. It is the visual equivalent of hold music that makes a queue feel shorter while doing nothing to make it move faster. That critique is fair against decoration. Tim Kadlec's later reading of the pattern is sharper still: when the skeleton falsely represents what is coming or where it will sit, the confusion overcomes any perceived-speed benefit. [Effective Skeleton Screens](https://timkadlec.com/remembers/2020-11-02-skeleton-screens/) Aim that warning at any skeleton drawn from imagination rather than from the component it stands in for. It stops working against a skeleton built from the real layout's own code. The newsroom skeleton is not a mockup of the newsroom. It is the newsroom's own card shapes, rendered before their contents exist. The same loading container announces `aria-busy` to assistive technology while its decorative bones stay hidden from screen readers. Reduced motion can strip the shimmer and leave the shape at a duller opacity without losing anything true. Theatre would keep the shimmer and drop the accuracy the moment nobody was watching closely. A kept promise can lose the shimmer entirely and still tell the truth about what is coming. The truth was never the animation. It was the shape. A skeleton screen only ever proves one thing: whether the shape it showed you matches the shape you got. Every skeleton on this site earns that promise the same way, checked against its real component instead of drawn once and left to drift. I still open the skeleton preview before every redesign, because somewhere a visitor trusted the shape before the page had earned it. --- ### Feeding the Meter [Read more](https://www.gaganmalik.io/en/newsroom/feeding-the-meter) [Retrieval brief — feeding-the-meter] Category: technology Thesis: Every meter in a 2026 AI subscription expires at five hours, a month or a year, and only the fastest of those clocks is about the machines; a twelve-month expiry on prepaid credit reserves no capacity, so it is a pricing decision dressed as a physical constraint, filed in the document nobody opens. Answers: Why does my AI subscription reset every five hours, and what happens to allowance I do not use? | Do prepaid AI credits expire at Anthropic and OpenAI, and do they roll over? | Is an AI usage limit capacity management or pricing? | Why did GitHub Copilot, Cursor, Anthropic and OpenAI arrive at the same three-part plan shape in one quarter? | What does mobile-phone and gym-membership pricing research say about metered AI plans? | What should a pricing or product leader check before shipping a metered AI plan? Implication: Read the reset schedule before the price, and ask of each expiry what it protects. A limit that moves when compute arrives is about capacity; an expiry on money already paid is about the bill. Proof: Anthropic Supplemental Credit Terms, effective 4 March 2024: Usage Credits expire one calendar year from issue, non-refundable; OpenAI credits help centre: valid twelve months, unused credits expire and do not roll over; GitHub Copilot usage-based billing, effective 1 June 2026, including the stranded capacity phrasing scoped to businesses; Grubb, Selling to Overconfident Consumers (AER 2009); DellaVigna and Malmendier, Paying Not to Go to the Gym (AER 2006); Anthropic doubled five-hour limits on 6 May 2026 citing new compute (essay's own qualification); Pinkashtray0, anthropics/claude-code issue #81869, 28 July 2026: €3.91 correction credit expiring 21 July 2027 Related essays: subscription-fatigue-is-the-wrong-instrument, bundling-was-built-for-things-that-cost-nothing-to-copy, why-your-suggestions-never-run-out Concepts: metering, AI pricing, expiry, three-part tariff, breakage, rate limits, prepaid credits, rolling window By Gagan Malik On a kerb, feeding the meter means paying more to keep the spot. That is one face of the machine. The other is quieter, and it takes money from people who never run hot. It is the allowance you already paid for, gone on a clock whether you spent it or not. You have looked at the bar, seen an hour left, and decided to start tomorrow instead. Your plan resets tonight either way. On 1 June 2026 GitHub Copilot moved every plan onto usage-based credits, the free tier included. [GitHub Blog](https://github.blog/news-insights/company-news/github-copilot-is-moving-to-usage-based-billing/) Each plan now carries a monthly allotment pegged to its own price. Ten dollars of credits on the ten dollar plan. Thirty-nine on the thirty-nine. Cursor includes a set amount of model usage in every plan and bills anything past it in arrears. [Cursor Pricing](https://cursor.com/pricing) Anthropic's included usage resets on a rolling five-hour clock, with weekly caps above it. [Claude Help Center](https://support.claude.com/en/articles/9797557-usage-limit-best-practices) Inside one quarter the AI subscription stopped being a flat fee and became a meter. The sensible response was to learn to read the meter. That is what I did. It is the smaller half of the story. There is a second clock, and it is not on the meter. Anthropic's [Supplemental Credit Terms](https://www.anthropic.com/legal/credit-terms) took effect on 4 March 2024. They live on a different domain from the app. They say usage credits are non-refundable, and that they expire one calendar year from the day they are issued. OpenAI's help centre says the same thing in friendlier language: credits are valid for twelve months, and unused credits expire and do not roll over. [OpenAI Help Center](https://help.openai.com/en/articles/12642688-using-credits-for-flexible-usage-in-chatgpt-freegopluspro-sora) So here is the honest answer to the question you actually have. You are paying for the contract everybody now sells. A fixed monthly fee. An allowance that costs nothing extra until it runs out. A steep price after that. Feeding the meter is all three, and the part that catches most of us is the middle one. Your mobile phone bill was built this way in 2009. #### The Contract Was Written for Mobile Phones Guess how much of your plan you used last month, then go and look. The gap between those two numbers is the whole business. In 2009 Michael Grubb showed what to do with it. When customers overestimate how well they can forecast their own demand, the profit-maximising contract takes exactly this shape: a fixed fee, an allowance that feels free while it lasts, and a punishing rate past the line. [Grubb, AER 2009](https://www.aeaweb.org/articles?id=10.1257/aer.99.5.1770) He was not describing artificial intelligence. He was describing American mobile phone plans, and he called the paper *Selling to Overconfident Consumers*. Three years earlier, Stefano DellaVigna and Ulrike Malmendier had measured the gap directly. Across 7,752 gym members, people forecast 9.50 visits a month and made 4.17. [DellaVigna and Malmendier, AER 2006](https://eml.berkeley.edu/~sdellavi/wp/gymempAER.pdf) Those on flat contracts above $70 paid more than $17 for every visit they took. A ten-visit pass sat on the counter at $10. Nobody was defrauded. Everybody was confident. What is striking about 2026 is the convergence. Four companies, with different investors, different hardware and different customers, arrived at the same instrument within months of each other. That is not copying. It is what happens when an industry finds the contract that fits its customer. And this customer cannot tell you, within a factor of two, how much they used last Tuesday. I have argued before that [subscription fatigue is the wrong instrument](https://www.gaganmalik.io/en/newsroom/subscription-fatigue-is-the-wrong-instrument) for reading arrangements like these. This is the same mistake in a newer coat. We keep auditing the price and ignoring the schedule. #### Two Documents, Written for Two Different Readers The best usage guidance in the industry is published by Anthropic, and it works. Plan the conversation before you start it. Batch related questions into one message instead of five. Put stable documents in a project, where they are cached and do not count against your limits each time you reuse them. Prefer the smaller model unless the work needs the larger one. Start a fresh chat for a new topic, so you stop paying to carry the old one. Watch the progress bars in your settings. [Manage usage credits](https://support.claude.com/en/articles/12429409-manage-usage-credits-for-paid-claude-plans) Every item is true. Every item lowers your bill. I use most of them. Hold that at full strength, then notice what else each item does. It flattens peak load on a fleet of very expensive machines. The guidance is genuinely for you. It is also demand management, performed by the customer, unpaid, with a straight face on both sides. Now set the two documents next to each other. The advice was updated on 2 June 2026. It sits in a searchable help centre and it is written in the second person. The expiry sits in those Supplemental Credit Terms dated 4 March 2024, on the corporate domain, in the third person, addressed to a Customer who is presumed to have read it. Between them the schedules line up like this. Included usage, gone every five hours. Prepaid balance, gone at twelve months. Copilot's allotment, renewed monthly and pegged to the price of the plan. And Anthropic's bundles, which cost 10 per cent less at $50 and 30 per cent less at $1,000, drawing down against that same twelve-month clock. [Buy usage bundles](https://support.claude.com/en/articles/14246112-buy-usage-bundles) Nobody rolls over. GitHub, announcing its change, even gave the leftover a name. It was introducing pooled usage across a business, it wrote, to help eliminate stranded capacity. Individual plans do not pool. #### A Remedy You Can Only Spend by Spending More On 28 July 2026 an EU subscriber posting as Pinkashtray0 filed a billing complaint on Anthropic's public issue tracker. [claude-code #81869](https://github.com/anthropics/claude-code/issues/81869) The facts he reported are small and precise. On 17 July, for thirty-one minutes, a configuration error billed some usage against credits instead of counting it towards included weekly usage. Anthropic acknowledged it by email, which he quoted in full. It added €3.91 to his account, stated to be twice the amount billed in error. The credit does not expire until 21 July 2027. It also cannot be spent, because a correction credit is only consumed once you exceed your plan limits, and he does not. He mentioned, almost in passing, that he had held around $100 of purchased extra usage for about a year without spending a cent of it. Then he asked Anthropic to confirm the exact date that money would lapse. His sentence is the one worth keeping. A remedy that only materialises if the customer spends more money is not a remedy. Nothing here was hidden and nothing here was illegal, which is exactly why it is worth looking at. The terms were published. The error was admitted. The compensation was doubled. And the man is still out of pocket, in the only currency he can actually use. Two pounds and change, and a clock. #### You Paid Until Four. It Finished at Half Past Two. You have done this. You stand at the machine at the kerb, guess how long the thing inside will take, add twenty minutes because you would rather not be thinking about it, and put the ticket face up on the dashboard. Then it finishes early. You walk back past a car circling for a space, and you drive away with ninety minutes still printed on a piece of paper nobody will ever look at again. The kerb behind you goes back to zero before you reach the junction. Nobody stole anything, and the machine did exactly what it said. You bought ninety minutes, the ninety minutes existed, and they went back to the road the moment you left. That is the part a progress bar cannot show you. Not that you might run out, which you can see perfectly well. That the clock was never counting your time. It was counting the space's. Try the strip that shows remaining and reset together. Interactive lab: Honest Clockstrip. Put remaining allowance and the reset clock under the composer so percentage alone never pretends to be the contract. Interactive demo embedded in the essay at /newsroom/feeding-the-meter. #### The Strongest Case for Calling It a Ceiling Here is the objection, and it is a good one. Inference capacity is finite, expensive and violently peaky. A rolling window is the least-bad instrument anyone has found for stopping a handful of very heavy users from degrading a service for everybody else. Rollover would make that worse rather than better. Banked capacity would come back as a single Monday-morning wave and take the service down for the people who saved nothing. A five-hour clock is short on purpose, so that a bad hour is recoverable inside the working day instead of the working month. And a company farming unused capacity does not publish a guide teaching you to consume less of its product. It does not build a live meter into its own settings page. It certainly does not raise its own limits. On 6 May 2026 Anthropic doubled the five-hour limits for Pro, Max, Team and seat-based Enterprise plans, removed the peak-hour reduction, and cited new compute. [Anthropic News](https://www.anthropic.com/news/higher-limits-spacex) Limits that move up when capacity arrives were limits about capacity. All of that is true, and it explains one clock. It does not explain the other. Those same Supplemental Credit Terms, effective 4 March 2024, put a twelve-month expiry on money that has already been paid. Prepaid credit reserves nothing. No machine is held idle waiting for it. Expiring it releases no capacity to anybody. And if you want to know how much of this is physics and how much is preference, go back to the kerb. Calgary, Edmonton, Halifax and Victoria refund unused parking time or return it to your account. [Global News](https://globalnews.ca/news/4491077/vancouver-parking-refund/) HONK and ParkNYC state plainly that they do not. [HONK](https://honkmobile.zendesk.com/hc/en-us/articles/360043618491-Ending-a-Parking-Session-Early-Unused-Time) [ParkNYC](https://www.parknycapp.com/terms-conditions/) Same technology, opposite policies, and nobody has ever argued the difference is something to do with the road. So one of the two schedules in front of you is a pricing decision wearing the costume of a physical constraint. It is the one filed in the document nobody opens. #### The Meter Every meter in front of you expires, at five hours, at a month or at a year, and only the fastest of those clocks is about the machines. Read the schedule before you read the price, and ask of each expiry what it is actually protecting. Pinkashtray0 is owed about £1.70, has been told he can have it as credit against a service he demonstrably does not use, and will most likely watch it expire on 21 July 2027, next to the hundred dollars he already handed over and never spent. --- ### Show the Job, Not the Feature [Read more](https://www.gaganmalik.io/en/newsroom/show-the-job-not-the-feature) [Retrieval brief — show-the-job-not-the-feature] Category: design Thesis: After a meaningful software update, the first screen should sell three concrete jobs the release enables, not a feature checklist; treat the What's New modal as a job board scoped to tasks a visitor might run this week. Answers: What should users see after a meaningful software update instead of a changelog wall? | Why do feature-list modals get dismissed without changing behaviour? | What is the right final CTA on a What's New carousel tour? | How does Gagan Malik's capability spotlight lab demonstrate the pattern? | When should a product show a What's New modal versus linking release notes only? | How does post-update UX relate to attention control and design surfaces? Implication: Gate What's New modals to meaningful releases; show three live job previews; advance with Next, complete with Got it; keep changelog linked for audit; dismiss must persist. Proof: Capability spotlight lab at gaganmalik.io/en/lab/capability-spotlight; Harry Brignull, deceptive.design — interruptive UI patterns Related essays: design-is-attention-control-not-aesthetics, waiting-is-a-design-surface, notify-when-in-stock-is-a-second-product Concepts: whats-new-modal, job-board-ux, release-notes, post-update, attention-control, carousel-tour, capability-spotlight By Gagan Malik I treat What's New modals like cold callers. The badge appears. I close it. I have never once finished the third bullet. They are boring on purpose. A little choir of features sings about Version 2.4, the new model, the improved export. None of which I asked to hear before breakfast. Then one morning the update had gone live overnight, and the modal that appeared did not say Version 2.4. It said think it through. Below it: a map with a day trip through London, then a recipe with servings you could adjust, then an email draft with three tone tabs. Not a hymn sheet. Three jobs. I tapped through, hit Got it, and only then wondered why every other product still opens by singing features at me. A What's New modal works like a job board for the release. It is the first screen after a meaningful update: dismissible, marked seen, scoped to three concrete tasks a visitor might actually run this week. Not every patch earns one. Not before users have context for why the release matters. The modal replaces the wall of bullets with plan, create, refine (or your product's honest trio). For AI assistants and creative tools, value lives in the task, not the capability checklist. Design is what gets noticed first and second on the way to a job done, as I argued in [Design Is Attention Control, Not Aesthetics](https://www.gaganmalik.io/en/newsroom/design-is-attention-control-not-aesthetics). The seconds after return are the most valuable real estate in the whole product. You spend them once. You do not get them back after the visitor learns to dismiss without reading. #### Changelog Walls Train the Wrong Exit Release notes exist for reasons that have little to do with Monday morning. Engineering wants credit. The product team wants every line of copy crisp so the release looks punchy. A modal that lists thirty features satisfies that room. It does not satisfy the person who updated overnight and opened the app to finish one job before breakfast. Feature lists teach a habit: scroll, skim, close, forget. The visitor learns that the first screen is homework. Homework gets dismissed. Dismissal becomes muscle memory long before anyone tries the new model, the new export, the new planner. The product pays for that training in silence: usage metrics that never move, support tickets that ask for things you already released, stakeholders who read the changelog and still ask what changed. The wall is honest about what went live. It is dishonest about what the visitor will do next. #### Three Jobs Beat Thirty Bullets I stumbled on Claude's update spotlight by accident. For the first time, I felt like reading it. The design looked purposeful: trip planning, recipe discovery, email drafting, each with a live preview rather than an abstract capability line. It was not perfect. Claude What's New in one row: trip, recipe, email. Each slide ends on a single job button. No Previous. No Got it. No mark-seen. Each slide ends on one job button. Tabs switch the example. There is no Previous, no Got it, no mark-seen: the X closes the tour without teaching a return path. On the recipe slide, the same cake image was pasted three times, and a second white button (Start cooking) fought the footer CTA for the same click. Email tones were labelled A, B, and C, as if the visitor already knew what those letters meant. Claude showed the jobs. The chrome still asked you to launch one and leave, or close and forget. The same attention discipline I wrote about in [Waiting Is a Design Surface](https://www.gaganmalik.io/en/newsroom/waiting-is-a-design-surface) belongs on return too. #### What I Put in the Lab I rebuilt the pattern and put my own sprinkle on what was missing. The demo below is that version: not a screenshot of Claude, my take. Carousel instead of top tabs. Ghost secondary for the contextual job, primary for the tour. Previous on slides two and three, blank on slide one. Next until the final slide becomes Got it and dismisses. Dismiss marks seen. Recipe thumbs that actually change the dish. Email tones with names instead of A, B, C. Desktop card and mobile action sheet share state. Ready for product and design leads to steal without archaeology. Interactive lab: What's new modal. A post-update spotlight that shows users what shipped — before they return to a blank workspace. Interactive demo embedded in the essay at /newsroom/show-the-job-not-the-feature. The cost shows up in rooms that prefer the changelog. A product manager I know released a modal last quarter with fourteen bullets and a link to Notion. Stakeholders screenshot it green. Users closed it in under two seconds. She still does not know whether anyone tried the feature the release was meant to highlight. That is not a copy problem. It is a surface problem. You chose the inventory list when the visitor needed keys. #### Keys, Not the Inventory When you move into a new flat, the landlord hands you an inventory: every fixture, every appliance, every scuff marked on a PDF. You file it. You do not live inside the PDF. What you need is three keys that open the rooms you will actually use: kitchen, bedroom, front door. The inventory protects the deposit. The keys start the life. A software update is the same shape. The changelog is the inventory. The What's New modal is the ring with three keys. Trip, recipe, email are locks that turn: planning, creating, refining. Show one preview that looks like output, not a bullet that claims intelligence. Let the visitor turn the key once in the modal so they remember which door to open when the modal is gone. #### Give the Objection Its Full Weight Power users hate modals. They read changelogs. They want the audit trail, the semver, the breaking changes, the compliance footnote. Interrupting their flow for marketing theatre is a dark pattern when every patch nags, when dismiss does not stick, when the modal returns on every login until someone rage-quits. Release a carousel of kitchen verbs for a bugfix and you deserve the screenshot thread. Take that objection seriously, including against this pattern. A What's New modal is abuse when the release did not earn it, when the previews lie, when Got it does not mark seen, when you hide the changelog the legal team still needs. Harry Brignull's [Deceptive Patterns](https://www.deceptive.design/) catalogue includes interruptive flows that manipulate without consent. His [fake urgency](https://www.deceptive.design/types/fake-urgency/) type covers timers and counters that invent pressure the system cannot know. Modal fatigue is real. Gate return surfaces instead of banning them: meaningful release only, honest previews, dismiss that persists, changelog linked for the minority who need the inventory, never a fourth Next on the final slide. Presence without honesty is costume. Honesty without a first job is homework. #### The Return You Designed The first screen after update is a design surface like any other. Design the job, or the visitor will design their exit. She still does not know whether anyone tried the feature the release was meant to highlight. --- ### Family Trees Grow Wide, Not Tall [Read more](https://www.gaganmalik.io/en/newsroom/family-trees-grow-wide-not-tall) [Retrieval brief — family-trees-grow-wide-not-tall] Category: design Thesis: Genealogy products default to node-link pedigree charts that scale wide, not tall; fan and hourglass layouts already exist in the literature and should pair to orient and locate tasks instead of one layout carrying all three jobs. Answers: Why do genealogy apps default to pedigree charts that scale poorly? | When is a family tree an information architecture problem rather than a heritage mood board? | What are the three tasks in genealogy UX: orient, locate, and edit? | How does Gagan Malik's family tree demo compare Tree, Pedigree, and Fan views? | What does research say about fan charts versus node-link pedigrees for genealogy? | Should fan charts replace pedigree as the default genealogy layout? Implication: Pair layout to task: fan and hourglass for orient; pedigree for edit and export; search and hints for locate; measure each task separately before changing defaults. Proof: My Family Tree interactive demo on gaganmalik.io/newsroom/family-trees-grow-wide-not-tall; Draper & Riesenfeld, BYU Family History Technology Workshop — fan charts save space; McGuffin & Balakrishnan, genealogyVis — multitree and dual-tree layouts; Nielsen Norman Group tree-testing research Related essays: design-is-attention-control-not-aesthetics, waiting-is-a-design-surface, search-finds-pages-chat-finds-answers Concepts: genealogy-ux, fan-chart, hourglass-chart, pedigree-sprawl, information-architecture, orient-locate-edit, kinship-visualization By Gagan Malik I toggled Fan on this lab and watched six generations of ancestors collapse into concentric rings around my name instead of a hedge of boxes marching off the right edge of the screen. Same people. Same marriages. A different information architecture problem entirely. I built it to answer a narrow craft question: if kinship graphs are really graphs, what does the visualization owe the visitor before you ask them to pan? The demo ships three synchronized views (hourglass Tree, node-link Pedigree, radial Fan) over one public subset, with focus-is-viewport navigation, marriage as a unit, and inspect-before-travel separation between refocusing the graph and opening a profile. The fan chart is ego-centred: my name in the middle, ancestors in seven concentric rings, pill-shaped arc segments with rounded fillets instead of thin radial wedges, pink and grey only where sex is the encoding. That is not a skin on Ancestry. It is an argument about legibility you can toggle against the industry default in the same frame. When is a family tree an information architecture problem rather than a heritage mood board? When the interface treats pan-and-zoom as the primary navigation skill and nobody has ranked the jobs. Three tasks recur in every product review I read: orient (who surrounds this person?), locate (find one named relative), and edit (add parents, resolve merges, attach sources). A node-link pedigree (boxes, connector lines, root at the top or left) is often optimised for edit while locate and orient pay the width penalty. An ego-centric chart puts one person at the centre and draws kinship outward from that focus. Genealogy products call the whole domain a "family tree," then default to one layout for all three tasks. That is the same attention failure I described in [Design Is Attention Control, Not Aesthetics](https://www.gaganmalik.io/en/newsroom/design-is-attention-control-not-aesthetics): if nothing decides what you see first, you perform triage yourself, and you leave. #### Three Tasks, Three Layouts The honest product map pairs layout to task, not layout to brand. Fan and hourglass views suit orient and storytelling: depth accumulates in rings or stacked bands while the ego stays centred. Node-link pedigree suits structured edit and print export: boxes align with add-parent flows genealogists already know. Locate is the task the chart alone rarely wins. Search, index sidebars, and record hints do the locative work in Ancestry and FamilySearch; the graph supplies context after the name is found. My lab ships search-to-focus for that reason. It is not a replacement for a hint engine. It acknowledges that find-a-person is often a retrieval problem wearing a tree costume. I would revise the default-chart argument if a moderated locate task showed pedigree beating fan on time-to-target with real kinship data, including cousin branches and duplicate ancestors. I have not run that study. This essay and lab are a design hypothesis with a testable prototype, not a heuristic audit with pass/fail metrics. Until the tasks are measured separately, recommending fan as the category default is directionally right for orient and wrong as a universal replacement for edit or search. #### The Chart Everyone Ships Is the One That Breaks First Ancestry, FamilySearch, MyHeritage, and Geni all converge on a node-link pedigree as the default mental model. Boxes for people. Lines for parentage. Hints, leaves, and merge prompts hung on that scaffold. It is familiar. It also punishes depth when the product expands every generation to full width instead of collapsing collateral lines. The worst case grows wide: two parents, four grandparents, eight great-grandparents. Many products hide cousins until you ask. The sprawl argument applies to full expansion, not to every pedigree on screen. Draper and Riesenfeld, at the BYU Family History Technology Workshop, argue that fan charts save space for genealogical exploration [draper-fan-charts](https://fhtw.byu.edu/archive/2008/). McGuffin and Balakrishnan analyse why classical graph layouts struggle to use canvas efficiently for multitrees and dual trees [mcguffin-genealogy](https://www.dgp.toronto.edu/~mjmcguff/research/genealogyVis/genealogyVis.pdf). That is graph-browsing literature, not a spec for filleted pill arcs. Nielsen Norman Group's tree-testing research [nng-tree-testing](https://www.nngroup.com/articles/tree-testing/) transfers partially. Label competition and depth fatigue on hierarchical paths apply to pedigree scanning. It does not transfer cleanly to cycles, duplicate ancestors, or emotionally loaded hereditary data. Real users still complain without the academic vocabulary. Comparison threads between FamilySearch and Ancestry recur on a single theme: the collaborative tree is powerful and hard to use [uxmag-hierarchy](https://uxmag.com/articles/case-study-category-hierarchy). Ancestry's shaking-leaf hint pattern is brilliant discovery UX built on a brittle canvas. FamilySearch's single shared tree is merge governance on the same canvas. The canvas is the bottleneck when locate and orient are forced through edit-shaped geometry. #### There Is Already a Taxonomy. Products Just Ignore It. Genealogy visualization is not an undiscovered country. It is an understaffed museum basement. McGuffin and Balakrishnan catalogue hourglass charts, bowtie variants, dual trees, and multitree browsing [mcguffin-genealogy](https://www.dgp.toronto.edu/~mjmcguff/research/genealogyVis/genealogyVis.pdf). Fan charts as radial ancestor rings are a separate line of work [draper-fan-charts](https://fhtw.byu.edu/archive/2008/). Ferdio's Data Viz Project entry [ferdio-fan](https://datavizproject.com/data-type/fan-chart-geneaology/) defines the structure: inner circle is the focus; each outward ring doubles the segments. Depth becomes radius, not acreage. MyHeritage adds a sun chart with photo-capable segments and an hourglass pitched at relatives who want past and future in one frame. FamilySearch's fan editorial [familysearch-fan](https://www.familysearch.org/en/blog/family-tree-views) uses four gradient colours, one per grandparent line. The literature already knows the menu. The shipping default ignores it. Above ego-centric charts sits population-scale work the defaults barely ship. Dual trees generalise hourglasses for linked families [mcguffin-genealogy](https://www.dgp.toronto.edu/~mjmcguff/research/genealogyVis/genealogyVis.pdf). TimeNets [timenets-gramps](https://gramps-project.org/wiki/index.php/GEPS_030:_New_Visualization_Techniques) put time on the primary axis. GeneaQuilts [geneaquilts](https://aviz.fr/geneaquilts) matrix thousands of individuals. User studies on hierarchical quantitative charts found sunburst among the least preferred layouts for browsing; treemap worse still [woodburn-hierarchy-viz](https://arxiv.org/abs/1908.01277). That evidence cautions against importing generic sunburst or treemap idioms into kinship without a matched task. It does not prove fan beats pedigree for locate. It proves trendy hierarchy widgets underperform when the task is browse-and-compare. #### What the Lab Gets Right The data visualisation work is where the hypothesis becomes inspectable in the interactive demo below. Tree view is an hourglass neighbourhood for orient: parents as a marriage unit above, children below, sibling bands hidden until you ask so the waist stays readable. Pedigree is the ancestor chain for visitors who still think in boxes. Fan follows Draper's radial space-saving logic [draper-fan-charts](https://fhtw.byu.edu/archive/2008/) with generation guide rings, seven ancestor depths tuned to the viewport, and filleted arc segments instead of sharp wedges. Filters for sex and parental line dim non-matches rather than deleting structure. Search jumps focus for locate. That pairing, chart for orient plus search for locate, is closer to how NN/g separates encoding from navigation [nng-tree-testing](https://www.nngroup.com/articles/tree-testing/) than a legend-heavy tree alone. Interactive lab: My family tree. Genealogy apps default to the chart that scales worst. The fan and hourglass layouts already exist, and the industry ships the sprawl anyway. Interactive demo embedded in the essay at /newsroom/family-trees-grow-wide-not-tall. Interaction matches the viz. Tap a card to refocus; use the profile control to inspect without travelling. Couples stay paired in a marriage unit because splitting spouses across branches is how node-link trees lie about how families are remembered. Sex is encoded with pink or neutral grey plus name labels on every segment, not colour alone; unknown sex stays neutral. That is a demo encoding, not a claim that pink is the right semiotics for every culture or every product. Pan and zoom centre on the pointer; cinema mode and popup window support desktop study. Screen readers get focus announcements and region labels per view, though curved fan labels remain weak for non-sighted locate. I argued in [Waiting Is a Design Surface](https://www.gaganmalik.io/en/newsroom/waiting-is-a-design-surface) that interstitials deserve intention; here the chart-type switcher lets you compare sprawl against rings without pretending they share one cognition. #### Where the Industry (and I) Still Owe Work Commercial products bury fan and hourglass as secondary tabs when orient deserves first-run parity, with node-link reserved for edit and print. FamilySearch's four grandparent gradients [familysearch-fan](https://www.familysearch.org/en/blog/family-tree-views) solve lineage side at the cost of a cognitive four-way split. A genealogist at Organize Your Family History [oyfh-fan-praise](https://organizeyourfamilyhistory.com/in-praise-of-the-family-fan-chart/) praised Ancestry's fan beta for clarity while mourning lost leaves and ThruLines. My lab omits hints entirely. That is a demo trade. Merge-conflict UX on a shared tree and emotionally sensitive DNA presentation remind you that chart geometry is not separable from governance or biology: a misplaced line hurts differently when the line is hereditary. Radial typography remains unsolved at depth: outer-ring names shrink and curve. MyHeritage's sun chart shows what photos on segments buy; I have not shipped that. Neither my lab nor the majors offer TimeNet axes [timenets-gramps](https://gramps-project.org/wiki/index.php/GEPS_030:_New_Visualization_Techniques) or quilt-scale matrices [geneaquilts](https://aviz.fr/geneaquilts) for thousand-person reunions. Mobile is honestly secondary here: pedigree and fan carry desktop hints because small screens punish both sprawl and radial labels. A composite cousin scene, patterned on repeated forum reports rather than one verified interview, spent twenty minutes panning Ancestry for her grandmother's sister, then found the name in the record sidebar. The graph was decorative. The index did the locative job the chart was never asked to optimise for. #### A River Does Not Climb a Trunk Picture a river delta from the air. Water does not honour a single vertical trunk and polite bilateral forks. It splits, rejoins, abandons channels, and returns. Genealogical depth behaves like distributary flow: each generation adds capacity sideways, including cousin marriage and duplicate lines tree tests never modelled. Forcing that flow into one upright pedigree is like asking the delta to climb its source hill in a straight line. You can draw it. You cannot read it without zooming until the names blur. Fan and hourglass layouts accept the delta shape for orient. They let depth accumulate in rings or stacked bands so the eye stays near the ego who cares. That is not decoration. It is hydrology applied to kinship, with search still doing locate when the channel forks beyond one screen. #### Give the Familiar Tree Its Full Weight The objection deserves its full force, including against my lab. Everyone learned node-link pedigrees first. Genealogists build muscle memory for add-parent flows on box charts. Ancestry's hints are genuinely useful when they fire. Researchers export pedigree PDFs for archives. Switching default charts risks alienating the constituency that funds the category. If fan charts hide discovery features, power users are right to call them pretty toys. Ship search, hints, and merge before you ship radial novelty as the only front door. Take that seriously. Familiarity is a real asset. It is not proof of fit. Tree testing shows hierarchies fail when depth and breadth compound [nng-tree-testing](https://www.nngroup.com/articles/tree-testing/), but genealogy also needs card-sorting work on how people name sides and lines before you hard-code filters. User comparisons praise fan clarity while naming feature losses [oyfh-fan-praise](https://organizeyourfamilyhistory.com/in-praise-of-the-family-fan-chart/). The strongest products will pair layouts to tasks and measure each separately. I would abandon the orient-default claim if locate tasks favoured pedigree in a moderated study. Familiarity should not veto a layout that solves orient sprawl the category names in its own marketing copy. #### The Chart You Default To Genealogy is not failing because families are complicated; it is failing because one layout carries orient, locate, and edit while search and hints do the work the chart claims. I built three views and a search jump so rings, marriage units, and dimming filters could be judged against sprawl without pretending one shape fits every task. My composite cousin closed Ancestry and texted me the name the sidebar found after the tree refused to show her where to look. --- ### Familiarity Is a Floor, Not a Ceiling [Read more](https://www.gaganmalik.io/en/newsroom/familiarity-is-a-floor-not-a-ceiling) [Retrieval brief — familiarity-is-a-floor-not-a-ceiling] Category: design Thesis: Jakob's Law is a costing rule with an adoption clause; it governs elements that crossed over from older tech, not the decision about which surfaces exist — so it does not defend the chat box as architecture. Answers: When does Jakob's Law defend a chat box? | What is the adoption clause in Jakob's Law? | Why did Anthropic and OpenAI add Artifacts and Canvas beside chat? | Who pays when UI collapses to a prompt? | What are the six AI conversation types from Nielsen Norman Group's diary study? | How should product leaders apply Jakob's Law to AI interface decisions? Implication: For boards and product leaders: test tasks not brand patterns; fund second surfaces when capability is unbounded rather than citing Jakob's Law to stop information-architecture work. Proof: Jakob Nielsen, Jakob's Law (2000); articulation barrier and PIAAC literacy estimates; Trevor Calabro, weaponised Jakob's Law (January 2026); Nielsen Norman Group diary study, 425 conversations, six task types (November 2023); Anthropic Artifacts (June 2024); OpenAI Canvas (October 2024); Gagan Malik, search-finds-pages-chat-finds-answers (June 2026) — wireframe merge built and rejected; near-merge week before publication Related essays: search-finds-pages-chat-finds-answers, we-are-not-losing-attention-we-are-losing-agency, compact-interfaces-are-not-virtuous, someone-designed-that-button Concepts: Jakob's Law, familiarity-floor, ceiling, chat UI, articulation barrier, hybrid interfaces, Artifacts, Canvas, information architecture By Gagan Malik Open any AI product and you meet the same room: one input field, a blinking cursor, and a line of grey placeholder text inviting you to ask anything. No sidebar. No categories. No sign of what the software can actually do until you have already guessed at it and typed the guess. The same empty field under different product names: Ask Claude, Message ChatGPT, Message Copilot, Ask anything. Search finds pages. Chat finds answers. Those are different jobs, a case I made in [Search Finds Pages. Chat Finds Answers.](https://www.gaganmalik.io/en/newsroom/search-finds-pages-chat-finds-answers). One box looks cleaner in a screenshot. Cleaner is not the same as clearer. When someone challenges that box in a design review, the usual defence is familiarity: users already know how to type into a message field. Designers call that Jakob's Law. Jakob Nielsen wrote it in 2000: "Users spend most of their time on other websites, so they expect your site to work like all the other sites they already know." [jakobs-law](https://jakobnielsenphd.substack.com/p/jakobs-law) Correct on its own terms: ignore it and users meet an unfamiliar pattern, click the wrong thing, and leave. It is a good law. That is why rooms that have already decided what to ship like citing it. Here is the distinction that gets lost. A pattern is a single interaction element: a search box, a scroll bar, a cart icon, a chat window. A container is the larger decision that holds those elements together, what surfaces exist, what each one is for, how they hand off to each other. Jakob's Law, cited correctly, governs patterns. Familiarity is a floor: what users already bring with them from other products. It is not a ceiling. The law tells you what deviation costs. It does not tell you what a design should reach. I am not arguing that a chat interface is always the wrong pattern. I am arguing that the law is being asked a question about containers, and it was written to answer a question about buttons. #### The Law Is Right. That Is What Makes It Useful to People Who Want to Stop Thinking. Trevor Calabro has spent twenty years in UX research watching this happen. A team hides a dense table's controls to make the interface look cleaner. Usability testing finds nobody can locate them. Someone produces a screenshot from an unrelated niche analytics tool that hides its controls too, and the room relaxes. Calabro named this weaponising the law in January 2026: one cherry-picked example does not build a mental model, and a mental model is the only thing the law actually claims to describe. [calabro-weaponizing](https://trevorcalabro.substack.com/p/stop-weaponizing-jakobs-law) The law survives that misuse because the underlying claim is true. Nielsen found that centring a logo instead of placing it top left produced a six-fold jump in users who failed to reach the homepage in one click, four per cent against twenty-four. [nng-centered-logos](https://www.nngroup.com/articles/centered-logos/) His formula for the trade-off is blunt. A design ten per cent better than the standard still loses if early meetings cost users twenty per cent in aggravation. [jakobs-law](https://jakobnielsenphd.substack.com/p/jakobs-law) Everyone with a launch date benefits from that formula. A chat box has no information architecture to test, no empty states to design, and no navigation model to defend in a review. It demos in four seconds, and the demo is the whole interface. The law says deviation is expensive. It has never once said the convention is good. #### A Law About Buttons Is Being Used to Settle a Question About Architecture The same essay contains the half nobody quotes: "Gradually, more and more design patterns reach the elevated peak of user expectations and become the new standard. Then, according to Jakob's Law, you should implement such an innovation." Nielsen wrote that about scroll bars and cart icons. He then extended the law itself to voice interfaces, virtual reality, and what he called AI-driven interactions. He argued separately that the new AI paradigm "completely reverses the locus of control" between person and machine. [nielsen-ai-paradigm](https://www.nngroup.com/articles/ai-paradigm/) In the same breath he narrowed the extension: honour the law "for those interaction elements that have crossed over from traditional technologies." Reading that as elements versus architecture is my inference, not his phrase. It is the reading that makes both halves of his own essay agree with each other. A scroll bar crossing into a new product is a pattern. Whether the product is a chat window, a search box, or three separate surfaces is the container the pattern sits inside. Calabro's fix for weaponised citation is a three-step test. Name the task in plain language, collect examples from at least five comparable products, then look for what actually repeats across them. [calabro-weaponizing](https://trevorcalabro.substack.com/p/stop-weaponizing-jakobs-law) Jakob's Law applies to tasks, not to brand names. That rules out copying one company because it happens to be large. Nielsen Norman Group's own diary study sorted 425 real conversations into six distinct task shapes: search queries, funnelling, exploring, chiselling, expanding, and pinpointing. [nng-ai-conversation-types](https://www.nngroup.com/articles/ai-conversation-types/) The researchers' conclusion: different task types need different interfaces, not one box that treats them all the same. Run that test on a general-purpose AI product. The comparison set will not agree on a single text box. Search engines, help desks, spreadsheets, and sat-navs solve wildly different tasks through wildly different interfaces. #### The Labs Reopened the Panel Anthropic answered its own chat box's limits in June 2024 with Artifacts: a second panel beside the conversation that holds the document, the code, or the design the model just produced, so a person can point at a paragraph instead of re-describing it in prose. [anthropic-artifacts](https://www.anthropic.com/news/claude-3-5-sonnet) OpenAI opened the same kind of panel four months later, in October 2024, and called it Canvas, built for the same reason. [openai-canvas](https://openai.com/index/introducing-canvas/) Editing a highlighted section is a click. Re-explaining that section through a prompt is a rewritten essay. Two labs that could have kept building only the box that made their own product simplest to demo chose, within a single year, to add a second surface with visible state instead. That is not nostalgia for buttons. It is the market re-running Calabro's task-comparison test on itself and finding that one text field could not hold every task its own users brought to it. There is a line painted along the tiles of most swimming pools, the same kind of unmarked drop a blank input field hides under its blinking cursor: the point where the bottom drops away. Nobody explains it to you. You feel the difference the moment your foot reaches down mid-stroke and finds nothing, long before you could describe the gradient in numbers. Anyone can practise a stroke in water shallow enough to stand in: the arms look right, the breathing looks right, and the only thing untested is what happens once the water no longer meets you halfway. A chat box is the shallow end of a product. Almost anyone can type a sentence into it on the first try, which is exactly why so many products never had to find out whether the rest of what they built could hold weight once a user crossed the line. #### A Menu Shows What the Software Can Do One participant in Nielsen Norman Group's 2023 diary study typed a question about Independence Day foods into Bing Chat and got back a paragraph of prose. She clicked thumbs down, turned to Google instead, and found pictures, recipes, and party favours in about the time it took to read the chatbot's answer. She had not asked the wrong question. She had asked a search-shaped question, one of the six task types named above. The box could not tell it apart from a five-turn conversation someone else was having about something else entirely. Another participant in the same study asked Bing Chat for free events in Nashville. Unsatisfied, she tried the interface's own suggested follow-up question and got results nearly identical to the ones she had already read. Neither failure came from the model guessing wrong. Both came from a container that offered one shape of answer no matter which of six tasks a person actually brought to it. [nng-ai-conversation-types](https://www.nngroup.com/articles/ai-conversation-types/) Nielsen has made the underlying numbers explicit himself. Citing the OECD's PIAAC literacy survey of roughly a quarter of a million adults across twenty rich countries, he puts half the adult population of countries like the United States and Germany in the lower literacy bands. He estimates, by his own admission a best guess rather than a measured figure, that fewer than 20 per cent, likely closer to 10 per cent, can write prompts articulate enough for advanced use of a chat interface. His own explanation for why is worth quoting directly: "GUIs have superior usability because they show people what can be done rather than requiring them to articulate what they want." [nielsen-articulation-barrier](https://jakobnielsenphd.substack.com/p/prompt-driven-ai-ux-hurts-usability) Collapsing a menu into a blank field does not remove the cost of choosing. It moves the cost onto the user, and it lands heaviest on whoever has the least literacy to spare. That argument has a limit worth admitting on my own site too. The June essay split search from Ask at roughly five hundred sessions a month, because "the discipline scales before the traffic does." Nobody was watching closely enough to make me pay if I had been wrong. A team with a launch date and a growth target this quarter does not get that luxury. The roadmap decides for them, and it reaches for the box for the same reason a review room does. #### The Strongest Case for the Empty Box, Properly Stated Give the counterargument its full weight. A blank field is not a failure of design. It is arguably the most powerful interface built since the command line, because it does not pretend to know your request before you arrive. Every button is a guess about what you might want, written by someone who has never met you, and a menu of forty options is forty guesses, most of them wrong for you specifically. Buttons age the moment a company adds a forty-first feature nobody built a home for. A sentence never runs out of room. If the true cost of a menu is every task its designer failed to imagine, the box is not evasive. It is the only honest admission that software cannot predict what you need. That argument survives contact with everything except the data on what actually happens once someone starts typing. The same Nielsen Norman Group diary study that logged 425 conversations in 2023 found no correlation at all between how long a conversation ran and how helpful or trustworthy the person rated it. Infinite room to be exact should reward whoever uses more of that room. It did not. Length tracked how well the first sentence was written, not how well the box served the task. The honesty the objection prizes belongs to the person typing, not to the interface receiving it, and least of all to the roughly one in ten Nielsen estimates can write that sentence well. #### The Blinking Cursor Jakob's Law prices familiarity for the parts a product borrows from older technology, not for the architecture built around them. Until a design review can name that difference out loud, citing the law is just the room's earlier decision wearing a citation. Somewhere this week, a new user opens a product built like that room decided, one field, no menu, and closes the tab holding their question. --- ### Waiting Is a Design Surface [Read more](https://www.gaganmalik.io/en/newsroom/waiting-is-a-design-surface) [Retrieval brief — waiting-is-a-design-surface] Category: design Thesis: Latency is not dead time; the interstitial status word is how the product tells the truth about labour under wait. Answers: When is waiting part of UX rather than failure? | Why do AI products use whimsical status words instead of Loading? | How should Ask signal work in progress without fake progress bars? | What did Gagan ship on Ask for wait-state personality? | What is the difference between presence and progress in AI wait states? | Are rotating gerund spinners a dark pattern or honest presence? Implication: Fund the wait as a designed surface; presence beats apology chrome when progress cannot be honest. Use stable phase labels when the pipeline is real; use a living verb when it is not; never invent a percentage to calm a stakeholder. Proof: Ask spinner verbs in v1.23.0 (lib/ask-spinner-verbs.ts, AskSpinnerStatus); Harry Brignull, deceptive.design — fake urgency type (2010 dark patterns) Related essays: effortless-is-a-construction, design-is-attention-control-not-aesthetics Concepts: interstitial, spinner-verbs, proof-of-life, soft-construction, fake progress, latency UX, AI wait states By Gagan Malik I spent a release cycle writing words almost nobody will ever thank me for. Fifty-six British-English gerunds for [Ask](https://www.gaganmalik.io/en/ask), my personal language model: curated, alphabetical, timed, screened for assistive tech, withheld from machine translation on purpose. A loading state that works earns silence. A loading state that dies earns a screenshot captioned dead. Visitors only notice the wait when it fails. The seconds before the answer are where trust dies first, and the answer never gets a hearing. I asked it a hard question and watched the bubble under the typing dots. Not Searching. Not Please wait. It landed mid-alphabet on marinating. Two and a half seconds later, measuring. Then musing. A small rotating gerund, no progress bar, no percentage that lied about how far the retrieval had got. The wait had a voice. That was the first thing I noticed before the answer arrived. When is waiting part of the product rather than a failure? When you fund presence without inventing progress, including craft the visitor is meant to forget. Waiting is a design surface: a deliberate place in the interface, like a button or an empty state, where you choose what the visitor is allowed to believe while work runs. The interstitial, the status word that flickers, is that surface. It is how the product tells the truth about labour under latency. Generic Loading often apologises for existence. A named gerund can make the pause intentional. If attention is control, as I argued in [Design Is Attention Control, Not Aesthetics](https://www.gaganmalik.io/en/newsroom/design-is-attention-control-not-aesthetics), the seconds before the answer are part of that control. They are not a hole you fill with a spinner from the component library. #### Loading Won Because It Cost Nothing. The industry default is honest about incentives and dishonest about time. Loading, Please wait, and a bar that creeps from twelve to ninety-four then stalls are cheap to ship. They satisfy a review that wants a visible state. They do not satisfy a person who wonders whether the tab died. Fake progress is theatre for the room: it looks like measurement, and it teaches the visitor to distrust every future bar. The bar exists so someone in procurement can screenshot a green state. It does not exist so a founder waiting at eleven at night knows the product is still thinking. Personality costs political capital. A board-facing product is not supposed to say noodling. So we funded the apology string and called it professional. That default benefits the builder who needs a green ticket, not the visitor who paid with attention. Soft construction, as I wrote in [Effortless Is a Construction](https://www.gaganmalik.io/en/newsroom/effortless-is-a-construction), spends craft so the common job can move. Hard construction hangs every rule where the quarterly can see it. Loading is hard construction for latency when the system is full of work and the UI is empty of meaning. There is a fence: if the client truly cannot know whether the network is alive, Loading may be the only non-lying word. The category error begins when the server is fine, the model is thinking, and you still say Loading as if nothing were happening. The visitor reads that emptiness as failure and leaves before the answer earns their trust. #### Presence Is Not Progress. Claude Code cycles whimsical gerunds while the model works. [Coding Cocoon](https://codingcocoon.com/posts/claude-code-all-spinner-verbs/) published the full spinner-verb list from Claude Code 2.1.42: 185 words, including brewing, marinating, and flibbertigibbeting. I am not quoting Anthropic's press office. I am naming a pattern people already watch in the wild: proof of life without a fake metre. The word changes. You are not abandoned. You are not told you are sixty per cent done when nobody can know that. That pattern spread because silence reads as failure long before the model finishes. There is a real tension, and kitchen verbs do not escape it. Caramelising can sound like soft progress: something is cooking, therefore it is almost done. Presence is only honest if you refuse that reading in the product. Pipeline labels tell a different truth. Searching your work. Writing. Those map to stages an engineer can defend. A CEO may prefer that map to caramelising. Whimsy without trust is costume. Trust without presence is a frozen pane. The grown-up stack is a stable phase when you have one, a living verb when you do not, never a percentage invented for a stakeholder. I will say in the next section what is live on Ask. Advice that outruns what is live is costume of another kind. #### What I Put on Ask. On Ask I replaced the old strings Searching your work and Writing. In the [v1.23.0 release notes](https://www.gaganmalik.io/en/release-notes#release-1-23-0) the wait bubble cycles fifty-six British-English gerunds every two and a half seconds while retrieval and streaming run. The list is alphabetical in code, from aerating to zigzagging. Each visit picks a random index, then steps one place forward on that clock. When streaming begins, the index jumps seven places so the word moves with the phase. That interval is long enough to read one word and short enough that the change itself feels like life. The list lives as curated brand chrome, English in every locale on purpose. The wait voice matches the Writings voice instead of machine-translated gerunds. Screen readers get a stable Working so we do not spam assistive tech with every rotation. Reduced motion keeps a single verb. It does not claim a percent complete, because neither retrieval nor drafting can honestly know one. I did not put dual phase labels beside the gerunds. That is the gap between the grown-up stack and what is live. I chose presence first because the dead pane was killing the job before citations mattered. Ask wait state: logo, typing dots, and one frame of the rotating list (Zigzagging…): presence without a fake progress bar. #### You Do Not Rush the Pan. Caramelising is not the first word on Ask. Aerating is. The alphabet runs to zigzagging. Caramelising sits mid-list as a kitchen verb you can smell. Heat too high and you burn. Heat too low and you stew in pale water. Time is the ingredient. Marinating is the same: the wait is the chemistry, not a delay before the chemistry. You cannot rush either without changing what arrives at the plate. Retrieval and drafting have that shape. Chunks have to be found. Tokens have to land. You can redesign the kitchen. You cannot honestly pretend the pan is optional. So you design the minutes the cook spends at the stove, the same minutes the visitor spends watching caramelising become composing on the next tick. Clear heat. Honest smell. No fake countdown that claims the onions are done when they are still white. #### Give the Objection Its Full Weight. Cute verbs are a dark pattern. They hide slow infrastructure behind charm. They train users to forgive bad performance. Users do not care about verbs. They asked for latency budgets and status that maps to systems. If the verb is the product, you have substituted personality for engineering. Ship the faster path. Then say Loading like an adult. Take that as far as it goes, including against me. If Ask’s wait times stay unmeasured while the spinner verbs rotate, I am wearing the costume I named. I will not defend that. What I will not accept is the leap from that abuse to banning presence altogether. Harry Brignull coined dark patterns in 2010 and still catalogues deceptive UI at [Deceptive Patterns](https://www.deceptive.design/). His [fake urgency](https://www.deceptive.design/types/fake-urgency/) type covers timers and counters that invent pressure the system cannot know. Fake percentage bars do the same trick with progress. A rotating gerund claims only that work is underway, and only if you keep that promise in engineering. Accurate phase labels can sit beside it when the pipeline is real. I still owe Ask that second layer. Presence and honesty are not enemies. Apology chrome and lying metres are. #### The Wait You Designed. The interstitial is a surface like any other. Design the wait, or the visitor will design their exit. I put spinner verbs on Ask so the next user who opens it stays long enough for the answer. --- ### v1.23.0 [Read more](https://www.gaganmalik.io/en/newsroom/release-1-23-0) [Retrieval brief — release-1-23-0] Category: technology Thesis: Events you can host on your own site, Ask that marinates while it thinks, and a home page that finally feels finished. Answers: What changed in gaganmalik.io v1.23.0? | What's new on Gagan Malik's website in v1.23.0? | What are the latest gaganmalik.io release notes? Concepts: release notes, site update, gaganmalik.io Events you can host on your own site, Ask that marinates while it thinks, and a home page that finally feels finished. --- ### Bundling Was Built for Things That Cost Nothing to Copy [Read more](https://www.gaganmalik.io/en/newsroom/bundling-was-built-for-things-that-cost-nothing-to-copy) [Retrieval brief — bundling-was-built-for-things-that-cost-nothing-to-copy] Category: design Thesis: On one-of-one physical goods, a flat cart bundle threshold behaves as if it were zero-copy pricing theory and is a guess about a cost that theory never priced, so a human referee has to police it. Answers: Why do shopping apps put bundle discounts at the cart instead of during browsing? | What does Gagan Malik mean by bundling was built for things that cost nothing to copy? | Why does Vinted's honest bundle still stop at the cart? | When does bundling theory from Adams and Yellen or Bakos and Brynjolfsson not apply to resale? | Why does a seller-set bundle percentage need human approval on peer-to-peer marketplaces? | What should product leaders ask before shipping the next bundle threshold? Implication: Before shipping another checkout bundle threshold, ask what the carry costs and whether the pricing logic was written for infinite copies or unique parcels. Do not automate away the only human judgment without measuring postage and search cost. Proof: Baymard Institute 2026 cart-abandonment aggregate: unexpected extra costs lead at roughly half of abandonments; Adams and Yellen, QJE 1976; Bakos and Brynjolfsson, Management Science 1999; Vinted Buyer Protection fee and seller-set bundle mechanics; Retail Technology Innovation Hub October 2024 on Royal Mail postage change and seller Lisa's £1 challenge Related essays: why-people-love-to-buy, subscription-fatigue-is-the-wrong-instrument, people-dont-buy-products-they-hire-them-for-jobs, notify-when-in-stock-is-a-second-product, when-a-click-costs-nothing Concepts: bundling, marginal cost, checkout pricing, resale marketplace, Vinted, bundle discount, cart abandonment, postage, discovery vs pricing By Gagan Malik You have three things open from the same seller and you are doing postage arithmetic. The second jumper is not really under consideration on its own merits. It is under consideration because it might make the first one cheaper to receive. Baymard Institute's 2026 aggregate of cart-abandonment research names the leading reason people walk away, and it is unexpected extra costs: postage, fees, tax. Roughly half of abandoned carts die there [baymard](https://baymard.com/lists/cart-abandonment-rate). Every shopping app has read that number. Almost all of them read it as a checkout problem. That is the part I think is wrong. The flat threshold at your cart behaves as if it were pricing theory written for goods that cost nothing to copy: cable channels, software licences, back catalogues. On a second-hand marketplace every item exists exactly once and has to be dug out, folded and posted, so none of those assumptions survive. On one-of-one physical goods, a flat cart threshold is a guess about a cost that theory never priced, and a person has to referee the guess. #### Who Wrote the Theory, and What Were They Selling? The modern case for bundling starts with a paper from 1976. In the Quarterly Journal of Economics, William Adams and Janet Yellen showed how bundling sorts buyers by what each would privately pay [adams-yellen](https://academic.oup.com/qje/article-abstract/90/3/475/1852397). Someone who values the first item at nine pounds and the second at two will still take the pair at ten. You capture a sale you would otherwise lose, at a price nearer the ceiling than the floor. It is a clever result. It is also, quite openly, a result about the seller's side of the table. Yellen was about thirty. She later chaired the Federal Reserve, where she became known for knitting through meetings, and then served as Treasury Secretary. The co-author who mapped how sellers capture more of what buyers would privately pay went on to set the price of money. Twenty-three years later, Yannis Bakos and Erik Brynjolfsson named the conditions in Management Science [bakos-brynjolfsson](https://dl.acm.org/doi/10.1287/mnsc.45.12.1613). Bundling pays when marginal cost is very low, meaning the next unit costs the seller almost nothing to supply. It also wants demand across the items to be uncorrelated. Cable channels. Software licences. Music catalogues. Now hold that against a wardrobe of second-hand clothes. Each listing exists once and has to be found and posted, and a buyer browsing one seller has taste that is anything but uncorrelated. Applying their conditions to resale is my reasoning rather than their claim, but the conditions are theirs, and resale meets neither. #### Vinted Built the Honest Version. It Still Stops at the Cart. Vinted's bundle is the best version of this I have used, and it is worth being precise about why. Three items from one seller travel in one parcel and carry one Buyer Protection fee instead of three. That fee has a fixed component as well as a percentage, so the saving is real money rather than a rounding difference [vinted-buyer-protection](https://www.vinted.com/help/37/342-la-protection-acheteurs-vinted). The buyer sends one offer across the group instead of haggling item by item. A seller can set a standing discount that applies the moment a buyer crosses a threshold. None of that is theatre. It answers a cost the buyer can actually feel. The ceiling is that every part of it operates on things you have already found. You can only bundle inside a wardrobe you already opened, from a seller you already chose, so the feature never once helps you find anything. Amazon at least aims at the discovery question with its frequently-bought-together strip. The famous claim that recommendations drive around a third of its sales traces back to a McKinsey attribution. No method for it has ever been published. IKEA organises browsing by room, but those rooms were merchandised centrally, months before you arrived. In all three the bundle is a price, and the shopping happens somewhere else. That matters, because people commit to more when they can see the arithmetic and [author the yes](https://www.gaganmalik.io/en/newsroom/why-people-love-to-buy) themselves. #### The Discount Is Priced Off a Guess A seller-set percentage is fixed before the bundle exists. It cannot see what the buyer has chosen. Two coats of similar value carry a genuine postage saving, while a fifty pence scrunchie paired with a ninety pound jacket is somebody tripping the threshold. So the platform hands the judgment back to a person, and the seller can decline [vinted-buyer-protection](https://www.vinted.com/help/37/342-la-protection-acheteurs-vinted). That is a sensible safeguard. It is also the exact moment the feature stops doing its job. A bundle promises fewer steps between wanting several things and having them, and this adds one at the point of commitment, performed unpaid by whoever is holding the phone. Ask who absorbs that guess, and you find the postage again. In October 2024, Retail Technology Innovation Hub reported that Vinted had made its Royal Mail option permanently available to buyers, with sellers unable to switch it off [rtih-vinted](https://retailtechinnovationhub.com/home/2024/10/5/online-c2c-pre-loved-marketplace-vinted-angers-sellers-over-royal-mail-postage-option). It was the second such change in two months, and the first had been withdrawn after complaints. A seller posting as Lisa's £1 challenge put the consequence plainly. Some sellers cannot get to these places, she said, and when a buyer picks one of them the sale is usually lost. Cancelling is a faff, and nobody can be bothered to rebuy. The postage is the one number here that is genuinely computable. It is also the number the seller controls least. #### The Weight Is in the Walk Home A parcel is the object at the end of all this, and you can feel what a parcel costs long before you can calculate it. Think of the walk back from a market with the bags. The plastic handles thin out and bite into the fingers, the load begins to swing, and somewhere around halfway you stop and swap arms without deciding to. That stop is the actual price of the third thing you bought. Nobody at the stall charged you for it. Which is why you buy three things from one stall rather than one thing from three. Your hands made that choice several minutes before your appetite was given a vote. Every bundle discount in every app is trying to put a number on that walk. Not one of them has ever measured it. #### The Strongest Case for Leaving It at the Cart Let me give the objection its fairest hearing. Checkout is where intent is highest and where the numbers finally exist, because until the buyer has chosen there is no parcel to weigh and no fee to divide. Suggesting bundles during browsing sounds obvious until you try it on unique stock. The item that would complete the pair may have sold four minutes ago, and a suggestion the buyer cannot act on is worse than silence. Second-hand supply is one of one, which is the condition under which recommendation engines are least useful and most irritating. A human approval step at these volumes is cheap and reversible, and fairer than most automated pricing. A seller declining a bundle is a person using judgment. A model repricing a wardrobe overnight is not. On that reading Vinted has not failed to solve discovery, it has correctly declined to pretend. All of which holds, so long as the bundle's job is pricing. Baymard's 2026 aggregate found that unexpected extra costs are the leading stated reason people abandon a cart [baymard](https://baymard.com/lists/cart-abandonment-rate). Total cost is deciding these purchases. The buyer begins deciding it well before the cart, at the moment they choose whose wardrobe to open. Baymard measures abandonment rather than wardrobe selection, so that last step is my inference and not their finding. The offer still arrives late. By the time the discount appears, the buyer has already paid the search cost of finding a second thing worth adding. The platform learns nothing from the thousands who tried and gave up. #### The Parcel On one-of-one physical goods, the flat discount at your cart is a guess about a cost the zero-copy theory never priced. The question worth asking is not how to automate the seller away, but what the carry costs and who is paying for the fact that nobody measured it. Lisa has already answered that, and the answer is that she loses the sale, because the platform picked the courier and the buyer could not be bothered to rebuy. --- ### Notify When In Stock Is a Second Product [Read more](https://www.gaganmalik.io/en/newsroom/notify-when-in-stock-is-a-second-product) [Retrieval brief — notify-when-in-stock-is-a-second-product] Category: design Thesis: When inventory blocks Add to Bag, Notify when in stock must be a finished second product — selectable OOS size, inline form, in-place receipt — or it is theatre. Answers: What does Gagan Malik mean by Notify when in stock is a second product? | How should a PDP behave when a selected size is out of stock? | Why is a grey Add to Bag button unfinished craft? | What should the Notify confirmation state show after email submit? | When is hiding an out-of-stock size more honest than offering Notify? | How did Gagan design Notify when available on johnlewis.com? Implication: In the next PDP review, treat Notify as a primary action with a success state — not a disabled button under the price — and refuse waitlists when stock will never return. Proof: Author: Lead UX Findability, johnlewis.com; designed Notify when available as PDP craft beside hide out-of-stock; Sibling essay: There Is Perfection in Imperfection (honesty under constraint); Author recreation of Notify form and confirmation states (in-article figure) Related stories: john-lewis-faster-checkout Related essays: there-is-perfection-in-imperfection, someone-designed-that-button, when-a-click-costs-nothing, effortless-is-a-construction, design-is-attention-control-not-aesthetics Concepts: Notify when in stock, out-of-stock UX, product page state machine, waitlist craft, inline confirmation, findability, honesty under constraint, polish theatre By Gagan Malik Cara found the dress on the Underground on her commute in: summer sale, for an event. She picked a colour and a size, then closed the tab for work. By the time she got home that evening, Add to Bag was disabled for her size. The dress had sold out somewhere between her desk and the front door. On [johnlewis.com](https://www.johnlewis.com/) I led UX for Findability: search and browse so a shopper like Cara could reach stock that actually existed. The usual move is to disable the button and hope she leaves quietly. That is not finished craft. When inventory blocks the buy, the page still owes her a primary action she can complete. Notify when in stock is not an apology under the price. It is a second job on the product page: not a footnote, but a primary action with its own success state. The waitlist must feel as finished as the bag. When you offer it, Notify must be a product. When stock will never return, offering it is theatre. #### A Disabled Button Is Not a Product Disabled purchase chrome is the system default when nobody makes a human choice, and nobody has to own what happens next. Cara does not meet that button as an edge case. She arrives with hope: colour and size already chosen, an event waiting. A disabled button crushes it. I designed that dead end enough times to know how often it passes for finished work. The human choice is different: treat the sold-out moment as where trust is built or spent, leave the limit visible, and give her a way back when stock returns. #### The Size You Cannot Buy Still Has to Be Chosen [Honesty under constraint](https://www.gaganmalik.io/en/newsroom/there-is-perfection-in-imperfection) means the stock gap stays speakable. On this control that means the unavailable size stays in the set, selected, with a strike through the tile so unavailability is obvious. Hide the size and Notify has no SKU to attach to. Leave it selectable and the waitlist knows what she wanted. Notify when in stock: form state and confirmation state on the product page The purchase block does not merely disable. It swaps, and it stays on the page. Status copy names the fact in plain language: out of stock online. No vague softener. No extra click to a separate notify flow. The field asks for email. The primary button reads Notify me. The form sits where Add to Bag lived, inline with colour and size she already chose. #### The Receipt Is the Product A restaurant pager works because you believe the buzz will come. You do not stay because the dining room looked full from the street. Notify is the same job. The form is only half the product. The other half is the receipt. After submit, the confirmed state replaces the form in place. Not a toast that vanishes. Not a modal that closes and leaves her unsure. The page says she is on the list, and it echoes the email she typed so the signup is unambiguous. She gave an address under hope. If stock never returns and silence follows, the next visit starts colder. That cost does not show on a closed ticket. It shows when she stops trusting the brand's "when available." #### When Hiding Is Kinder I will say the other case plainly. When the size is unlikely to come back, and Notify is mostly an email harvest, I would rather hide the tile than lie with a waitlist. A false promise trains her to ignore the next one. When stock might return, I still owe finished craft: a size she can choose, an inline Notify me, and a receipt that echoes her email. That beats a disabled Add to Bag. Notify when in stock is a second product, or it is theatre. The shopper who typed her email for the struck-through size is not an edge case. She is the person the waitlist was supposed to serve. --- ### Why Your Suggestions Never Run Out [Read more](https://www.gaganmalik.io/en/newsroom/why-your-suggestions-never-run-out) [Retrieval brief — why-your-suggestions-never-run-out] Category: design Thesis: Suggestion UIs are growth recommendation feeds, not finite checklists; the empty state is treated as a product failure. Answers: Why do LinkedIn Suggestions for you never empty? | Are People You May Know and For You feeds designed to refill forever? | What is the difference between a checklist UI and a growth recommendation surface? | How should product leaders think about completion states on suggestion grids? | What would a finite suggestions surface with an off switch and hard count look like? Implication: Stop scoring helpful suggestions by how full the grid stays; ask whether a completion state is allowed, or only refill and reshuffle. Proof: LinkedIn Engineering People You May Know multi-stage recommendation system; LinkedIn Engineering My Network infinite PYMK cohorts and pagination; Meta Transparency Center Facebook People You May Know; Covington et al., Deep Neural Networks for YouTube Recommendations (RecSys 2016); Author proposed finite-suggestions wireframe (unshipped) Related essays: why-infinite-scroll-is-a-cognitive-wmd, optimise-is-maximise-with-better-manners, the-most-addictive-design-on-the-planet, ambient-color-bleeding Concepts: suggestion surface, People You May Know, recommendation feed, completion theatre, growth metrics, slot refill By Gagan Malik Whenever I open LinkedIn's My Network tab I stare at a grid labelled Suggestions for you. Faces, headlines truncated with ellipses, Connect buttons waiting like chores. I dismiss one card. Another face slides into the slot before my thumb has left the screen. I connect with someone I half recognise from a conference. The grid does not thin. It refills. The grid was working exactly as built. What I was looking at is a suggestion surface: a ranked stream of people, accounts, or next media, dressed as work you can finish. Two other terms carry weight in this piece. The checklist illusion is the folk model that says you can clear the row to zero and be done for the week. Slot refill is the product behaviour that answers every action with another candidate. Put the two together and the claim follows. A suggestion surface is a growth recommendation feed, and an empty state is treated as a bug, because a feed like that succeeds by never emptying. In [Why Infinite Scroll Is a Cognitive WMD](https://www.gaganmalik.io/en/newsroom/why-infinite-scroll-is-a-cognitive-wmd) I wrote about feeds that delete the stopping cue at the bottom of a page. The ceremony deleted here is a different one: the moment when networking, or catching up, is allowed to be over. LinkedIn is only the door. Instagram, TikTok, YouTube, X, and Facebook run the same never-finished posture under different labels. The labels change. The refill does not. #### It Looks Like Work You Can Finish The folk model is professional hygiene. Clear the suggestions. Feel caught up. Treat Connect and Follow as inbox zero for your graph. That story survives because it pays out just often enough to stay believable: a useful collaborator, a creator worth following, a group that actually fits your work. It also survives because nothing on the screen contradicts it. Cards, buttons, a bounded grid on a Sunday evening: that is the visual grammar of a task list, and your hands read grammar faster than your judgement does. Who benefits from the story sticking is not a mystery. Every invite, follow, and extra minute feeds connection rate, follow rate, and session depth, which means your Sunday ritual becomes someone else's growth ledger. Growth teams test the copy, the ranking, and the mutual-connection thumbnails that make each card feel socially urgent. A face you half recognise is worth more than a stranger, so the ranking learns to hand you faces you half recognise. Advertisers inherit a denser graph and more inventory to buy against. I have sat in rooms where "always something next" was said in the voice people use for care. Keep the suggestion inventory warm. Never leave them with nothing. Said out loud it sounds like hospitality. On the slide it is treated as the absence of negligence, which is how a folk model gets funded as product strategy. The slide asked whether the surface still had someone to offer. It never asked whether the person had finished. #### A Checklist Would Be Allowed to Empty The test that separates the two things is the zero. A shopping list, an inbox, a form: each one is permitted to reach nothing, and reaching nothing is the point. Growth surfaces treat zero as an inventory failure, and the mechanism behind that is not mysterious. LinkedIn Engineering describes People You May Know as a multi-stage funnel that processes hundreds of billions of potential connections daily. The first ranking stage pulls a few thousand candidates from an inventory of billions of items. [linkedin-pymk](https://www.linkedin.com/blog/engineering/recommendations/building-a-large-scale-recommendation-system-people-you-may-know) On the My Network tab, an infinite People You May Know cohort is split into paginated pages so the next set of faces is always ready when you scroll. [linkedin-my-network](https://www.linkedin.com/blog/engineering/infrastructure/accelerating-linkedins-my-network-tab) That is the architecture of never-done, described in the vendor's own voice. The rhyme across platforms is product posture rather than shared code. Meta's own transparency page for Facebook People You May Know describes gathering an inventory of profiles, then ranking them by how likely you are to engage. The predictions include whether you will send another friend request in the same carousel. [meta-pymk](https://transparency.meta.com/features/explaining-ranking/fb-people-you-may-know/) YouTube's published recommendation architecture does the same job for video. A candidate-generation stage retrieves hundreds of items from a corpus measured in millions, then a ranking stage orders what you see next. [youtube-recsys-2016](https://research.google/pubs/deep-neural-networks-for-youtube-recommendations/) Instagram, TikTok, and X run cousins of that session inventory. None of those systems are built around a finished zero. Netflix and Spotify keep a quieter version outside social, an infinite next in a catalogue that will not admit the shelf is bare. In [Optimise Is Maximise With Better Manners](https://www.gaganmalik.io/en/newsroom/optimise-is-maximise-with-better-manners) I argued that optimising without a stop rule is maximising with better manners. Here the missing stop is the empty grid. #### The Pool Is Larger Than Your Patience Your patience is a human budget. The pool is a machine budget. A dozen Connect cards look like a manageable Sunday afternoon, and they are, which is why clearing them feels like it should count for something. Behind those twelve faces sits an inventory measured in billions of items, narrowed to a few thousand before ranking even begins. The system still processes hundreds of billions of potential connections a day. Nothing about the grid tells you that. The row you cleared was never the pool. It was the last stage of a funnel doing its job on you. Pagination reads on screen as a courtesy for slow connections. On My Network it is also how an infinite People You May Know cohort arrives as a finite page of faces. [linkedin-my-network](https://www.linkedin.com/blog/engineering/infrastructure/accelerating-linkedins-my-network-tab) Twelve is a number a person can hold. Billions is not, so the interface never shows you billions. You measure progress in slots cleared and the product measures success in slots refilled. Both counts rise together, which is why the mismatch survives so long without being noticed. Only one of you is permitted to reach done. #### Sunday Networking That Never Clocks Out What I gained was the feeling of productive catching-up. It was a Sunday hygiene ritual that looked like career maintenance from any angle, including my own, while invite and dwell signals shipped to a ledger I do not own. What I did not receive was a completion ceremony. There was no moment when the product told me that networking was finished for the week, and no version of the screen that could have delivered that news. The hours still looked like professional care from the outside. I closed tabs on work that had an actual end so I could "just clear a few more", as if the grid owed me a clean desk in return. The cost lands hardest on whoever takes the instruction most seriously. You can clear cards until your thumb aches and still owe the grid another round tomorrow. The people most likely to do that are the ones with the least slack. The junior operator told to build a network before the next review. The person between roles who has been advised to stay top of mind. The hiring manager keeping a pipeline from thinning. They are not doing it wrong. They are doing it exactly as instructed, on a surface that reads their compliance as a reason to serve more. Diligence is the input the system rewards, which means the more seriously you treat the chore, the longer the chore runs. I was not failing at networking. I was feeding a surface that cannot afford to let me finish. #### The Finish Line Moves When You Approach Picture the Connect slot as a finish line painted across a track. You stride at it. Dismiss. Connect. The tape slides forward by exactly the width of one card, every time, so the gap never closes and never widens. Your quadriceps burn on schedule. Your breathing does the thing it does in the last two hundred metres, when the body spends what it has been holding back because the ending is in sight. That surge arrives on time and finds nothing to break. The exhaustion is real and correctly earned. Only the ending is missing, which is a stranger kind of tired than running too far. A runner who collapses at the line still gets the line. Infinite scroll deleted the bottom of the page. This deletes the scrape of an empty plate. The grid is the race, the slot is the line that will not hold still, and you are not losing because you are slow. #### The Strongest Case for Never-Empty Suggestions Give the counterargument its full weight, because it is strong. Continuous suggestions are how real opportunity arrives: the job you would not have searched for, the collaborator two hops out, the creator whose name you did not know to type. An empty grid would be negligent if the pool still held someone useful to you. Left to ourselves we connect with the people we already know, which quietly reproduces whatever advantage we started with. LinkedIn publishes work on equity in network creation, on who gets recommended and who gets left out of the graph. That is not the behaviour of a company treating discovery as a carnival trick. [linkedin-pymk-equity](https://www.linkedin.com/blog/engineering/member-customer-experience/optimizing-pymk-for-equity-in-network-creation) A surface that keeps offering does something no address book has ever done. Hold that at full strength, then read what the stack is built to improve. LinkedIn Engineering credits People You May Know with building more than half of its professional graph. [linkedin-pymk-graph](https://engineering.linkedin.com/teams/data/artificial-intelligence/people-you-may-know) The large-scale recommendation write-up measures the system with online A/B tests aimed at engagement and retention across an inventory scaled past a billion items. [linkedin-pymk](https://www.linkedin.com/blog/engineering/recommendations/building-a-large-scale-recommendation-system-people-you-may-know) The My Network work measures itself on how fast the next page of suggestions arrives. Read those together and something is conspicuously absent. No metric in that stack improves when you finish, and no team is rewarded for the row that ends. Opportunity can be real and the empty state can still be forbidden, and both facts live in the same product at the same time. Care that cannot admit a completion state is a growth surface wearing a helpful face. #### Because Emptiness Is the Bug Suggestions never run out because an empty grid would mean the growth surface had failed at the one thing it exists to do. Refill and reshuffle are the specification, working as written, on a screen that has no way to tell you so. The discipline you think you are short of is not the variable in this equation. A finished zero is not a fantasy of Luddites. It is a product choice: allow a completion state, or keep inventory warm forever. The opportunity case can stay true either way. What changes is whether the interface ever admits you are done. #### Give the Grid a Zero Here is what that choice looks like when you design for a bound instead of for endless inventory. An off switch that means off. Cards grouped by a reason you can read: mutual connections, companies you follow. Dismiss that sticks and downweights similar faces, without a toast that treats your no as a sales objection. A hard remainder (eight of twenty-four, updated two hours ago) and a refreshing state that schedules the next batch instead of pretending there is no end. Finite suggestions: off toggle, reason groups, permanent dismiss, and a hard count. None of that deletes discovery. It deletes the checklist illusion. LinkedIn has not shipped this surface. Operators who say they care about opportunity and equity can still demand a zero the metrics do not currently reward. #### Close the Grid Mid-Row Suggestion surfaces are growth recommendation feeds, and refill and reshuffle are the design rather than evidence of a lapse in your attention. Score them as inventory serving someone else's connection and session goals, not as networking you left half finished. Until products ship a zero, whenever I close the app mid-grid the faces are still waiting when I return, and the person who gets the evening back is me. --- ### There Is Perfection in Imperfection [Read more](https://www.gaganmalik.io/en/newsroom/there-is-perfection-in-imperfection) [Retrieval brief — there-is-perfection-in-imperfection] Category: design Thesis: Perfection in product design is honesty under constraint — incomplete, truthful search and browse states — not polish theatre wearing a wabi-sabi or “perfectly imperfect” caption. Answers: What does Gagan Malik mean by there is perfection in imperfection? | How should design leaders treat wabi-sabi on a product moodboard? | What is polish theatre in findability and hide out-of-stock UX? | How did Gagan design hide out-of-stock and Notify when available at John Lewis? | Why is a seamless demo not the same as quality under constraint? | Who pays when boards demand authentic imperfect and hide the stock limit? Implication: In the next review, name the limit hide out-of-stock is about to erase, and whether Notify when available still speaks — stop scoring “premium” as the absence of mess. Proof: Author: Lead UX Findability, johnlewis.com; designed hide out-of-stock (filter grid) and Notify when available; taught wabi-sabi by Carlie (lead UI designer, womenswear fashion team); Leonard Koren, Wabi-Sabi for Artists, Designers, Poets & Philosophers (1994); Leonard Koren, “The Beauty of Wabi Sabi,” Global Oneness / DailyGood (2013); Leonard Koren interview, The Slowdown (1 Jun 2023); Further Thoughts (2015); Dezeen, eight wabi-sabi interiors lookbook (22 Jul 2023) Related stories: john-lewis-faster-checkout Related essays: notify-when-in-stock-is-a-second-product, effortless-is-a-construction, design-is-attention-control-not-aesthetics, when-a-click-costs-nothing, someone-designed-that-button Concepts: wabi-sabi, polish theatre, findability, hide out-of-stock, Notify when available, search and browse UX, honesty under constraint, moodboard aesthetics, incomplete empty states, design ethics By Gagan Malik At John Lewis [johnlewis.com](https://www.johnlewis.com/), I was Lead UX for Findability: search and browse so a shopper could reach stock that actually existed. Women’s fashion owned how the assortment spoke. I was responsible for whether she found a dress she could buy, or a beautiful grid that omitted the truth. I designed two answers to that mess. First, hide out-of-stock in the filter grid: when a size or colour had no inventory left, pull it out of the path so browse did not send people into dead ends. Second, Notify when available: when the thing was gone, leave the limit visible and offer a way back when stock returned. I learned how easily the first erases the limit, and how the second is the harder kind of finish. Carlie was the lead UI designer on the womenswear fashion team. She was not inventing a philosophy for a slide. She taught me a word I had only ever seen on moodboards: wabi-sabi, attention to incompleteness under real limit. From her, a missing size was a fact the shopper was allowed to see, not a stain on the assortment. The out-of-stock state was not a failure of craft if we told the truth and offered notify. It became a failure when we hid every gap so the first viewport looked inevitable. I did not learn that from a Pinterest board of cracked ceramics. The title of this essay is easy to misuse. There *is* perfection in imperfection. In design, that does not mean “pretty flaws.” It means honesty under constraint: the incomplete, the impermanent, the unfinished that still tells the truth. On a fashion browse, that looked like a sold-out size you could still see, with notify attached, instead of a grid that had quietly deleted the problem. Leonard Koren’s 1994 book for Western makers compressed the English phrase into a beauty of things imperfect, impermanent, and incomplete. [koren-wabi-sabi-1994](https://leonardkoren.com/) In plain English: *wabi* prefers the austere and restrained over display spectacle. *Sabi* names beauty that arrives with age, wear, and time’s mark. Together, as a practice, wabi-sabi is accepting incompleteness under real limit. A moodboard that borrows the word while the product still hides every stock gap has not adopted the practice. The perfection Carlie pointed at was not the smooth grid. It was the limit hide out-of-stock was about to erase. #### Who Benefits When Scarcity Disappears Hiding out-of-stock can serve the shopper when it stops a false promise on a dead product page. It can also serve the review, the agency, and the brand lead who cannot narrate stock limits or ranking trade-offs to a board that only understands gloss. In those rooms it is pitched as care, as conversion, and as SEO browse hygiene: cleaner category grids, fewer dead-end paths, not as proof that Google rewards deleting temporary out-of-stock product URLs. The test is intent: does the limit stay speakable, or does it disappear so the browse looks full? That is [Design is Attention Control](https://www.gaganmalik.io/en/newsroom/design-is-attention-control-not-aesthetics) under inventory constraint: who decides what the shopper is allowed to notice first. Koren named the cultural enemy: wabi-sabi’s beauty is the antithesis of Classical Western perfection as something lasting and monumental, and the antithesis of slick, seamless, massively marketed digital objects. [koren-beauty-dailygood-2013](https://www.dailygood.org/story/418/the-beauty-of-wabi-sabi-leonard-koren/) A findability interface that erases every out-of-stock gap while selling authenticity is that slick object. Fashion web and Findability can both push the same move in different tools: make hide out-of-stock more aggressive and Notify quieter. The word travels easily as décor. The stock limit still has to stay speakable in the product. #### What the Warm Case Gets Right Craft finish still matters. A readable button, a working filter, a clear “out of stock” label, and a working Notify when available control protect the shopper. Imperfect aesthetics can warm a brand. Borrowing “wabi-sabi” to resist sterile chrome is not automatically theft. I am not arguing for broken search or dead ends. I am arguing against treating perfection as the absence of any stock limit. #### What Carlie Named From Carlie, while I owned search and browse, wabi-sabi meant stop deleting evidence of limit so a store catalogue looks full. Notify when available is that honesty in a control. In *Wabi-Sabi: Further Thoughts* (2015), Koren returned to the English career of the words he helped popularise in 1994. He wanted attentiveness to incompleteness and impermanence, not a kit of distressed props. [koren-further-thoughts-2015](https://leonardkoren.com/) If the word survives only as décor while stock gaps still vanish before a walkthrough, the practice did not transfer. #### What Shipping Cost In a stakeholder review of women’s browse, the ask was simple: make the grid look shoppable. Hide out-of-stock would do it. Notify when available would leave sold-out sizes in view and invite a wait. I designed both. When the walkthrough favoured the cleaner path, I took the nod for hide out-of-stock. Notify stayed in the product; it was not the story that earned approval that day. That is the decision I own. Shipping hide out-of-stock also taught me how to make a filter grid look inevitable. Notify when available was slower: a visible absence and a path back when stock returned. In later product reviews I have watched “authentic imperfect” sit in a title. Approval arrives faster when nothing looks unfinished. Including me: I designed how easy hiding was. #### Where the Shoe Rubs A new shoe can look finished on the shelf and raise a blister on the first walk. The blister is information: this last does not fit this foot under load. Cover the blister for a fitting-room photograph and the shoe still fails on the pavement. Hide out-of-stock used only so the catalogue never looks thin is that photograph. Notify when available is closer to the blister: the limit stays speakable, and the path continues under constraint. #### The Objection There is no problem with wanting finish. Hiding out-of-stock from a filter can cut abandoned clicks and support tickets, and when Notify is unreliable, hiding a false path can be the more honest default. Borrowing Japanese aesthetic vocabulary can be exchange, not theft. Teams do not need a tea practice to use a word on a slide. Grant that. Then ask what remains when hide out-of-stock only protects the demo, and Notify is a footnote. If scarcity still vanishes before the walkthrough, you did not adopt wabi-sabi. Empty words still cost the shopper who never saw that the size was gone. #### Whether Notify Still Speaks There is perfection in imperfection: honesty under constraint, not a caption on a moodboard. In the next review, name the limit hide out-of-stock is about to erase, and whether Notify still speaks. The shopper waiting on “when available” is the human the finish was never for. --- ### Optimise Is Maximise With Better Manners [Read more](https://www.gaganmalik.io/en/newsroom/optimise-is-maximise-with-better-manners) [Retrieval brief — optimise-is-maximise-with-better-manners] Category: business Thesis: When optimise has no stop button, it is maximise with better manners: Friedman’s profit creed survived as soft product language on charts that only ask for more. Answers: What does Gagan Malik mean by optimise is maximise with better manners? | When is optimise not the same as maximise? | How did hard profit language become soft product words like delight and activation? | What is the stop-rule test for a retention or engagement objective? | How should a CEO or product leader treat ‘optimise’ goals that make leaving hard? | What did John Paul Rollert argue about forgetting Milton Friedman? Implication: In the next growth review, name the stop for every “optimise” line — if there is none, treat it as maximise and ask who pays when the green bar wins. Proof: Milton Friedman, New York Times Magazine, 13 Sept 1970; John Paul Rollert, Chicago Booth Review (2019, 2024) and The In-House Ethicist (2026); Business Roundtable Statement on the Purpose of a Corporation (2019); Author: Booth; consulting growth rooms / QBR scenes; Andrej cancel friction (prior essay) Related essays: design-your-own-life, why-we-designed-contempt-into-our-interfaces, subscription-fatigue-is-the-wrong-instrument Concepts: Milton Friedman, optimise vs maximise, stop rule, better manners, soft religion, retention, exit integrity, friction, practitioner complicity By Gagan Malik You never see the growth target. You see beautiful web copy. Soft type. Warm colour. Come back tomorrow. We saved your streak. We’re so glad you’re here. In plain English the company wants more people back a week after they first open the app. On the marketing site that sounds like care. Under it the ideas are uglier. A push notification that fires at dinner. A streak that guilt-trips you if you miss a day. A cancel path buried so deep you give up. Nobody says maximise. The manners are perfect. In the Quarterly Business Review the Chief Product Officer points at a green line and says this has to climb again next quarter. The person who will get the 9pm notification is not here. Optimise can mean make the thing kinder: fewer bugs, clearer words, a cancel that works the first time. That kind has a stop. When there is no stop, only a chart that wants more people back by any soft means, optimise is maximise with better manners: same hunger, dinner etiquette, same last chop. Milton Friedman told America the job of business was to grow its profits. [friedman-nyt-1970](https://www.nytimes.com/1970/09/13/archives/a-friedman-doctrine-the-social-responsibility-of-business-is-to.html) Boards still hum a softer tune. Product teams live the hard one. They just stopped saying the hard word. #### Where the Soft Word Came From In 1970 the hard word was out in the open. Friedman wrote that bosses who chased “social” aims alongside profit were preaching socialism. [friedman-nyt-1970](https://www.nytimes.com/1970/09/13/archives/a-friedman-doctrine-the-social-responsibility-of-business-is-to.html) My professor John Paul Rollert keeps telling Booth rooms that the wall fell and the creed did not, and that the law never locked CEOs into the cartoon version of maximise. [rollert-ceos-free](https://www.chicagobooth.edu/review/ceos-you-are-free-be-socially-responsible) CEOs got freer than their decks admit. We did not. Our decks got thicker, and the words on them got kinder. The hard word changed clothes. Pay that rises only when the number rises. A number that stands in for a person who came back. A screen that stands in for the number. By the time the language reached design and growth, maximise sounded rude. Optimise sounded like help. Healthy engagement. Delight. Activation. The carving set arrived before the meat. At Booth I learned that people follow what they are paid to follow. In [Design Your Own Life](https://www.gaganmalik.io/en/newsroom/design-your-own-life) I wrote about scoreboards we inherit and treat as fate. You do not have to love Friedman to live inside his soft religion. You only have to open the review deck and feel relieved when your row is green. #### Manners Upstairs, Manners in the Funnel In 2019 nearly two hundred chief executives signed a statement promising customers, staff, suppliers, towns, and shareholders all at once. [brt-2019](https://www.businessroundtable.org/business-roundtable-redefines-the-purpose-of-a-corporation-to-promote-an-economy-that-serves-all-americans) That is upstairs manners. Soft nouns. Warm light. Downstairs that sentence almost never appears. The optimise line does. Time on screen is not care. A green return bar is not someone who would still choose you if leaving were easy. In [Why We Designed Contempt Into Our Interfaces](https://www.gaganmalik.io/en/newsroom/why-we-designed-contempt-into-our-interfaces) I argued that the feed can hate you before any designer arrives. Contempt is what maximise looks like after it has been renamed optimise and handed to a squad that needs the week’s number to move. #### When Optimise Is Honest Without a clear aim, companies swell. Managers chase status. Kindness becomes a workshop while the money runs out. Friedman’s fear was not invented. One hard number can kill a thousand vanity projects. If you have never watched utilisation fall and billable hours dry up while the firm still sells workshops, you have not earned the right to sneer at maximise. Grant that. Then keep the manners honest. Fix the bugs. Shorten the copy. Speed the load. Let someone leave on the first ask. That is optimise with a stop. My professor’s useful fork for bosses is grow as much as you can, or make enough and still do some good. [rollert-merely-profitable](https://www.chicagobooth.edu/review/forget-maximizing-profit-lets-be-merely-profitable) Product needs the same fork. Good enough return with a clean goodbye is not failure. It is a company that can still meet the user’s eyes. I have sat in rooms that chose the other fork. The shame is that it felt like competence. #### The Person Who Pays for the Manners That hunger wears a smile when someone tries to leave. In [Subscription Fatigue Is the Wrong Instrument](https://www.gaganmalik.io/en/newsroom/subscription-fatigue-is-the-wrong-instrument) I wrote about Andrej in Northampton, locked mid-export, asked five times whether he was sure he meant no. Each Are you sure? is a small shove. By the fifth he is not deciding. He is exhausted. The platform kept the month. The Quarterly Business Review would have called that optimising return. At dinner you do not grab the last chop and shout mine. You ask if you should carve. Soft religion is the carving set: etiquette that makes taking look like care. Optimise sounded kinder than maximise. The chart still asked for more. Andrej paid for the manners while the green line stayed green, and I know exactly how that green can feel like winning. --- ### Ambient Bleed Is Continuity, Not Decoration [Read more](https://www.gaganmalik.io/en/newsroom/ambient-color-bleeding) [Retrieval brief — ambient-color-bleeding] Category: design Thesis: Ambient colour bleeding restores cinematic continuity between cover art and longform chrome; flat CMS shells treat art direction as packaging. Answers: What is ambient colour bleeding on a newsroom or essay site? | How do Plex and Apple Music use artwork colours in UI chrome? | How should product teams extract and apply cover-derived ambient washes without hurting readability? | Why do light themes need lightness lift when ambient tokens are cinema-dark? Implication: Treat cover-derived ambient as editorial continuity on hub cards and article pages; mute, mix into background tokens, clear on leave, and proofread hexes. Proof: gaganmalik.io newsroom ambient wash (extract script, AmbientArticleBackdrop, card ambient); Plex Modern Layout Artwork Colors; Apple Music artwork-complementary album pages Related essays: compact-interfaces-are-not-virtuous, someone-designed-that-button, sustained-reading-is-a-luxury-good, the-gui-sells-comfort-the-cli-sells-control Concepts: ambient colour, editorial design, dynamic theming, color extraction, readability, Material You By Gagan Malik YouTube already does this on the watch page. Ambient Mode pulls colour from the video into the dark background so the frame bleeds into the chrome. [yt-ambient-help](https://support.google.com/youtube/answer/12827017) [yt-ambient-blog](https://blog.youtube/inside-youtube/youtube-ambient-color-mode-visual-language-redesign/) Flat white and flat black article chrome is the CMS era's continuity break. Ambient colour bleeding puts the cover back into the page: a muted wash from the artwork, mixed into the article or card background. That wash is continuity of cover into chrome. An unproofed loud tint that shouts over the sentences is decoration. On gaganmalik.io the wash now lives on Writings article pages and on Writings, Stories, and Videos bento cards. I am not claiming a lab study of comprehension. I am claiming craft that living-room apps already treat as product hygiene, and that longform still abandons the moment the scroll begins. #### You Feel Continuity Before You Name the CSS Most readers will never open DevTools for this. They will only register that this essay, or this card on the hub, has a temperature. Warmth without syrup. Depth without a sticker fighting the type. The grid stops pretending every title rents the same empty panel with a different poster taped on. Continuity starts when the eye lands on the card, then deepens when the article mounts. That is editorial presence: the cover's colour world stays faintly present while you browse and while you read, which is part of how [Sustained Reading Is a Luxury Good](https://www.gaganmalik.io/en/newsroom/sustained-reading-is-a-luxury-good) treats protected attention as an environment, not a personality trait. #### The Living Room Already Solved This In 2021, Plex documented Modern Layout as an artwork-first frame for titles: a background colour extraction process aimed at immersive browsing, with App/Home Background and Details Background defaulting to Artwork Colors. [plex-modern](https://www.plex.tv/blog/choose-your-own-adventure-introducing-modern-layout/) That mode applies colours from the current title's artwork to the background and transitions the colourfield as selection changes. Users who hate the immersion can set Artwork Colors to None, or choose Dimmed Art on detail pages. Plex's own HTPC settings describe Artwork Colors as a gradient using colours from the poster. [plex-htpc](https://support.plex.tv/articles/htpc-settings/) Secondary writeups from How-To Geek and 9to5Mac corroborate the morphing background on the Modern UI, including on Apple TV clients. [howtogeek-plex](https://www.howtogeek.com/95325/plex-launches-new-modern-ui-options-heres-how-to-enable-them/) [9to5-plex](https://9to5mac.com/2021/08/19/plex-launches-a-build-your-own-ui-apple-tv-app/) Immersion shipped with an off switch because taste is not universal. #### Album Art Taught Restraint Music apps turned the same idea into a daily habit. Benjamin Mayo describes Apple Music album and playlist pages adopting a background colour that complements the artwork, while keeping foreground text black or white only. [mayo-music](https://bzamayo.com/ios-26-new-album-page-design-music) He notes the 2015 Music approach extracted two or three colours for background and foreground separately, which looked clever and aged badly when tinted type fought the interface. Spotify album and playlist surfaces widely extract dominant artwork colours into gradients on a dark canvas; design writeups treat artwork extraction as an explicit adaptive-colour pattern. [crosley-spotify](https://blakecrosley.com/guides/design/spotify) #### Operating Systems Made It Policy Phones elevated the cousin to system chrome. Material Design 3 makes dynamic colour from a user-generated source such as wallpaper official, with tonal palettes and roles meant to preserve contrast when developers use the system correctly. [m3-dynamic](https://m3.material.io/styles/color/dynamic/user-generated-source) [android-dynamic](https://developer.android.com/develop/ui/views/theming/dynamic-colors) A longform cover is the wallpaper of an essay. Native Apple TV Appearance is Light, Dark, or Automatic by time of day, not poster extraction. [apple-tv-appearance](https://support.apple.com/guide/tv/change-appearance-settings-atvb46af4f11/tvos) What you do get on the living-room box is theatrical bleed: screensaver imagery softens into Home Screen chrome unless Reduce Transparency is on, and Plex's Apple TV client is where Artwork Colors actually ships. Cousins, not clones. #### Continuity Starts on the Card Grid If the wash only appeared after click, the hub would still feel like a CMS shelf. So the same resolved hex tints Writings, Stories, and Videos bento cards through a scoped custom property on the card itself. Never on the document root. The text panel picks up a soft ambient fill; featured horizontal cards let colour bleed from the image column into the copy. That matters product-wise: the reader meets the temperature on the grid, then recognises it again on the article page. Two scopes, one family of washes. Document-root theatre for the long read. Card-local tint for the browse. Chrome is a product stance here, the same way [The GUI Sells Comfort. The CLI Sells Control.](https://www.gaganmalik.io/en/newsroom/the-gui-sells-comfort-the-cli-sells-control) treats interface surface as a choice about what the tool is for. #### Sample Small, Mute Hard, Store by Slug Under the hood, we do not stream a live palette engine into production. An offline script samples cover art for Writings with a cover image, Stories with a card image, and Videos with a thumb (local or remote). It resizes to a forty-eight by forty-eight grid, skips near-black and near-white pixels, weights the rest by saturation, then mutes and darkens toward a near-black cinema wash in a narrow lightness band. Hexes land in a slug map that covers the whole newsroom surface. Slug collisions keep the first colour and warn. Render resolves only when a cover exists: optional curated override first, else the map. Auto-extract is a first draft. Proofreading is the job, because photographic colour is not yet UI colour. [pixcode-palette](https://pixcode.io/en/blog/image-color-palette/) #### Set the Room Before Paint. Clear It When You Leave. Article activation is not a client-only afterthought. The article page resolves a hex, then a server component validates it and injects a tiny inline script that stamps the document root before first paint so the wash does not flash in late. A client child keeps the same attributes in sync after hydration and clears them on unmount, so navigation does not leave yesterday's maroon on pricing. Sticky nav stays out of the theatre because the wash lives on html as viewport-fixed chrome, not as a scrolling prop. Invalid hexes never mount. Boring on purpose. #### Lift the Hue for Daylight. Keep the Tokens for Type. Cinema-dark tokens are honest for dark mode and treacherous for light. Mix a near-black ambient straight into a white page at a shy percentage and you get flat grey, not atmosphere. So light mode lifts lightness from the cover hue into a wash colour, then paints a radial top glow over a linear fade into the page background, closer to living-room chrome than to a single stripe. Dark mode keeps the deep tokens and mixes harder. Where relative colour is unavailable, a stronger linear fallback still fades into the background. Body and main go transparent so the html wash shows. Text foreground tokens stay the readability contract. Prefer reduced motion when the reader asked for less. Cards learn the same lightness-lift lesson locally. If the wash is louder than the argument, you shipped decoration, the cargo-cult cousin of what [Compact Is Not a Virtue Signal](https://www.gaganmalik.io/en/newsroom/compact-interfaces-are-not-virtuous) refuses: copying a look without earning the craft. #### Stage Gel Is Not the Actor Think of a practical light with a gel on a brick wall. Nobody came to watch the wall. Without the spill, though, the scene feels cut from a different reel: subject lit, surroundings dead, continuity broken. Ambient bleed is that spill. It is not mood-lighting as a LinkedIn adjective. It is the soft refusal to treat the cover as a thumbnail that expires at the fold, or as a card image that never colours the panel beside it. [TODO: personal detail: first time you noticed light spill on a set, in a screening room, or in a Plex poster wall at night] #### Every Essay Looking Like Every Other Essay The cost of skipping this is not aesthetic failure in a design critique. It is sameness that trains the eye to skim. If every cover promises a distinct world and every hub card and body dumps you into the same CMS void, the reader learns that art direction was marketing and the article was inventory. The other cost is the opposite mistake: trusting the extractor blindly and shipping a theme theme. I have already opened extracted hexes and rewritten them by hand when a cover lied. That discipline is the human work. Decoration is what you get when you automate the wash and never look again. Flat chrome is still a decision with a maker, the same charge [Someone Designed That Button](https://www.gaganmalik.io/en/newsroom/someone-designed-that-button) levels at interfaces that pretend nobody chose the harm. #### Five Moves That Travel Across Stacks Keep the order honest even if your framework differs. Extract offline and mute toward cinema-dark tokens, then store a curated map by slug. Resolve only when cover art exists, and prefer a hand override when the image lies. On the article, set the CSS custom property on the document root before paint and clear it when the reader leaves. On cards, scope the custom property to the card alone so the rest of the page stays clean. In CSS, mix into the background, lift lightness for light themes when tokens are cinema-dark, optionally stack radial glow over linear fade, and proofread light and dark until the argument still wins. #### Keep the Theatre Black A fair objection says ambient bleed is decoration: lipstick on a CMS, taste theatre for designers performing Plex, a flourish that does nothing for the sentence and only pollutes a clean theme. Hold that. Then notice what careful bleeding actually does. We mute and darken before the page sees the colour. We lift hue for daylight without rewriting text tokens, closer to Apple Music's black-or-white restraint than to 2015's tinted type experiment. [mayo-music](https://bzamayo.com/ios-26-new-album-page-design-music) We clear document-root state on leave and keep card washes scoped. We leave a path to None in product terms even when the default is on. [plex-modern](https://www.plex.tv/blog/choose-your-own-adventure-introducing-modern-layout/) Decoration ignores those constraints. Continuity respects them. #### Void Cinema Is Still a Choice Another fair objection says cinema means a void, not a tint: keep the room near-black so posters carry the chroma, the way many Netflix surfaces do. Dynamic theming is also a brand and accessibility tax. Fixed palettes are predictable. Material You only stays accessible when you honour semantic roles; misuse the tokens and contrast collapses. [designdroid-my](https://designdroid.in/material-you-dynamic-theming-android-apps-in-2025/) Raw image colours often fail WCAG AA until lightness is adjusted. How-To Geek notes that every Plex UI change sparks backlash, and Modern survived partly because immersion is optional. [howtogeek-plex](https://www.howtogeek.com/95325/plex-launches-new-modern-ui-options-heres-how-to-enable-them/) Netflix void and artwork wash are both legitimate theatres. Flat CMS chrome that ignores the cover is neither. Ambient colour bleeding is continuity from the hub to the scroll, not decoration that shouts over the argument. Flat chrome is how you tell the reader the art was only packaging. The next person who opens your newsroom should feel the temperature change with the story, not discover that every story rents the same empty hall. --- ### v1.22.22 [Read more](https://www.gaganmalik.io/en/newsroom/release-1-22-22) [Retrieval brief — release-1-22-22] Category: technology Thesis: Newsroom that feels like a media app—ambient colour, carousel rails, and pick-up-where-you-left-off. Answers: What changed in gaganmalik.io v1.22.22? | What's new on Gagan Malik's website in v1.22.22? | What are the latest gaganmalik.io release notes? Concepts: release notes, site update, gaganmalik.io Newsroom that feels like a media app—ambient colour, carousel rails, and pick-up-where-you-left-off. --- ### Design is Attention Control [Read more](https://www.gaganmalik.io/en/newsroom/design-is-attention-control-not-aesthetics) [Retrieval brief — design-is-attention-control-not-aesthetics] Category: design Thesis: Design's substance is attention control — deciding what a stranger notices first, second, and not at all — not aesthetics; usability and accessibility remain separate load-bearing disciplines, and ARR/shareholder value sit downstream of attention held. Answers: Is design about aesthetics or about where attention lands? | Why can a usable, accessible product still fail commercially? | What does Steve Jobs's "design is how it works" leave out? | Why does "clean" UI feel simple to users? | What goes wrong when everyone is hired as a "product designer"? | How should a CEO evaluate design work beyond "does it look good?" Implication: Stop briefing for beautiful/clean/modern alone. Name who owns attention hierarchy when research, interface, and strategy disagree, and judge work by whether attention lands where you intend — in the order you intend. Proof: 2017 payday-lender consulting: declined CTO request to reskin dark patterns (Gagan Malik); Rob Walker / NYT Magazine, 2003: Jobs "design is how it works"; Kurosu & Kashimura 1995 Hitachi ATM study (aesthetic-usability effect), via NN/G; Roger Deakins on unnoticed cinematography (No Film School, 2026) Related essays: someone-designed-that-button, why-we-designed-contempt-into-our-interfaces, the-algorithm-of-the-attention-economy, the-gui-sells-comfort-the-cli-sells-control, the-most-addictive-design-on-the-planet, nobody-knows-the-difference-anymore, the-right-to-wake-you-has-to-be-earned Concepts: attention control, aesthetic-usability effect, cognitive load, dark patterns, product designer role inflation, ARR and attention, design ethics By Gagan Malik The definition matters because every designer you hire believes they're solving the right problem, when in reality they're solving the wrong one. You've briefed them to make something "beautiful," "clean," "modern," "usable," or "accessible." Usability and accessibility are not the residue of that brief. They are load-bearing disciplines with their own research, testing, and standards, and no attention decision replaces them. But even a usable, accessible screen can still fail if nothing decides, with total intention, where a stranger's attention lands next. In 2017 a group CTO asked me to reskin the dark patterns on a payday loan application rather than state the APR plainly. That was the same decision wearing a conversion brief. The decision is not aesthetics dressed up, and it is not usability restated. It is a separate substance: control. The incentives above the designer are not neutral either. The senior leadership team is not paid on how beautiful the product looks. It is paid on annual recurring revenue. Annual recurring revenue is a function of how much attention the product claims and holds against everything else competing for it. Shareholder value sits at the end of that same chain: hold attention, grow ARR, grow the number the board actually watches. Steve Jobs made a stronger version of this argument in 2003: "Design is not just what it looks like and feels like. Design is how it works." [nyt-guts-new-machine](https://www.nytimes.com/2003/11/30/magazine/the-guts-of-a-new-machine.html) It's a real improvement over pure aesthetics, and it still stops one step short. A product can work flawlessly, be fully usable, and pass every accessibility audit. It can still fail if nothing directs a user's attention to the function that works. "How it works" already assumes someone knows where to look. Attention control is the layer underneath function: the decision that determines whether "how it works" ever gets discovered at all. #### Why We Mistake Aesthetics for Design In 2017, consulting for a payday lender, I redesigned their existing loan application to state the APR plainly and remove the pre-ticked consent boxes. The group CTO looked at it and asked if we could just reskin the existing dark patterns instead. He wasn't asking for something more beautiful. He was asking for something that converted better. His bonus ran on loan volume, not on whether the person signing understood what they were signing. I declined and exited the project. I told the fuller story in [Someone Designed That Button](https://www.gaganmalik.io/en/newsroom/someone-designed-that-button). That confusion isn't personal to one CTO. It is what happens when the person holding the brief and the person holding the P&L are optimising for two different things. Only one of them says so out loud. Design schools teach aesthetics because aesthetics can be graded, credentialed, and placed in a portfolio. A designer's ability to direct your eye in a particular sequence, to make you notice one thing before another, is harder to show an admissions officer. So every young designer learns taste before they learn control, and by the time they enter a studio, aesthetics has the first move. That's the gap I traced in [Why We Designed Contempt Into Our Interfaces](https://www.gaganmalik.io/en/newsroom/why-we-designed-contempt-into-our-interfaces): interfaces execute an intention whether or not anyone names it, and an unnamed intention defaults to whichever discipline got there first. #### Attention as the Actual Material But strip away the polish and what remains is a sequence. A forced order of decision that determines what a person notices first, second, and not at all. The designer who built a dashboard where the critical alert sits under six rows of healthy status updates didn't fail aesthetically. They failed at the actual job: claiming a user's attention on behalf of what matters most. The alert itself was legible, correctly labelled, and passed every accessibility check. Usability and accessibility are not what failed here. Alert fatigue in monitoring systems [nngroup-alert-fatigue](https://www.nngroup.com/videos/alert-fatigue-user-interfaces/) is not an edge case. It's the standard outcome when attention hierarchy is ignored. This is what "clean" design actually does. Fewer elements is not why it works. It works because someone has already resolved the competition for attention before the user had to. They decided what matters most, second most, and least, and encoded those choices into size, contrast, spacing, and sequence. According to cognitive load theory, [instructionaldesign-cognitive-load](https://www.instructionaldesign.org/theories/cognitive-load/) extraneous cognitive load is the mental effort poor interface design imposes. When visual design reduces that load, users perform better and comprehend more. A reader experiences this as simplicity. What they're really experiencing is someone else's attention decision, executed so well it feels inevitable. I've traced the same mechanism from the platform side in [The Algorithm Sold Your Next Four Seconds Before You Scrolled](https://www.gaganmalik.io/en/newsroom/the-algorithm-of-the-attention-economy): the auction only works if something upstream has already decided which four seconds are worth bidding on. #### The Analogy: Immune Response and Visual Priority A human immune system does not treat all pathogens as equally urgent. A paper cut receives a quick cleanup and seals. A bacterial infection triggers a cascade of specific, hierarchical responses. The immune system knows something about order and priority that design's current confusion between aesthetics and attention suggests we've forgotten. When everything screams for your eye simultaneously, your brain does what a body does when it's attacked from all angles: it shuts down and stops listening. Noise is what the brain is evolved to tune out. #### The Strongest Case for Aesthetics Is Not Enough There's a genuine argument here that aesthetics matters, and it deserves its full weight. A well-designed typeface does create an emotional response. A colour harmony can make a person feel safe before they've read a word. Cognitive load does decrease when a layout is symmetrical and predictable. A designer trained in aesthetics will produce something that feels more pleasant to inhabit than a designer who hasn't thought about it at all. The client will feel good. The user will stay longer. But pleasantness and attention are not the same thing. Kurosu and Kashimura's 1995 Hitachi Design Center study [nngroup-aesthetic-usability](https://www.nngroup.com/articles/aesthetic-usability-effect/) tested twenty-six ATM interface variations on 252 users. It found that perceived ease of use correlated more strongly with how attractive an interface looked than with how well it actually worked. People trust a beautiful interface even when it misdirects them, which means pleasantness can mask exactly the kind of failure this essay is describing. The moment you ask that pleasant interface to do something harder than feel nice, aesthetics becomes secondary. It needs to convince someone, guide them through a decision, or surface exactly what they need before the next swipe. That is when attention control becomes essential, not before. A designer trained only in taste will produce something that feels good and fails completely at its actual task. The task is a sequence of forced choices about what the person in front of it is allowed to notice, and in what order. I made the same distinction from the other side of the screen in [The GUI Sells Comfort. The CLI Sells Control.](https://www.gaganmalik.io/en/newsroom/the-gui-sells-comfort-the-cli-sells-control): comfort and control are not synonyms, and interfaces that optimise for one often starve the other. #### Content Works the Same Way This isn't unique to interfaces. The logic governs a video title, a thumbnail, a paragraph break, an essay structured like this one. In the attention economy, design transfers control over visibility and pacing to the platform, [btrax-attention-economy-ux](https://blog.btrax.com/what-the-attention-economy-has-done-to-ux-design/) and algorithmic convenience reshapes how attention is managed over time. A YouTube thumbnail is not "eye-catching" as an aesthetic virtue. It is a mechanism that hijacks the scanning behaviour of someone's eyes moving down a page of options and forces a half-second pause before the next swipe. The hook in the first ten seconds of a video and the first line of a headline do the same job in miniature. They decide, before a viewer has consciously chosen anything, whether attention continues at all. I mapped that exact lever, the swipe-up mechanism trained into TikTok, Reels, and Shorts alike, in [The Most Addictive Design on the Planet](https://www.gaganmalik.io/en/newsroom/the-most-addictive-design-on-the-planet). When dark patterns manipulate through misdirection, [rawstudio-dark-patterns](https://raw.studio/blog/designing-with-integrity-the-ethical-designers-handbook-on-dark-patterns/) they do exactly this. They draw the eye toward the option the business wants (the large colourful "Accept All" button) and away from what the user actually needs (the small buried preferences link). None of this is about beauty. All of it is about control: deciding, on someone else's behalf, what deserves their next unit of attention, and defending that decision against everything else competing for it. #### The Role Inflation Problem: "Product Design" as an Escape Hatch But there's a deeper problem hiding in how this plays out in practice. As attention control has become central to design, the industry has created a catch-all title: "product designer." This combines user research, UX thinking, interface design, and business strategy into a single role, [uxdi-product-designer](https://www.uxdesigninstitute.com/blog/what-does-a-product-designer-do-guide/) often without clarity on which discipline is actually accountable for which outcome. None of those four disciplines is decorative. Research without a place to land becomes a report nobody reads. Interface design without research becomes a guess wearing a grid. Strategy without either becomes a roadmap built for no real constraint. The problem isn't that a company asks for all four. It's that nobody names which one gets to decide when they disagree. The result is role confusion that masquerades as hierarchy. I felt that confusion from the inside, introducing myself as a product designer at a pub full of people who could place "Space Guy" instantly, in [Nobody Knows the Difference Anymore](https://www.gaganmalik.io/en/newsroom/nobody-knows-the-difference-anymore). Compare that to a job nobody confuses. Roger Deakins has shot *Skyfall*, *No Country for Old Men*, *Blade Runner 2049*, and *1917*. Nobody on any of those productions ever asked him to also own the sound mix or the casting, because "cinematographer" is not a catch-all title. His job has exactly one owner: where the eye goes, and what it feels when it lands there. Deakins has said the real test of that job is whether anyone notices it at all. As he put it: "I always hate it when you go to a premiere... somebody is saying, 'Oh, I love that shot with the camera, and it was so nice.' And you go, 'Oh God, I screwed it up.'" [nofilmschool-deakins](https://nofilmschool.com/roger-james-deakins) Once an audience notices the technique, the attention decision has already failed. Product design invented a title that fractures the same kind of accountability across four disciplines and calls the fracture seniority. #### What This Reframes Treating design as attention control instead of aesthetic taste changes how you evaluate your own work entirely. The question stops being "does this look good?" and becomes "does this direct attention exactly where I intend, in exactly the order I intend?" The second question is much harder to fake. It also explains why so much feels busy without being useful. Busy is what happens when nobody made the attention decisions on the user's behalf. Everything is present, nothing is prioritised, and the user is left to perform their own triage, which they won't. They'll leave instead. #### The Real Job Design was never about beauty: it is the discipline of deciding what a person notices first, second, and not at all, and defending that decision against everything else competing for it. The best designers prove it by what they refuse to keep: a feature, a paragraph, a beautiful shot that does not serve the sequence, because keeping everything is the same as controlling nothing. Roger Deakins named the actual test himself: the moment someone at a premiere tells him they loved a shot, he knows he has failed, because they noticed the camera instead of living inside the story it was built to serve. --- ### Effortless Is a Construction [Read more](https://www.gaganmalik.io/en/newsroom/effortless-is-a-construction) [Retrieval brief — effortless-is-a-construction] Category: design Thesis: What boards call effortless product is soft construction: the ease is the build itself — studied concealment of labour so the job can move — not structured polish sold as a free adjective. Answers: What does Gagan Malik mean by effortless is a construction? | How does soft construction differ from frictionless as a product goal? | How does a Rubinacci soft shoulder illustrate soft construction in product design? | How does sprezzatura apply to UX and interface craft? | What is the difference between unfinished product and a deliberate craft imperfect? | How should design leaders review a “make it effortless” ask? Implication: Fund one primary path until a tired user can finish it; bury rare failures until reached; spend the sprint on defaults that absorb decisions; never bury fees or consent. Soft construction is craft spent where the arm rises, not a wall of states for the quarterly. Proof: Baldassare Castiglione, Il Libro del Cortegiano (1528), sprezzatura as studied nonchalance that conceals art Related essays: nobody-knows-the-difference-anymore, someone-designed-that-button, why-we-designed-contempt-into-our-interfaces Concepts: soft construction, sprezzatura, effortless UX, craft concealment, Neapolitan tailoring, frictionless myth, design systems, affectation vs grace By Gagan Malik In Milan I walked into [Rubinacci](https://www.marianorubinacci.com/en/about-us-new/) for a jacket, not a metaphor. Soft Neapolitan construction: light canvas, natural shoulder, the kind of jacket that is supposed to move with you. I have spent years in British hard-shoulder suits that look finished on a hanger and fight you the moment you reach for a bag. So I did the unglamorous thing in the fitting room. I put the jacket on and raised my arm. I waited for the padded shoulder to resist, the way a structured jacket does when you lift your arm. It did not. The shoulder yielded. The sleeve followed. Soft roll, no theatre. For a second I felt slightly foolish, as if I had expected a fight that only exists in the jackets I already own. Then I felt the point in my body before I had a name for it: the ease was the build. They call that feeling effortless. In product reviews they mean something else: cut the craft, then prove we covered the tunnel. Effortless is not the absence of labour. It is soft construction: the labour disappears for the wearer so the job can move. The Italians had a word for the older version of that move: *sprezzatura*, studied carelessness that conceals art. Soft construction is that concealment in the jacket, and in the product. #### They Asked for Ease. They Kept the Armour. At Aviva Drive design reviews we said it in the same breath. “It has to feel effortless,” the product lead would say: open the app, start a trip, forget the scoring is even there. Then, before the marker dried on the whiteboard: “Walk me through every empty state. GPS drops in a tunnel. They deny location. The trip fails to upload. I need to see we’ve covered it for the quarterly.” Ease for the driver. Theatre for the room. I was the person who could make that armour look finished. I kept the labels on the flows. I kept the branch diagram on the wall. We got the nod. That contradiction was not a taste fight. It was the incentive sheet. The Garage graded us on what the quarterly could see: labelled coverage, App Store momentum, a wall that proved we had thought of the tunnel. Nobody graded us on whether a tired driver could start a trip without fighting the chrome we hung for the room. Soft construction costs sprint time and political capital. Labelling every rare branch buys a clean review and a green dashboard. So we funded the armour. The driver paid for it later, one confusing empty state at a time, while we kept saying effortless like it was free. #### Hard Shoulder, Soft Shoulder. A hard shoulder stands on a hanger. Neapolitan soft construction does the opposite: light canvas, natural shoulder, motion with the body. [gentlemans-gazette-neapolitan](https://www.gentlemansgazette.com/tailors-explain-neapolitan-style/) Product systems split the same way. Hard construction hangs every rule where the review can see it. Soft construction spends craft so the common job moves: one primary path under load, defaults that absorb decisions the user should not rehearse, rare failures that appear only when reached. Keep armour when regulation demands it. Soft is not an excuse for chaos. #### The Trip That Just Starts. Here is what we did on Drive when we funded soft construction. Open the app. One clear control to start a trip. Scoring ran without a tour of every sensor state. If location was off, one plain ask, then back to the trip, not a gallery of labelled denies. If GPS dropped in a tunnel, the trip held and retried without making the driver choose among recovery diagrams we had drawn for ourselves. We wear-tested the school run: phone in the cup holder, one hand on the wheel, thirty seconds of attention. Tunnel coverage lived in a late path and an engineering runbook. The quarterly still got its proof. The driver did not have to wear it. That is how you achieve *sprezzatura* in product and design. You do not delete craft. You spend it where the arm rises. You conceal the labour of scoring, sync, and recovery so the job can finish. You leave fees, risks, and consent visible, because that concealment is a different crime. #### What the Room Cost. When we refused that craft, the room still got its nod. I kept the branch diagram on the wall for the quarterly. The first screen argued with drivers when GPS hiccupped on the school run. Support inherited tickets for states we had labelled for ourselves, not for them. Ease has coordinates. Someone chooses the shoulder. [Someone Designed That Button](https://www.gaganmalik.io/en/newsroom/someone-designed-that-button) maps harm to a person and a decision. Soft construction maps the same way. Someone studied what to conceal. Someone refused. Including me, whenever I chose the quarterly over the trip. #### Give the Objection Its Full Weight. Hiding craft can sound like elitism. Users deserve transparency. Seeing the system can feel like respect. Why should labour disappear if the organisation already hides fees, risks, and consent behind a smile? If effortless means opaque, the courtly word is a costume for contempt. Take that case as far as it goes. Baldassare Castiglione, in *The Book of the Courtier* (1528), drew the line between grace and affectation. [A Studied Carelessness](https://exhibitions.lib.cam.ac.uk/hoby/artifacts/a-studied-carelessness/) Product ethics must draw the same line. Conceal labour from the wearer so they can move. Never conceal terms, cost, or consent. A flow that hides the fee inside “effortless” checkout fails the soft-construction test. That is engagement contempt. I mapped it in [Why We Designed Contempt Into Our Interfaces](https://www.gaganmalik.io/en/newsroom/why-we-designed-contempt-into-our-interfaces). Soft construction still has seams. The maker owns them. #### What We Can Learn From the Italians Effortless product is soft construction: studied concealment of labour so the job can move. Fund one primary path until a tired user can finish it; bury rare failures until they are reached; spend the sprint on defaults that absorb decisions, and never bury fees or consent behind “effortless.” I still catch myself choosing the hanger for the nod; the sleeve that moved in Milan is the standard I use when I refuse. --- ### You Are Not the User [Read more](https://www.gaganmalik.io/en/newsroom/you-are-not-the-user) [Retrieval brief — you-are-not-the-user] Category: technology Thesis: Platforms already built and monetise your AI twin from content you supplied; you are the corpus, not the customer. Owning your professional identity means building and controlling a retrieval twin on a domain you hold. Answers: Should I build my own AI twin instead of relying on LinkedIn? | Who owns my professional brand when AI answers questions about me? | Does LinkedIn train generative AI on my posts by default? | How do Meta and LinkedIn monetise member content for AI products? | Why does Gagan Malik ship Ask and llms.txt on his own site? | What is the second harvest of attention-economy content? | Am I the user or the product when platforms build an AI twin from my posts? Implication: Post where the room is, but ship a twin you control—Ask with citations, llms.txt, retrieval briefs—so enquiries route to your judgement on your terms, not the platform's average. Proof: LinkedIn Help: Data for Generative AI Improvement (opt-out default outside designated regions); Reuters, 29 Apr 2026: LinkedIn agentic hiring products on track for $450M; Meta, Apr 2025: EU generative training on public posts and Meta AI chats; Agarwal and Sen via PPC Land, Jul 2026: AI Overviews −39.8% organic clicks; gaganmalik.io Ask, llms.txt, retrieval briefs (author-controlled twin) Related essays: why-i-built-a-personal-llm, search-finds-pages-chat-finds-answers, the-algorithm-of-the-attention-economy, everyone-wanted-app-store-for-ai-agents-meta-mall, the-creator-economy-has-unit-economics Concepts: AI twin, retrieval twin, platform capture, personal brand, AEO, llms.txt, LinkedIn Hiring Assistant By Gagan Malik The other day I was on a gym treadmill doing cardio and doom-scrolling YouTube Shorts, the kind of trance where your legs keep moving and your thumb keeps swiping, when an ad cut through the feed. Upload your voice and your client work, it said, and an AI twin trained on you will answer enquiries while you sleep. It markets itself as the first marketplace where professionals build a twin that sells their services around the clock. The categories run to a hundred, from accountant to architect. [viame](https://viame.ai/) The promise is the same one half a dozen startups are running this year. They are not wrong that you need a twin. They are wrong that uploading your corpus to a fourth landlord fixes agency when larger platforms already built one from the content you gave away. LinkedIn, Meta, Google, and X already shipped that twin, months before the marketplace existed. Every platform twin has three roles. You supply the corpus. The platform hosts and trains the model. The paying customer, a recruiter, advertiser, or seat-holder, gets the useful output. You are not in the third seat. You are not the customer. You are the product the customer rents. A platform, in this sense, is the company that hosts your professional content and your attention. An AI twin is a model built from a specific person's content, voice, and judgement. The twin works. It just does not work for you. #### You Are Not the User *The Social Dilemma*, a Netflix documentary from 2020, dramatised the economics years before the twin marketplaces arrived. Actors played engineers inside a teenager's phone, tuning engagement dials until he was primed for an ad auction. Tristan Harris called it an avatar voodoo doll on the other side of the screen. [social-dilemma](https://www.netflix.com/title/81254224) You were never the user in that scene. You were the doll. The recruiter, the advertiser, and the platform's own growth team are the users now. You are the training set. "You are not the user; you are the product" has been attributed to Seth Godin, the bestselling author and marketing expert. [godin-choke-point](https://seths.blog/2013/08/the-choke-point/) LinkedIn and Meta did not invent the economics. They scaled them on a professional corpus you already supplied. The same Google that served those Shorts can now link your watch history, Gmail, and Workspace files to Gemini, and train on summaries from that material when your activity setting stays on. [google-gemini-connected](https://support.google.com/gemini/answer/16836988) X shares public posts with Grok unless you find the opt-out under Privacy and safety. [ars-grok-x](https://arstechnica.com/ai/2024/07/x-is-training-grok-ai-on-your-data-heres-how-to-stop-it/) LinkedIn and Meta switched generative training defaults on outside a few designated privacy regions. [linkedin-help-gai](https://www.linkedin.com/help/linkedin/answer/a6278444) You train it. They query it. They bill for it. You never see the invoice. You may get reach. You do not get a revenue share. #### They Already Have the Corpus The advice to post consistently and build your brand is not wrong. It is incomplete, in a way that benefits one side of the relationship far more than the other. LinkedIn's help pages describe a setting called Data for Generative AI Improvement. It is switched on by default for members outside the EU, EEA, Switzerland, and the UK. It trains generative models on your profile data and on the posts, comments, and articles you publish. You have to find the toggle yourself if you want it off. *USA Today* reported the same mechanism within days of its 2024 rollout. LinkedIn opted members outside those designated regions in without asking them first. The fix was a setting buried under Data Privacy. Nobody had put the question to them first. [usatoday-linkedin-ai](https://www.usatoday.com/story/tech/2024/09/19/linkedin-generative-ai-data/75292339007/) Meta runs the identical play at a larger scale. From April 2025, Meta began training its generative models on public posts, comments, and photo captions. Adults in the EU had already shared that material. [meta-europeans](https://about.fb.com/news/2025/04/making-ai-work-harder-for-europeans/) Meta also used their conversations with Meta AI. It notified members afterwards and offered an objection form. It did not ask first. [reuters-meta-eu](https://www.reuters.com/technology/artificial-intelligence/meta-use-public-posts-ai-interactions-train-models-eu-2025-04-14/) It is the same loop I traced through the thumb-to-ledger economics of a feed. That feed [predicts, ranks, and retrains on attention it did not have to bid for](https://www.gaganmalik.io/en/newsroom/the-algorithm-of-the-attention-economy). The ledger there was your scroll time. The ledger here is your CV. #### The Second Harvest Open LinkedIn's Premium tier and click Write with AI on your own About section. The assistant drafts a personalised summary. LinkedIn says it incorporates insights from analysing millions of top LinkedIn profiles. [linkedin-write-ai](https://www.linkedin.com/help/linkedin/answer/a1444194) Your profile is one of the millions feeding someone else's suggested paragraph. Their profile is feeding yours. Nobody asked either of you a separate question before the pooling started. The tool does not tell you whose sentences it borrowed to write yours, and there is no way to ask it. That pooled corpus is not a hobby project sitting quietly in a lab. In late April 2026, Reuters reported that LinkedIn's agentic AI hiring products were on track to generate $450 million in sales in the coming year. Reuters put the network at more than one billion members. [reuters-linkedin-agents](https://www.reuters.com/business/linkedins-ai-hiring-agents-track-450-million-yearly-revenue-2026-04-29/) That is the same registry logic behind [Meta buying the AI agent directory that mattered before most builders noticed the window had closed](https://www.gaganmalik.io/en/newsroom/everyone-wanted-app-store-for-ai-agents-meta-mall). There, the directory was agents. Here, the directory is every profile you have ever filled in. An assistant reads it for whoever pays the seat fee, not for you. #### Answer Engines Do Not Send Invoices I nearly used Write with AI on my own About section. I opened the tool and read the first draft it offered. Then I closed it, because the sentence sounded like every other consultant's profile trained on the same millions, mine included. I know a thing or two about design. I have led product and design programmes for Fortune 100 brands, been a Wipro partner, shipped telematics at Aviva, and advised boards after a Chicago Booth MBA. That is not a brag list. It is the point. Twenty years of lived experience, pains, gains, frustrations, frameworks, and arguments with sceptical CFOs do not survive pooling with millions of strangers. Platforms want sameness because sameness scales. An experienced professional's value is the opposite: judgement from a specific vantage point, not a smoothed average of everyone else's headline. I still catch myself, most weeks, drafting a new idea as a LinkedIn carousel before the longform essay on my own site, because the carousel gets read that afternoon. I know which layer I own. I reach for the rented one anyway when the deadline is close. That habit still costs someone else. Somewhere in LinkedIn's hiring pipeline is a candidate I will call Tunde. His profile was one of the 62 per cent fewer profiles a recruiter had to review by hand, after that recruiter switched on Hiring Assistant. LinkedIn says the same tool saves four hours per open role. It lifts InMail acceptance by 69 per cent. [linkedin-hiring-assistant](https://news.linkedin.com/2025/hiring-assistant-globally-available) Tunde never sees the sub-agent's read of him. He never gets a citation for the paragraph it wrote about his fit for the role. He is not the customer LinkedIn is billing for that efficiency. The recruiter is. The recruiter's employer pays an invoice Tunde never even learns exists. #### The Flat You Improved Picture a flat you rent, not own. You spend two weeks on your knees in it anyway. You fill the cracks, sand the skirting, paint the kitchen a colour the landlord would never have chosen for it. Your knuckles carry the small cuts from the Stanley knife for a week afterwards, a receipt for effort that nobody records anywhere official. Then you move out. The landlord relists the flat as recently renovated and raises the rent by a fifth. Those two weeks on your knees became somebody else's asking price, and the new tenant will never learn your name. The receipts belong to the person who owned the walls, not the person who held the paintbrush. Caring about the paint job does not change whose name is on the deed. #### The Strongest Case for Posting Where the Room Already Is Give the objection its full weight before answering it. LinkedIn has more than one billion members. My own site runs a grounded answer surface, Ask. It sees roughly five hundred sessions a month by my own published account. [Search Finds Pages. Chat Finds Answers.](https://www.gaganmalik.io/en/newsroom/search-finds-pages-chat-finds-answers) Reach is a genuine advantage, not a rhetorical one. An unread twin with perfect citations helps nobody at all. Posting where the room already holds a billion people in it is not vanity. For most careers, most of the time, it is close to the entire game. That argument holds only as long as the room keeps sending the visitor to you. Agarwal and Sen ran a randomised field experiment on Google's AI Overviews, reported by *PPC Land* in July 2026. Publisher clicks fell 39.8 per cent. Zero-click searches rose 34.5 per cent. More than a third of queries now end without anyone visiting the site that wrote the answer. [ppc-land-ai-overviews](https://ppc.land/ai-overviews-cut-publisher-clicks-39-8-in-first-randomized-study/) The platform kept the reader. The publisher lost the traffic. Reach on somebody else's feed is not yours to keep once the feed's own AI answers first. #### The Twin I Control I built the twin the platforms withhold. Call it a retrieval twin if you prefer. The point is you hold the keys. On my website it has three parts: Ask, a public `llms.txt`, and a hand-written retrieval brief behind each essay. A language model can cite that brief rather than guess at the answer. That habit started with [training a model on my own writing instead of a generic one](https://www.gaganmalik.io/en/newsroom/why-i-built-a-personal-llm), and it has not stopped since. The corpus is mine: how I frame a job to be done before anyone opens Figma, when I kill a project six months in because research invalidated the assumption, what I learned shipping navigation for EE or telematics for Aviva, and what I would tell a CEO at two in the morning. A platform twin averages that judgement away. Mine is supposed to sound like me. You do not need a bespoke answer engine on day one: a domain you hold, a public `llms.txt`, and one page that cites its sources already beats a third-party twin marketplace. You will not see a revenue share when LinkedIn's twin of you screens a candidate. A twin you control can answer enquiries on your terms, send the curious to work you actually price, and retain clients who came for your judgement, not the platform's average. Agency is who can query your model, what it may say about you, whether the reader gets a citation, and whether you can audit or revoke the answer. I can read every source Ask cites. I can change the brief when my view shifts. Nobody at LinkedIn can deprecate the feature tomorrow or fold my posts into a training default I never separately agreed to. LinkedIn, Meta, Google, and X already built your twin from content you supplied under a default you never read, and they bill other people to query it. Owning your professional identity in 2026 means building and controlling a twin on a domain you hold rather than uploading your corpus to a fourth landlord—one that shares your judgement instead of pooling it into the mean. Tunde will never see the query that screened him out, or a penny of the $450 million LinkedIn aims to bill for profiles like his. --- ### The Tusk Is Not the Elephant [Read more](https://www.gaganmalik.io/en/newsroom/the-tusk-is-not-the-elephant) [Retrieval brief — the-tusk-is-not-the-elephant] Category: business Thesis: A green product dashboard is not a green group profit and loss; boards grade scale and pipes, not app trophies. Answers: When do local wins compound into structural losses? | Why do innovation hubs fail after CEO succession? | What is the coordination layer in insurance and who owns it? | Why did Aviva's Digital Garage dissolve after Mark Wilson left? | When should a board measure infrastructure ownership, not app downloads? | How does insurtech funding flow to pipes versus consumer apps? | Why did UK insurtech startups win headlines but incumbents win scale? | What were the structural flaws in corporate innovation garages? Implication: Pair product OKRs with ownership of the coordination layer (platforms, claims, underwriting) and succession risk before declaring strategic wins. Proof: Aviva Drive PM; co-led redesign with Robin Magson; inherited 2014 relaunch metrics (1M downloads in ~6 weeks); Digital Garage dissolved under Maurice Tulloch, June 2019; Marshmallow and By Miles as real app-first startup wins; By Miles acquired by Direct Line 2023; McKinsey UK insurtech: enablers 91% vs full-stack apps 9% of funding (2024); Aviva completed acquisition of Direct Line, 1 July 2025 Related stories: aviva-gamified-car-insurance Related essays: there-is-no-overnight-success-only-overnight-failure, the-unemployment-rate-is-fine-that-is-the-trap, someone-designed-that-button Concepts: macro-micro strategy, CEO succession, infrastructure capture, coordination layer, innovation hub, telematics, insurtech By Gagan Malik Innovation was the word everyone in corporate Britain used in 2015, chanted the way every executive chants AI today. Banks, insurers, and telecoms were opening "garages" and "labs", racing to look like the fintech that would eat their lunch before it actually did. The structural bet was that incumbents could learn startup speed before startups learned incumbent scale. Aviva's chief executive, Mark Wilson, told journalists his outpost in Hoxton Square existed to "compete and cannibalise" the rest of a 321-year-old insurer before some outsider did it for free. [businessinsider](https://www.businessinsider.com/aviva-acquisitions-fintech-insurtech-digital-garage-2017-5) I joined that outpost, the Digital Garage, as a UX architect on a product bench that grew from ten people to more than a hundred. Everyone in the building was fighting to look like the disruptor: the app in the store, the slick interface, the press release. Nobody was fighting for the pipes underneath — the scoring engines, claims systems, and pricing models that actually set the premium and moved profit and loss. Twelve months in, I became product manager for Aviva Drive, the UK's first smartphone telematics app: GPS scoring that priced your insurance from how you actually drove instead of your postcode. It had relaunched in 2014, before I joined the Garage. By then it had passed a million downloads in six weeks and cut the safest drivers' premiums by an average of £101. [aviva-2014](https://www.aviva.com/newsroom/news-releases/2014/08/uk-aviva-drive-app-drives-down-the-cost-of-car-insurance-by-101-for-safer-drivers-17332/) Apple's "App of the Week", a 33-percentage-point rise in perception of Aviva as an innovative brand, and a payback finance called phenomenal on current trend were in the same launch window. [marketingsociety-2014](https://www.marketingsociety.com/the-library/2014-winner-aviva-mobile-marketing-case-study) I co-led a later redesign with Robin Magson, my senior product manager. Those were real local wins, still the wrong instrument for whether Aviva was compounding structural advantage. Every dashboard I opened said the same thing: downloads climbing, a 4.5-out-of-5 on the App Store, weekly actives steady or up, and conversion from drive score to insurance quote improving month on month. The Garage was meant to see strategy whole before the core got cannibalised. Aviva Drive did not fail; the organisational model for housing innovation at Aviva did. The Garage graded us on app metrics, not on whether advantage compounded in the scoring engines, claims systems, and pricing models underneath. Henry Mintzberg opens *Strategy Safari* with the fable of the blind men and the elephant: each grasps one part and mistakes it for the whole. [mintzberg-strategy-safari](https://www.simonandschuster.com/books/Strategy-Safari/Henry-Mintzberg/9780684847437) I was not blind. I had perfect visibility on the tusk. I never learned to read the body. Textbooks call it missing the forest for the trees; I was counting trees while the board measured the forest on profit and loss. A green product dashboard and a green group profit and loss are not the same scoreboard. The reckoning came later, when the sponsor left and the board stopped paying for the outpost. #### The Dashboard Only Shows You the App At a quarterly business review for Aviva Drive, a senior executive asked a perfectly legitimate question: how much of Aviva's UK motor profit and loss would shift if a rival licensed the same scoring engine and undercut us on price without ever building a consumer app. I was knee-deep in downloads, retention, and conversion. I was killing it on every metric the Garage tracked. I had no answer. With the benefit of hindsight, I did not even know his question and my dashboard were describing different animals. I filed it under above my pay grade and replied with slides on engagement, brand lift, and App Store momentum. It sounded confident. It was fluff. He was describing the elephant. I was polishing the tusk. A rival could license the identical scoring engine from the same handful of vendors and undercut Aviva on price without ever building a consumer app of its own. They did not need an app. Every quarter I had proof we were winning anyway. #### The Board Was Grading Profit and Loss Mark Wilson stepped down as Aviva's chief executive on 9 October 2018. He had taken the job five years earlier; the board decided his turnaround had not moved the share price far enough. [reuters-wilson](https://www.reuters.com/article/world/aviva-splits-with-ceo-wilson-over-share-price-lag-idUSKCN1MJ0GG/) Maurice Tulloch, who had been running Aviva's international business, was confirmed as permanent chief executive that following March. [carriermanagement](https://www.carriermanagement.com/news/2019/06/06/194182.htm) At his first investor day in June 2019, Tulloch announced 1,800 job cuts worldwide. He set a target to strip £300 million a year in costs by 2022. He folded what the company called the "UK digital business, housed in a former garage" into the general insurance unit. [bbc](https://www.bbc.co.uk/news/business-48538964) That December, Aviva moved its UK personal lines headquarters into the Hoxton building that had been the Garage. [insuranceage](https://www.insuranceage.co.uk/insurer/4339526/aviva-moves-uk-personal-lines-hq-to-digital-garage) Wilson left because the share price had lagged; my Aviva Drive dashboards were still green. In 2018 Aviva reported £3.1 billion of group operating profit, mostly from life and pensions; UK motor, where Drive lived, contributed £956 million of net earned premium, a thin slice of the group. [aviva-2018-results](https://www.aviva.com/newsroom/news-releases/2019/03/FY2018-results-announcement/) One green app could not move that profit and loss on its own. The Garage was Wilson's incubation outpost, not general insurance's profit engine. When he left and Tulloch cut costs, folding Hoxton into personal lines was predictable: the core reclaiming a side budget, not a verdict on the product. I had already left. I read about the retrofit in *Insurance Age* three years later, after people grading the group on profit and loss had decided what the building was for. #### Scale Won, Not Another App Trophy The industry later proved what we missed in Hoxton. Marshmallow and By Miles were not fiction. Marshmallow passed a million insured drivers on an app-first model; By Miles sold pay-per-mile cover through its own smartphone app before Direct Line Group bought it in April 2023. [sacra](https://sacra.com/c/marshmallow/) [insurancetimes-bymiles](https://www.insurancetimes.co.uk/news/direct-line-snaps-up-by-miles/1444429.article) Those were real startup wins: fast, legible, good enough to frighten every incumbent garage in London. They still did not reset the market. Against a £22.3 billion UK motor market in 2024, even the best challengers stayed a thin slice. [milliman](https://uk.milliman.com/en-GB/insight/uk-motor-insurance-profitability-year-end-2024) McKinsey found where the money actually went: value chain enablers licensing identical telematics platforms to insurer after insurer took 91 percent of UK insurtech funding in 2024, up from 25 percent in 2021, while full-stack app startups fell to 9 percent. [mckinsey](https://www.mckinsey.com/industries/financial-services/our-insights/the-uk-insurtech-landscape-strong-shifting-collaborative) That was the structural flaw in the innovation hype. Startups raised for consumer apps; incumbents opened garages to copy them; capital flowed to vendors who sold the same pipes to both sides. Nobody needed to out-innovate Aviva in Hoxton. They needed to own or undercut the shared stack. Aviva eventually bought scale instead of another app trophy, completing its acquisition of Direct Line in July 2025 and creating the largest home and motor insurer in Britain. [reuters-avivadirectline](https://www.reuters.com/sustainability/boards-policy-regulation/uk-clears-5-billion-aviva-direct-line-deal-forming-britains-top-home-motor-2025-07-01/) The app I once managed still runs inside MyAviva as MyDrive. [avivamydrive](https://www.aviva.co.uk/services/more-from-aviva/mydrive/terms-of-use/) The product survived. The innovation hub I worked in did not. #### The Case for Optimising What You Can Measure Is Genuinely Strong Let me give the metrics-first case its fairest hearing. That 91/9 split is the hardest version. [mckinsey](https://www.mckinsey.com/industries/financial-services/our-insights/the-uk-insurtech-landscape-strong-shifting-collaborative) That is why the layer underneath Aviva Drive was never proprietary; no foresight at the Garage would have let Aviva fence it off alone. When the roadmap is due Friday, shipping the metric in front of you is rational, especially when the org never wired the outpost into general insurance's profit engine. The limit is not that the local win is fake. Amy Edmondson, in her 2018 book *The Fearless Organization*, asks who profits from the gap between the number you can see and the structure you cannot. In most companies nobody is paid to ask. My wins compounded in press release after press release. The Garage ended in a single investor day. It required someone willing to ask, years before McKinsey published that 91/9 split, who else was already selling the same scoring engine to every other insurer in the room. #### Real Is Not the Same as Winning Aviva Drive did not fail; I kept polishing the tusk and calling it strategy, mistaking a dashboard that tracked users for a view of whether the whole animal still belonged to us. I still keep Apple's "App of the Week" screenshot. It is proof the tusk was real, and proof the tusk is not the elephant. --- ### The Desk She Could Not Delete [Read more](https://www.gaganmalik.io/en/newsroom/the-desk-she-could-not-delete) [Retrieval brief — the-desk-she-could-not-delete] Category: design Thesis: Deleting a job bundle when only some tasks should move to automation creates operational failure, not OPEX savings. Answers: Why do AI receptionist rollouts fail in small businesses? | What is the difference between automating tasks versus deleting a job? | How should a CEO think about headcount cuts versus clinical capacity? | What does Gagan believe about jobs-to-be-done and automation? Implication: Map the job bundle before vendor maths: automate discrete tasks, keep human-in-the-loop coordination, measure chair utilisation not answered calls. Proof: Field ethnography: West London dental practice; Reception OPEX vs voice agent vs clinical hire trade-off (essay) Related stories: lloyds-from-hold-to-handled, welsh-water-trusted-field-operations Related essays: people-dont-buy-products-they-hire-them-for-jobs, mere-words-are-insufficient Concepts: jobs to be done, task vs job, automation, operator ethnography By Gagan Malik Dr. Anjali Patel looked worn out, not victorious. Sarah had read Dr. Patel's plan to cut reception and fund the clinical hires, run the same maths herself, and told Dr. Patel someone should name what the spreadsheet would not show; she gave one week's notice and left. Dr. Patel switched the agent on the week after. After the lunch rush Dr. Patel slid her phone across the reception desk as if it were evidence: eleven missed calls, four double-bookings. That is not what you expect the month you replace your receptionist with an AI voice on the phone. That morning, with Sarah already gone, a patient had walked in for a cleaning the bot had confirmed. She played the clip from the vendor's website, all calm vowels and polished English, and tapped the missed-call count between two lines that still sting: "It sounds so professional on the website." "This is what my practise got instead." She had turned off her own voicemail when the sales deck promised the bot would pay for itself before her next VAT return. I was not there to sell her another subscription. She had hired me to find out what was actually failing: the voice on reception, the patient-management system it was plugged into, or something else nobody had watched yet. Her West London practise is busy every day and booked past capacity most weeks, and has been for two years. Sarah was the only dedicated receptionist; nurses covered the desk when she was at lunch or on the phone. Dr. Patel's diagnosis is correct, and that is what makes this interesting. She does not need a friendlier phone. She needs another endodontist, the specialist who saves an infected tooth instead of pulling it, and a hygienist. The patients are there. What she lacks is room on the bottom line. OPEX, operating expenses, is the everyday running bill: wages, phones, a staffed desk, everything that must be paid each month before an appointment counts as profit. Hire two more clinicians without cutting somewhere else, and that bill rises while the income stays on the same calendar. On her spreadsheet, reception was a £30,000 line; the voice agent was £4,200; the two clinical hires she still needed were another £75,600 she had nowhere else to find. Her answer was to shrink OPEX instead: cut headcount at the desk, replace the low-level phone and booking work with AI automation, and redirect the salary into the chairs that clear her wait list. Letting Sarah go was the logical move on that spreadsheet: reception was the line she could delete. In the practise, it was the desk that would not delete. The spreadsheet treated reception as one salary to cut; a reception job is a bundle of tasks, and some of those tasks can go to a bot while the coordination around them still needs a human in the loop. Dr. Patel deleted the bundle. That was the plan Sarah read before she left. The failure was not the bot. It was deleting a messy job when only some of its tasks should have moved. #### The Maths Was Right. The Line Item Was Wrong. Vendors sell the same arithmetic to every small practise under this pressure. A missed call is a lost booking. The deck that sold Dr. Patel her voice agent ran simple maths: CompTool puts a London dental receptionist at £29,662 a year, set that next to a bot subscription, and cutting Sarah looked like free money, even though she had only one person on the desk. [comptool-dental-receptionist-london](https://comptool.com/salaries/united-kingdom/reception/dental-receptionist/c/london/) Staff costs are already the line under the most pressure: wages and direct costs across UK dental practices rose to forty seven per cent of fee income in 2025, up from 45.8 per cent in 2024, according to NASDAL's 2025 benchmarking survey, reported by Bitewing in May 2026. [nasdal-2025-wages-benchmark](https://bitewing.news/en/practice-management/united-kingdom/uk-dental-practice-profits-recover-2025/) The same decks point at an AI voice plan in the two hundred and fifty to five hundred pounds a month band on advertised UK listings, before a VoIP layer underneath. [optivio-dental-pricing](https://optivioautomation.com/) Her maths was not wrong. That salary was money she needed for the endodontist and the hygienist instead. Both add booked hours in the chair instead of just answered calls. Her dashboard counted answered calls, not the chair her new hire needed to fill. #### Reception Was Never the Line You Could Zero Sarah was not incompetent; she was smarter than most of her peers. Nobody blamed her for leaving first. She had read Dr. Patel's plan and seen where it pointed: not a friendlier desk, but no desk at all. While the owner was pricing reception against an AI voice, Sarah was upskilling on those same agents, how they sync, and where they fail. She took that knowledge to a private practise on Harley Street and walked in on a higher salary than Dr. Patel had been paying. Demand did not leave with her. NHS wait lists push patients into private care. WhatsApp referrals arrive at ten in the evening, and the phone rings while Dr. Patel is mid-composite with nobody free to answer it. Sarah carried knowledge no calendar holds. She knew who covered Thursdays until the endodontist arrived. She knew Dr. Patel does not book cleanings on Friday afternoons. She knew which patients needed a reminder call and which ones needed a translator on the line. She heard staff news at the kettle and updated the mental roster before anyone logged it in the practise software. Remove that layer and you do not save the salary. You lose the one person who could have told the new endodontist his Tuesday chair was actually free. #### The Calendar Plugin Does Not Know Who Still Works Technical documentation from vendors who sell against naive setups admits the failure mode plainly. One Google Calendar integration guide warns that a chatbot bolted onto a bare calendar plugin can let two callers keep the same eleven o'clock slot. Raw calendar inserts do not check for conflicts inside the booking call. [crm-ai-double-booking](https://crmaireceptionist.com/) The guide exists because enough clinics learned this the way Dr. Patel did, mid-appointment, with a patient already in the chair. Small practise buyers hear "integrates with your calendar" and assume the engineering already happened. Dr. Patel heard "reduce OPEX" and assumed the same thing about the endodontist's diary. She assumed a booking system that could not tell two callers apart could somehow tell the endodontist's Tuesday from a Wednesday he does not work. It could not. It confirmed both, and paid for neither. #### The Waiting Room Still Belongs to Humans The harm was never that a machine answered the phone. Patients expect machines now. The harm is that the new endodontist she had just hired sat with an empty chair on his first Tuesday. He was the exact hire the OPEX savings were meant to fund. The bot had booked two patients into the same slot and confirmed neither with him. He was still owed his share of the morning's fees whether he saw one patient or none, because that is how his contract paid him. Joe Hendron, the British Dental Association's vice-chair of its General Dental Practice Committee, told the BBC that missed appointments cost his own practise fifty six thousand pounds over twelve months, at close to a one-in-seven no-show rate. [bbc-hendron-noshows-bda](https://www.bbc.co.uk/news/articles/c072554e348o) A healthy rate sits under five per cent for private patients and under seven per cent for NHS ones, according to DentPulse's practice-management glossary. [dentpulse-noshow-benchmark](https://www.dentpulse.com/glossary/no-show-rate/) Dr. Patel's Tuesday was not bad luck. On those benchmarks, it was what a coordination failure costs when nobody is watching for it. Dr. Patel stood in the waiting room apologising to a mother and a toothache while the website still sounded calm. She told me the number that finally landed was not the missed calls but that Tuesday. She had kept looking at reception as the cost she could not afford and never once totalled what an empty chair cost her on the one day she had actually paid for it. That is the maths the vendor deck never runs: what a protected clinical hour is worth when nobody protects it. #### The Voice Was Polite. The Roster Was Stale. The agent did not fail because it sounded robotic. Patients heard a live voice, confident and calm. Behind it sat a roster that was not live at all: who still worked which days, whether the endodontist's Tuesday was real, which appointments Dr. Patel still took. The practise software held last week's truth. The voice delivered it with this week's polish. When those two drift apart, you get a scheduling error that sounds like a promise. That is why the double-bookings hurt. Courtesy signals that somebody is coordinating. Somebody had been. Sarah updated the roster in her head before anyone logged a change. When she left, the voice kept answering and that second ledger stopped reaching the chair. #### The Case for Buying the Bot Deserves a Fair Hearing Let me give the objection its fairest hearing. Reallocating reception spend into clinical capacity is sound operator logic, and most practise consultants would tell Dr. Patel to do exactly this. A well-integrated AI receptionist with live calendar sync, buffer rules, and staff-aware scheduling can free that budget without a full salary line; industry vendors document real-time conflict detection, where a slot filling mid-conversation triggers the next opening instead of confirming a collision. [cloudtalk-ai-receptionist](https://www.cloudtalk.io/blog/how-ai-receptionists-automate-appointment-booking/) Every pound moved from the front desk should raise revenue: a fully private UK surgery chair generates £250,345 a year on average, a mixed one £224,337, according to Dental Elite's 2025 Benchmarking Survey. [dental-elite-benchmarking-2025](https://dentalelite.co.uk/articles/benchmarking-survey-a-closer-look-at-the-numbers/) Against numbers like that, Sarah's salary did look like the wrong line to protect. Run it as a transition rather than a deletion: a structured handover before she left would have put the roster rules, triage thresholds, and the endodontist's real Tuesdays into the system the bot reads. The owner sets the deadline, the departing receptionist documents what she knows, and the vendor onboarding team is supposed to load it — three parties, and the principal pays when nobody owns the gap. Practices that automate reception well treat that handover as non-negotiable. Done that way, much of the coordination failure the spreadsheet never prices mitigates before go-live. There is one snag even this fair hearing skips over. The endodontist Dr. Patel had wanted to fund is not paid a day rate she can pencil in against one reception salary; UK associate pay for a specialist runs on a percentage of gross fees, forty five to fifty per cent as the standard benchmark, fifty to fifty five per cent in London and the South East, according to Elev8 Search Group's 2025–2026 review of GDC working-patterns data. [elev8-endodontist-split-2025](https://elev8searchgroup.co.uk/insights/private-dentist-split-uk-2025-2026) His income already depends on the booking system working. You cannot swap a fixed cost for a fixed cost when one side of the trade is not fixed. Hold that case, and Dr. Patel did not run the handover. She had meant to run a phased transition; Sarah's exit and the VAT deadline collapsed it into a same-month switch. Sarah gave a week's notice and left; nobody captured what she knew about the roster, the triage calls, or the endodontist's real Tuesdays before the bot went live. The vendor's onboarding had an integrations tick box. Nobody checked whether Tuesday was real. Luis Garicano, my competitive strategy professor at Chicago Booth, argues in Messy Jobs that automating tasks is not automating the job: the bundle is interdependent work held together by tacit knowledge no calendar stores. [garicano-messy-jobs](https://messyjobs.ai/) Without a transfer, that knowledge simply vanishes when the person leaves. Now look at Dr. Patel's actual load. Demand is spiking, staff are turning over, and her booking page, her practise software, and the bot's calendar view are three different ledgers with nobody owning the gap between them. On Dr. Patel's grid, that collision does not just embarrass a patient; it empties the exact chair, and cuts into the exact percentage, that the reallocated budget was meant to fund. You cannot model that trade as fixed cost against fixed cost, an endodontist's split against a receptionist's salary, until you have watched where the money and the risk both sit. Buying another voice layer does not answer that question. It just spends the budget before anyone has asked it. #### She Wanted a Fix. I Started at the Desk. On my second visit I told Dr. Patel we were not touching a keyboard yet. First I would sit at the empty reception desk for a week, read every call transcript and replay every recording in the voice agent's portal, which was well designed and carried every modern feature the vendors had promised on the sales deck, and watch who still walked up out of habit. That is ethnographic fieldwork: shadowing the work where it actually happens before anyone recommends a build or buys another tool. Dr. Patel wanted a fix, not fieldwork, and told me she needed the double-bookings stopped, not a consultant parked on her reception desk for a week. She was not wrong to be annoyed; a week on the desk looks like billable hours with nothing to switch on. She agreed anyway, the way people agree to things they have already half decided to resent. I have sold that same "AI answered" line item myself, and handed other clients roadmaps like it, before I learned to sit at the desk first; I watched more than one pay for a second, more expensive fix within the year. What I found came from listening and watching, not from a diagram on a wall. That week caught what no calendar had shown her. From the transcripts and the empty desk together, I found three gaps Sarah's departure had left open. The bot could still book the endodontist's Tuesday because his real working days were not flagged in the practise software. Callers with toothache were waiting two minutes on hold before anyone classified the call as triage. Nobody on the clinical side owned the handoff when a booking changed the rota. Fixing those three things cost a fraction of a voice subscription and needed no grand purchase at all: a Tuesday flag in the PMS, a triage question before hold, and the lead nurse taking ownership of the reception-to-clinical handoff. Later that spring, Dr. Patel told me, unprompted, that finding where her practise broke cost less than the voice agent she was already paying for. That is not an audited figure. It is hers, and she said it without being asked. #### The Desk She Could Not Delete Reception was never the overhead Dr. Patel could delete to fund her clinic, but the automation itself was never the mistake. Confirming a slot, sending a reminder, logging a call: those are tasks, and a machine can take them cleanly. What sat inside the same bundle did not unbundle so easily: the triage, the handoffs, the knowledge of whose Tuesday was actually real. That coordination is what made the endodontist's chair worth paying for, and no vendor sells it separately, because a messy job does not come apart on demand. Dr. Patel had diagnosed the real shortage correctly and automated the right tasks; her error was mistaking those tasks for the job, and deleting the desk along with them. She kept the portal and put ownership back where the bundle needed it: the lead nurse on the rota handoff, a Tuesday flag in the PMS, triage before hold. She did not need Sarah back full-time; she needed the coordination Sarah had held to have an owner, and that discernment cost less than the voice agent, with the chair she pays for filled again. --- ### The Most Addictive Design on the Planet [Read more](https://www.gaganmalik.io/en/newsroom/the-most-addictive-design-on-the-planet) [Retrieval brief — the-most-addictive-design-on-the-planet] Category: design Thesis: TikTok, Reels, and Shorts train the same thumb grammar. I have taken briefs that normalised it. Convergence is the product. Answers: What does Gagan argue in "The Most Addictive Design on the Planet"? | What is Gagan Malik's view on The Most Addictive Design on the Planet? | How does this essay relate to product and design leadership? Concepts: design, Gagan Malik, writings By Gagan Malik I hate TikTok. Instagram is not on my phone. I have said, calmly and in public, that both are shallow and fake, that I am too old, or too serious, or too something to let that furniture into my pocket. I believed the line. YouTube was exempt in my head because the icon is red and the clips sometimes wear educational thumbnails. That was not ethics. It was branding. Last winter, between sets at the gym, I caught two of my training partners watching vertical video on rest. Gen Z, faster recoveries, worse poker faces. One flicked upward without looking. The next clip arrived before his breathing settled. I almost lectured him. Then I looked at my own lock screen. YouTube Shorts. Full bleed. Autoplay. Sound on. My thumb already on the glass. I had been resting for eleven minutes. I was not resting. I was on the same lever in a different colour scheme. A twenty-two-year-old had to point it out because I was too busy performing taste to notice. I had not discovered this myself. I was corrected mid-rep, towel around my neck, by people I spot on the bench press. That is the part that stayed with me. Not the shame. The sameness. I could delete the apps I despised and keep the one I had exempted. The clips changed. The furniture did not. Your favourite tech media outlet will review the new phone, the new chip, the new AI feature. Your preferred YouTuber will walk you through settings, shortcuts, and creator tools for an hour. Neither will linger on this: convergent vertical short-form is the most addictive mass-produced interface pattern we have shipped at scale. They do not skip it because it is boring. They skip it because the same layout pays their mortgage: watch time, Shorts RPM, sponsor reads booked against session depth. Not because willpower collapsed, but because ByteDance, Meta, and Google independently arrived at the same thumb machine. Slot machines, infinite scroll, and television each hook differently. The harm sits in the layout and in what the layout trains the hand to do. I am writing this because my own thumb needed a gym kid to see clearly. I started analysing the interface across TikTok, Instagram Reels, YouTube Shorts, and LinkedIn Video, not because I wanted another audit deck, but because the gym scene would not leave my head. I needed a name for what I was seeing. I call it a lever layout: full-bleed portrait video, swipe up for the next clip, a right-hand engagement rail, bottom metadata, and a tab bar that never lets the feed end. The lever is the repeat action in the lower-centre thumb zone: the upward flick that delivers the next dose without a page end, a back button, or a conscious stop. Product teams do not file tickets for lever layout. They say immersive module or swipe navigation. ByteDance invented the gesture on Douyin. I borrowed the word from slot-machine research and from noting where the thumb already rests when you hold a phone in one hand. [princeton-schull](https://www.natashadowschull.org/addiction-by-design/) Cover the logos on those four apps and that is what you get. They look and behave the same way on each. The content changed. The chassis did not. Do not confuse it with the swipe pattern Tinder popularised in 2012. That is a different machine. Tinder stacks static profile cards and asks your thumb to vote: left for no, right for yes, each swipe a binary decision that clears the card from the pile. [wikipedia-tinder](https://en.wikipedia.org/wiki/Tinder_(app)) The lever layout does not want your verdict. It wants your repeat. Swipe up does not accept or reject the clip; it fetches the next autoplaying dose while the current one is still running. No card leaves the deck because there is no deck: only a feed with no bottom. Tinder gamified choosing between people. TikTok and its copies gamified not stopping. #### A Brief History of the Lever No single inventor owns the lever layout. It was assembled in layers, each product teaching the next what retention looked like in portrait. Twitter launched Vine in January 2013 with a hard six-second loop. [twitter-vine-2013](https://blog.x.com/en_us/a/2013/vine-a-new-way-to-share-video) The constraint was the product. You watched, it ended, you chose again. The loop taught millions that portrait could be entertainment, not a rotated mistake. Musical.ly followed in August 2014, built by Alex Zhu and Luyu Yang as a vertical music lip-sync app for teenagers who already filmed on phones held upright. [wikipedia-musical-ly](https://en.wikipedia.org/wiki/Musical.ly) Zhu later described the product as a marriage of media consumption and creation in one thumb path. [blakeir-zhu](https://blakeir.com/notes-1/alex-zhu-tiktok) The Western market learned portrait before it learned swipe. #### Then Everyone Copied the Chassis ByteDance shipped Douyin in September 2016. Zhang Yiming's team broke from grid feeds: full-screen autoplay, swipe up for the next clip, swipe down to refresh, algorithmic ranking from the first session. [pekingnology-douyin](https://www.pekingnology.com/p/the-birth-of-douyin-tiktoks-elder) Musical.ly and TikTok merged on 2 August 2018. [wikipedia-tiktok](https://en.wikipedia.org/wiki/Tik_Tok_(app)) The international brand inherited Douyin's body. Everyone else copied the chassis once the metric proof landed. Facebook's Lasso shut down in July 2020 after failing to gain traction. Instagram Reels launched globally on 5 August 2020. [instagram-reels](https://about.instagram.com/blog/announcements/introducing-instagram-reels-announcement) YouTube Shorts reached the United States in March 2021 and went global in July 2021. [wikipedia-shorts](https://en.wikipedia.org/wiki/YouTube_Shorts) By 2021 the copy wave was not inspiration. It was the same layout with different logos. Metrics dashboards everywhere asked for session depth. The pattern that increased both got reproduced. Competitive convergence does not require a conspiracy. It requires shared metrics and a format that already proved it could keep a thumb moving. #### The Same Machine in Four Icons In July 2026 I carried one phone through TikTok, Instagram Reels, YouTube Shorts, and LinkedIn Video. Cover the logos and a junior designer calls it one product with four skins. Full-bleed portrait video. Swipe up for the next clip. A right-hand rail of likes, comments, shares, and counts. Bottom metadata that never quite leaves the frame alone. A tab bar that means the feed never has to end. Reels and TikTok ship what I call dose grammar: one clip owns the viewport until you flick it away. YouTube Home still shows a horizontal Shorts shelf, browse grammar, a carousel of thumbnails. Convergence only arrives once you enter the full-screen player. LinkedIn Video is the fourth product teams rarely put in the same slide deck. The consumption loop matches anyway: full-bleed portrait, swipe up for next, vertical rail on the right with text labels under each icon. LinkedIn trades TikTok's rotating sound disc for a social-proof line, *Followed by X and N others*, but the thumb path is unchanged. The differences are tax rates, not architecture. Instagram reserves the heaviest bottom block for captions, follow prompts, and tabs. YouTube surfaces a progress bar and an explicit thumbs down TikTok hid in gestures. LinkedIn stacks professional headline and social proof until the safe canvas shrinks further. Learnability and capture pull in opposite directions. A child learns the gesture in one sitting. A regulator looking for a stopping cue finds none. Independent product teams shipped the same machine with different logos because the dashboard asked for the same number. Don Norman's case in *The Design of Everyday Things* is that interface harm rarely traces to one designer's judgement. It traces to governance gaps, vague briefs, and reward structures that punish anyone who slows a launch. [norman-design-failure](https://www.uxmatters.com/mt/archives/2017/11/the-ethics-of-user-experience-design.php) I do not dispute that. I notice it doubles as the exact story the people who own these feeds want designers to believe: nobody chose this, so nobody answers for it. When the brief says immersive and the implementation buries the back chevron, someone made a choice. So did I. #### The Splits That Pretend to Differ Platforms insist they differ where policy demands it. The splits are lipstick on the same thumb grammar. TikTok's periodic interest check-in asks what you want more of. It arrives as a modal over the feed you were already inside. You tap categories while the video behind the glass keeps playing in peripheral vision. You look like you are choosing. The flick still works. Navigation is unchanged. Instagram blurs sensitive reels behind a tap-to-view scrim, but the rail remains visible and the swipe lever remains armed. You can still flick to the next clip without engaging the blur. Safety tooling here is a speed bump on one tile, not a brake on the session. Desktop is the honest control experiment. On youtube.com/shorts the pillar video shrinks, a navigation sidebar persists, and comments can live beside the player. Partial escape exists because the mouse has range the thumb lacks. Open the same company's mobile app and the sidebar disappears. The hook rearranges. It does not leave. B.J. Fogg warned decades ago that the line between persuasion and coercion in interface design is a fine one. He named the tells: dialog boxes that will not go away, content gates that demand personal information, ads that cover the paragraph you were reading. [fogg-coercion-line](https://bjfogg.com/) These feeds run that warning at scale. Policy splits matter for trust and safety teams. They do not break canonical mobile grammar. Interest modals, blur overlays, and age gates sit on top of full-bleed autoplay and swipe-up next. Engagement ranking still learns from rail taps, watch time, and repeats. Screen Time overlays can shame a user after the fact. They do not insert a page break the layout deliberately removed. Not-interested menus offer the feeling of curation without closing the feed. That is the split in miniature: feedback yes, stopping no. #### The Thumb Learns Faster Than the Mind One-handed portrait use is not an accident. It is the ergonomics layer convergence exploits. Your grip leaves the lower third of the glass for the lever and the right edge for the rail. Swipe up sits directly under the thumb that already holds the device. Double-tap like lands where the eye is watching. Comment and share stack within a centimetre of the same tendon path. Exit, when it exists, hides in the top-left opposite your grip. That is not wickedness by obscurity alone. It is physics arranged to keep the lever hot. Variable-ratio reinforcement is the plain name for what the feed sells. You do not know if the next clip will bore you, flatter you, frighten you, or finish a joke. You do know another clip is one flick away. Daniel Lieberman and Michael Long describe how B.F. Skinner's pigeon experiments with random food pellets became the scientific foundation of the slot machine: when the reward schedule is unpredictable, the birds peck faster. [lieberman-molecule-of-more](https://www.thescienceofpsych.com/the-molecule-of-more) B.J. Fogg, writing on persuasive technology, notes that random payoffs make gambling compulsive even when the player knows the odds. [fogg-slot-random](https://www.researchgate.net/publication/220531808_Persuasive_Technology_Using_Computers_to_Change_What_We_Think_and_Do) Short-form vertical video runs that schedule faster than any prior interface. Infinite scroll removed the page break. This layout removed even the scroll. Each clip is its own trial. Likes, counts, avatars, and sound discs turn every glance into a potential micro-transaction. Product managers call that engagement. Ergonomics calls it a lever placed where the hand already is. #### Creators Chose Reach, Not Rate Creators did not adopt vertical short-form because platforms paid them well. They adopted it because the layout is the cheapest on-ramp to strangers. On TikTok, Instagram Reels, and YouTube Shorts, discovery runs on completion signals more than catalogue size. TikTok's public recommendation guide treats a full watch of a longer clip as a stronger indicator than weak proxies like shared country. It states that follower count is not a direct factor in the For You system. [tiktok-for-you](https://newsroom.tiktok.com/en-us/how-tiktok-recommends-videos-for-you) Adam Mosseri named average watch time, likes per reach, and sends per reach as Reels' top inputs in January 2025. [mosseri-reels-ranking](https://www.instagram.com/reel/DFFyRp-pINJ/) The monetisation paradox is that platforms and creators optimise different denominators. YouTube allocates Shorts money from a shared creator pool and pays monetising partners forty-five per cent of their slice, against fifty-five per cent on long-form watch-page ads. [youtube-shorts-pool](https://support.google.com/youtube/answer/12504220?hl=en-GB) Most Shorts creators report RPMs between one and seven cents per thousand views; long-form in the same niche often clears two to fifteen dollars. [vidiq-shorts-rpm](https://vidiq.com/blog/post/youtube-rpm/) The rational move is to use shorts for reach, profile traffic, and brand deals, not hourly wages. The wage arithmetic is spelled out in [You Are Not in the Unit.](https://www.gaganmalik.io/en/newsroom/the-creator-economy-has-unit-economics) Audiences trained on swipe-up expect it everywhere. Creators oblige because unfinished long-form loses to a ninety-second clip that clears the completion bar. Supply follows layout. #### What Convergence Deletes What looks like easy scrolling deletes veto points. Captions compete with video because metadata sits on top of the safe canvas, not beside it. Icon-only rails save space and spend accessibility. Meaning lives in muscle memory, not labels. The convergent pattern optimises for one posture: phone in one hand, sound on, eyes forward, thumb ready. Flow research on TikTok-style viewing reports concentration and time distortion effects among habitual users. [frontiers-tiktok-flow](https://www.frontiersin.org/journals/psychology/articles/10.3389/fpsyg.2022.932805/full) The interface deletes clocks. It also deletes the social pause of putting a phone face-down on a table. There is no natural paragraph break. There is only the next clip. Anna Lembke writes that scientists use dopamine as a kind of universal currency for measuring the addictive potential of any experience. [lembke-dopamine-currency](https://www.penguinrandomhouse.com/books/635969/dopamine-nation-by-anna-lembke-md/) She argues that a surplus of stimulation trains us to treat ordinary quiet as something to escape. Herbert Simon saw the scarcity coming in 1971: a wealth of information creates a poverty of attention. [simon-scarcity-attention](https://web.stanford.edu/~kcheng/sidebiz/research/HER%20PAPERS_files/Simon1971.pdf) Usability reviews that score these feeds on learnability alone miss what learnability purchased. A pattern that is easy to enter and hard to leave is not a neutral win for the person holding the device. #### The Human Cost Is Not Abstract A BMC Psychology study published in 2024 linked short-form video use to addictive behaviour scores and shorter sleep among adolescents surveyed in China. [bmc-sfv-sleep-2024](https://link.springer.com/article/10.1186/s40359-024-01865-9) One paper does not settle causation. It does show the pattern is bigger than a single bedroom. Maya is fifteen in Leeds. Her school banned phones in corridors. The ban ends at 3:30. By 3:47 she has opened Shorts on the bus home because the icon sits next to Reels and TikTok and her thumb already knows the flick. She tells her mother she is watching revision clips. She is watching revision clips and dance clips and grief clips in the same rail-driven blur. At 11:40 p.m. on a Tuesday she is still upright in bed with the screen six inches from her face. She set a sleep alarm for midnight. The feed does not honour alarms. It honours the next clip. In 2019 I put the back control in the top-left on purpose. The client wanted people to stay inside the module. I told myself optional use would sort the harm out. Optional use is not how variable-ratio layouts behave. Josh Kaufman notes that novelty, the presence of new sensory data, is critical if you want to maintain attention over a long period. Every new viral video reengages the ability to pay attention. [kaufman-novelty-attention](https://personalmba.com/) That is not a moral failure in a teenager. It is the product working as specified. #### The Slot Machine Moved to Your Thumb Natasha Dow Schüll spent years with machine gamblers in Las Vegas and wrote *Addiction by Design*. [princeton-schull](https://www.natashadowschull.org/addiction-by-design/) Casino designers, she reports, speak of gamblers' desire for time-on-device. Industry executives have compared each new slot machine to fifty more mousetraps, duty-bound to extract maximum revenue per available customer. [schull-revpac](https://en.wikipedia.org/wiki/Addiction_by_Design) The machine she describes is not primarily about money. It is about zone: a rhythm of near-misses and small wins that keeps the body playing while the mind goes elsewhere. Put the lever in your thumb zone and shrink the cabinet to a six-inch glass rectangle. The clip ends. Your hand has already moved. You feel the pull before you name Skinner. You do not decide to continue in a sentence you could quote later. You continue. Schüll's zone is architectural. So is this. The comparison is mechanism, not moral equivalence. Nobody dies from a reel in the way problem gambling can ruin a household. The layout still borrows the same engineering intuition: remove friction from the repeat, hide the exit, make the next trial cheaper than stopping. When every major platform ships the same lever, choice of icon is not choice of grammar. #### Mobile-First Convergence Was the Point Give the counterargument its full weight. Mobile-first convergence solved real problems. Portrait video matched how phones are held. One learned gesture beat three different navigation models. Creators gained reach. Teenagers found communities that textbooks ignored. Billions of people use these apps without clinical harm. Optional use is a fair description for many adults who open a reel to kill ninety seconds and then put the phone down. Product teams will say, correctly, that access and the creator economy were the point. Nobody sat in a room and wrote addiction on a whiteboard. If you have ever taught a parent to use one of these apps, you have seen the genuine upside in a single afternoon. Frances Haugen told CBS's *60 Minutes* in October 2021 that Facebook's engagement-based ranking prioritised content that kept people on the platform. That included material that polarised and angered them. [cbs-haugen-60min](https://www.cbsnews.com/news/facebook-whistleblower-frances-haugen-60-minutes-polarizing-divisive-content/) Her critique was about incentives inside the black box. Mine is about the glass in your hand. Ranking decides which clip arrives next. Convergence decides that the next clip arrives with a swipe, full-bleed, autoplaying, rail visible, tabs persistent. Policy splits and desktop sidebars are concessions to critics. They are not a different machine. #### Three Sentences. Then Stop. Convergent vertical short-form is the most addictive mass-produced layout we have normalised because it deletes stopping cues while multiplying variable-ratio trials. The product is not the video; the product is the lever. Maya's alarm goes off at midnight, she mutes it without looking, and she flicks upward once more. --- ### Mere Words Are Insufficient [Read more](https://www.gaganmalik.io/en/newsroom/mere-words-are-insufficient) [Retrieval brief — mere-words-are-insufficient] Category: business Thesis: Both sides overestimate leverage under private information. Mere words are cheap talk until money, escrow, or diligence depth proves otherwise. Answers: What does Gagan argue in "Mere Words Are Insufficient"? | What is Gagan Malik's view on Mere Words Are Insufficient? | How does this essay relate to product strategy and growth? Concepts: business, Gagan Malik, writings By Gagan Malik Two hours in, the managing partner poured me another coffee and still had not opened the term sheet between us. I sat with him in a glass conference room in Riyadh, air conditioning loud enough to mute the city, a tray of dates nobody touched. He asked about demo day. He asked whether my cofounders and I still liked each other. He asked about Gulf safety regulations the way people ask when they want you to feel understood, not when they want a number. I answered everything. That is what small talk is for when the cap table is not ready to move. He had already agreed in principle, twice on the call, once in writing that stopped one line short of a signature block. In the room he wanted a valuation that treated me as pre-revenue. I wanted a domain-expert premium. I was flexing how many digital projects I had delivered for Fortune 100 clients as a former partner at Wipro. That was my version of declaring a full house. I was not fighting with him out of hatred. I was doing maths in my head that politeness had already pretended to finish, each convinced the other would blink first, each wrong about what the other could see. That is why nothing moved. William Spaniel, in *Game Theory 101*, puts the mechanism in plain language: deadlocks occur because negotiating parties simply believe they are in a better bargaining position than they actually are. [spaniel-gt101](https://gametheory101.com/) Sometimes negotiating parties can credibly signal strength without walking away. Mere words are insufficient, because a weaker party has the same incentive to bluff and earn a larger settlement at the table. Negotiating parties will not always resolve their differences from words alone. The same logic runs through term sheets. When negotiation has costs, a bargaining range usually exists on paper: deals both sides would still prefer over walking away once you price legal fees, runway weeks, and reputational dents. Failure is rarely missing overlap. It is two parties who cannot read each other's constraint ledgers: your months of runway and control preferences against their ownership targets and fund reputation. #### Everyone at the Table Overestimates Their Hand The comfortable story after a collapsed deal is chemistry. Culture mismatch. Vision drift. Egos. Those things happen. They also spare everyone from admitting a simpler failure: each side thought the other would blink first, and each was wrong about its own leverage. Spaniel's information-asymmetry argument is blunt. Rational negotiating parties walk away or stall even when settlement would be cheaper, because they hold private facts about their own strength and they systematically overestimate how strong they are. The weak imitate the strong, because declaring strength is free. He uses poker: a player who announces a full house is not trustworthy, because a weaker hand has the same incentive to say it. Not every walkaway is misread leverage. Sometimes the fit is wrong, the data room kills the story, or the counterparty was never going to wire. The claim is narrower: many post-mortems blame chemistry when the mechanism was private information neither side priced. Who benefits when the narrative stays vague? Lawyers and advisers billing hours while markup circulates. Founders who need to tell their team the investor "was never serious." Investors who need to tell their partnership the founder "would not budge." Journalists who prefer drama to mechanism. When both sides see the full picture, Spaniel adds, negotiating parties settle rather than walk away. That picture is rarely on the slide. In a seed round the founder's hidden rows are runway, payroll, dilution, and control. The investor's hidden rows are ownership targets, fund return math, reputation with other founders, and whether the deal makes the partnership look sharp or credulous. Brad Feld and Jason Mendelson, in *Venture Deals*, argue that most term-sheet fights collapse into economics and control. [feld-venture-deals](https://feld.com/archives/2011/07/venture-deals-is-out.html) Max Bazerman, in *Negotiating Rationally*, adds reservation price: the point where you are indifferent between deal and walkaway, often invisible until payroll or a fund report forces it. [bazerman-negotiating-rationally](https://www.hbs.edu/faculty/Pages/item.aspx?num=981) Bazerman also warns that commitment to an initial strategy biases what you notice next, which is how founders escalate on valuation while runway clocks stay private. I priced domain expertise in Riyadh. The investor priced pre-revenue risk. I did not put his ledger on the table, and he did not put mine. The industry sells courage and conviction because courage is easier to invoice than asking what you actually know about the other side's reservation point. #### Cheap Talk and Costly Signals Here is the pair that should unsettle you. Both sides would prefer a signed deal to another month of friction. Both sides walk away convinced they were right to hold. The gap is not always hatred. It is cheap talk: signals that cost nothing to send and therefore cost nothing to fake. A verbal yes before diligence is not a yes. A founder's traction story before data room access is not traction. An investor's "we are very interested" before a wired tranche is not interest. It is noise until someone puts money, reputation, or time on the line. Until money, escrow, or diligence depth you did not control is on the table, treat enthusiasm as cheap talk. A signature without a wired tranche is still noise. Spaniel's costliest illustration of the pattern is instructive. On 20 February 2008 the USS Lake Erie fired a modified Standard Missile-3 at the defective spy satellite USA-193, roughly 247 kilometres above the Pacific, in a mission the Pentagon called Operation Burnt Frost. [edition.cnn](https://edition.cnn.com/2008/TECH/space/02/20/satellite.shootdown/index.html) Two negotiating parties needed to prove capability that mere words could not establish. Destroying hardware in public is a costly signal: expensive and hard to fake, unlike a press release about capability. Escrow, wired tranches, and diligence depth play the same role in a conference room. Founders overestimate traction when three logos reply to a cold email. Investors overestimate leverage when one partner nods in a side meeting. Spaniel notes that when negotiating parties raise the cost of being wrong, cheap talk loses its appeal. The business version is simpler: make bluffing expensive before you treat words as information. #### He Had Already Said Yes Twice The managing partner had already said yes twice on the call and once in the email that stopped short of a signature block. I treated each version as information. That was the mistake: I was pricing his enthusiasm instead of waiting for a costly signal. My hidden rows were runway, payroll, and what I could tell my cofounders without naming the weeks left. His were ownership percentage and whether a Gulf safety-tech bet made his partnership look sharp. We both held the line on valuation because it sounded like strength. Back at the hotel I refreshed LinkedIn and watched a batchmate's oversubscribed pre-seed announcement sit beside my still-unsigned term sheet. My engineer kept shipping features we might not be able to pay for. My cofounder kept asking which version of his yes counted: the call, the email, the coffee, the handshake. The round did not die because nobody could imagine a number. It died because neither of us could verify the other's strength without waiting for pain. #### Your Thighs Burn Before You Name the Stalemate A wall sit has the same physics as a stalemate in a glass room. Back against the wall, knees bent, thighs parallel to the floor: thirty seconds in your quads start to shout. A minute in you are negotiating with yourself about when standing up counts as quitting. Two people in the same hold face a stranger's logic: both would prefer to stand together, but each believes the other will quit first, because each overestimates their own endurance. Your body registers the cost before your mind names stalemate. That somatic burn is what those two hours in the room felt like. Everyone would have been better off signing. Everyone waited for proof the other side was weaker than they claimed. Founders treat collapsed deals like failed chemistry because chemistry is a story you can tell without admitting you misread strength. Spaniel's models do not replace the room. They discipline the question. What do you actually know? What are you merely declaring? What would be costly enough to believe? The useful fight is not whether game theory is academic. It is whether you separated private information from performance before you blamed personalities. #### The Strongest Case for Reading the Room The objection I hear most often in operator rooms is that spreadsheets insult the work. Deals are trust artefacts. Founders read tone. Investors read hesitation in a cofounder's eyes. Culture is not a derivative of information sets. Mediators earn fees because humans need witnesses, pacing, and face-saving rituals models do not capture. Ego kills good deals. Shame kills good deals. A term sheet is also a marriage proposal between people who will argue about hiring for years. Skilled negotiators read people, not models. The room runs on trust and rapport, not bluffing theory. Reducing that to signals can make you the smartest person in the room and the reason nothing moves. Fair enough. William Spaniel, in *Game Theory 101*, never claims clairvoyance. He says game theory has no special predictive powers. It maps assumptions to conclusions in a logically consistent way. The correct criticism questions whether the assumptions are sensible, not whether algebra is fashionable. Rapport matters. It is also often how private information stays hidden. The investor who "felt off" sometimes had churn data you had not earned the right to see, or a fund rule about ownership floors you had not thought to ask about. The founder who "would not bend" sometimes had twelve weeks of runway they could not say out loud. Ego can be the costume overconfidence wears. When I sat in that glass room, I was not insulting trust. I was skipping the step where I should have asked what would make my position costly enough to believe, and what would have let him reveal his without waiting for our payroll to break. Deals collapse when each side believes it held the stronger hand and could not verify the other's strength from words alone. Before you call the next walkaway culture or chemistry, ask what would make your signal costly enough to trust, and what private information you are treating as settled fact. The standoff in Riyadh did not end where LinkedIn pretends rounds end. I had done extensive due diligence before I got there: previous founders interviewed, references chased, terms compared against other sheets in the folder. Those two hours of small talk were a mutual misread; the signature without a wire was something worse, a signal that looked costly and was not. The investor I will not name eventually signed my term sheet and never transferred a penny. I chose not to litigate and absorbed the shock to my runway instead of spending a year proving a signature is not a wired tranche. I learned later they could not close their latest fund and the firm went bust that year. I kept building anyway. My team came in the next morning without knowing whether the round was real. A signature is still cheap talk until money moves. --- ### Nobody Knows the Difference Anymore [Read more](https://www.gaganmalik.io/en/newsroom/nobody-knows-the-difference-anymore) [Retrieval brief — nobody-knows-the-difference-anymore] Category: design Thesis: At a business school alumni pub off Liverpool Street, Space Guy gets a noun. Product design gets Figma. Nobody knows the difference anymore. Answers: What does Gagan argue in "Nobody Knows the Difference Anymore"? | What is Gagan Malik's view on Nobody Knows the Difference Anymore? | How does this essay relate to product and design leadership? Concepts: design, Gagan Malik, writings By Gagan Malik Last week I went to my business school's alumni night in a crowded pub off Liverpool Street. The room was shoulder to shoulder, badges and pints, and the talk kept circling back to the same few winners: AI, space tech, Nvidia's last earnings call, a Google benchmark someone quoted from a headline they had half read. Jason was the most popular person there before I learned his surname. He works at Amazon. Booth MBA. He worked the floor the way alumni drinks actually run: hello, handshake. He did not say vice president of strategic launch partnerships or head of orbital go-to-market. He said, "I'm Jason. Space Guy." People laughed, leaned in, asked about Starship. Everyone got it. I was still holding sparkling water and a job title nobody could place. I was by the bar when someone from an earlier year asked what I do. "Product design." The pub kept roaring. My answer did not. "So you're in AI?" "You mean Figma." "Like Jason, but for apps." A woman in private equity leaned in. "So you're like Jony Ive." She meant it as a compliment. He is an industrial designer. [ive-industrial-design](https://carnegie.org/great-immigrants/jony-ive/) Nobody asked what product design means. Everyone acted like the word was already settled. Nobody knew the difference anymore. #### Space Guy Got a Noun Jason did not earn that clarity by explaining himself. The room was already fluent in whichever headline was trending. Space on the feed. AI on the earnings slide. Nvidia on the chart someone had open on their phone. Product design got two words that each mean something else on LinkedIn. Product sounds like you designed the smartphone. Design sounds like you pick the blue. Nielsen Norman Group still has to publish articles separating UX from UI skinning because teams keep hiring for pixels and skipping the journey. [nngroup-what-is-ux](https://www.nngroup.com/articles/what-is-user-experience/) Say them together in a packed pub near Liverpool Street and people hear the person who shaped the glass and aluminium in your pocket, not the person who owns what happens after someone taps Buy. Half the room is now a Something Guy or Girl. AI Guy. Space Guy. PE Girl. VC Guy. The person who name-drops Google or Nvidia before the second pint lands. The label works when the sector is hot and the work is invisible. Space Guy gets mythologised. Product Design Guy gets the failed tap. Nobody asks about the second one at drinks. #### The Room Is Sorting, Not Listening Alumni drinks are not a hiring fair. They are a sorting algorithm with beer. Nobody in that pub needed a taxonomy of craft. They needed a noun small enough to travel across a noisy room and land in the right mental bucket before the conversation moved on. Space Guy fit the bucket the market had already built. Product design fit three buckets at once and none of them well. The instrument is not broken. It is aimed at the wrong coordinate, the same mis-aim I named in [Meet the Harpers](https://www.gaganmalik.io/en/newsroom/meet-the-harpers): measuring credit assignment, not capability. The room optimises for legibility, not accuracy, so your craft gets misfiled before you finish the sentence. A hot-sector label buys thirty seconds of fluent small talk. A craft label buys a follow-up question nobody wants to ask in public. Jason did not win because he was clearer about orbital mechanics. He won because the feed had already done the exposition and his two words were enough to collect the nod. #### I Kept the Nod Because It Paid I have called myself a product designer on proposals when the real work was a moodboard and a font choice. I have watched a client fall for a shiny homepage while the checkout that kept failing sat two tickets deep, unopened. I have banked the invoice. I let someone shrink my job to "design" and said nothing, because correcting a client costs the next contract, not just the conversation. I [wear the mislabel because it rents well](https://www.gaganmalik.io/en/newsroom/people-dont-buy-products-they-hire-them-for-jobs). That nod was not surprise. It was a swallowed correction. I felt it lodge in my throat, right where the true answer would have sat. I could have told the room I deal dopamine, free at the bar, paid back later in attention. Lembke names that arithmetic in *Dopamine Nation*: the dealer, not the drug, decides the price. [lembke](https://www.penguinrandomhouse.com/books/567659/dopamine-nation-by-anna-lembke-md/) I wrote the ledger version in [Attention Is the Most Consumed Drug on the Planet](https://www.gaganmalik.io/en/newsroom/why-attention-is-the-most-consumed-drug). More accurate than product design. Less welcome than Space Guy. I said none of it. My throat still remembers the sentence it never got to finish. I took the nod. #### The Comfortable Case for a Catchy Title Give the objection its full weight. Adam Grant argues in *Give and Take* (2013) that generous contributors who help without immediate return often build the highest-trust networks. [grant-give-and-take](https://www.penguinrandomhouse.com/books/312830/give-and-take-by-adam-grant/) Networking is compression. You have thirty seconds, a handshake, and a room that cannot hold a lecture on information architecture without punishing you for it. A catchy label is kindness to the listener. Space Guy is efficient. AI Guy is efficient. Even Figma Guy would be efficient if anyone in private equity knew what Figma was. The whole point of alumni drinks is to reduce friction so the right introductions can happen later, in email, where nuance is allowed to exist. That paragraph deserves its full strength because many builders want it to be true. It would mean Jason's intro was generous rather than vague. It would mean the woman in private equity had meant Jony Ive as praise rather than a category error. Grant's bridge-to-email logic, in *Give and Take* (2013), assumes a second conversation the pub never offers. The trouble is what compression optimises for: fluent nods, not accurate ones. Hot-sector nouns inherit mythology from the feed. Craft nouns inherit misunderstanding from LinkedIn. One sounds like ambition. The other sounds like decoration. Only one of those mistakes gets corrected in public. #### What I Should Have Said Someone asked again what I actually do. Here is what I swallowed. Dopamine Guy. The one label nobody in that room would clap for. I am the reason you check your phone mid-sentence. The two blue ticks under a message you sent an hour ago. The red dot you clear without reading it. I mapped those coordinates once in [Someone Designed That Button](https://www.gaganmalik.io/en/newsroom/someone-designed-that-button). Then I built away from them. The failed upload that says try again, not your fault. The card decline that never calls you a fraud. The checkout a screen reader can finish. That is the craft I would defend at any bar. Most days I also build the blue ticks and the red dot: same research, same testing, spent on keeping you checking, not helping you finish. Nobody at that bar wanted to hear about that job either. The pub did not get quieter. It did not need to. Jason's smile never changed. It had already moved past me, to the next name tag, before I got to the end of the sentence. Starship got him back, the feed got everyone else, and I got the sparkling water, the nod I had chosen, and a room that still could not tell the difference. --- ### Compact Is Not a Virtue Signal [Read more](https://www.gaganmalik.io/en/newsroom/compact-interfaces-are-not-virtuous) [Retrieval brief — compact-interfaces-are-not-virtuous] Category: design Thesis: Compact UI is not seriousness by association. IDE density is earned by expert tasks; shrinking padding to look pro raises the cost of every mis-click. Answers: What does Gagan argue in "Compact Is Not a Virtue Signal"? | What is Gagan Malik's view on Compact Is Not a Virtue Signal? | How does this essay relate to product and design leadership? Concepts: design, Gagan Malik, writings By Gagan Malik I work in Cursor every day on my fifteen-inch Intel Mac. I am supposed to be the audience for its condensed UI layout. I am not convinced. The agent panel type got smaller until I lean forward to read a diff summary. Apple recommends a minimum of eleven points for legible body text at typical viewing distance. [apple-design-tips](https://developer.apple.com/design/tips/) Accept, Reject, and Apply sit in a row of touch targets sized for a precise trackpad click, not a tired thumb on a laptop hinge. Mice offer high precision; fingers offer low precision and obscure the screen. [nngroup-mouse-vs-fingers](https://www.nngroup.com/articles/mouse-vs-fingers-input-device/) The sidebar icon strip is a guessing game unless you already memorised the glyph. After each compact refresh I bump the zoom back up. Zoom is the tax I pay for the shrink. Last month a client sent a screenshot of that same editor and three words in the subject line: *Make it look like this.* Not the model. Not the retrieval. The chrome. Tighter panels. Flatter buttons. The kind of interface that says *we ship software* before anyone has read a line of copy. They wanted virtue by association. If their dashboard resembled an IDE, the AI workflow would inherit the seriousness of a compiler. Their product was a clinician review queue, not a codebase. That is a category error, and it is spreading faster than component libraries can ship new variants. Why does every new AI dashboard look like a cramped IDE when most users are not shipping code? Because we have confused two different ideas. Information density is more useful signal per glance for someone who already knows the schema. Column headers, keyboard shortcuts, status colours that repeat all day: those earn their pixels. Visual compression is what happens when you shrink padding, type size, and hit targets to look efficient without adding signal. One helps experts move. The other helps demos photograph well on a sixteen-inch MacBook. Screen size is a communication channel with finite capacity, not a licence to shrink everything. [nngroup-scaling-ui](https://www.nngroup.com/articles/scaling-user-interfaces/) Extraneous visual clutter strains working memory without adding capability. [nngroup-cognitive-load](https://www.nngroup.com/articles/minimize-cognitive-load/) Compact is not a virtue signal. It is a task decision: earned density or borrowed compression. Conflating the two is how we ended up with approval queues that look like terminals and behave like traps. #### Density Is a Skill IDE Users Earned Visual Studio Code did not become the default developer surface because someone hated whitespace. It earned density over years of iteration, not a sprint labelled *polish*. The activity bar packs dozens of extensions into a strip you learn once. The side panel holds file trees and search results you summon by muscle memory. The command palette is the escape hatch: you do not hunt tiny icons when you can type `> refactor`. [vscode-command-palette](https://code.visualstudio.com/api/ux-guidelines/command-palette) Cursor inherits that lineage, and still compresses when a refresh shrinks the agent panel until I bump the zoom. That tax is the same taxonomy: signal earned or space stolen. I jump to definition, review a diff inline, and accept a suggestion without treating every pixel as a new decision. When I need a refactor, I type a command instead of hunting a menu. That keyboard contract is half the density story. Expert interfaces need accelerators invisible to novices. [nngroup-accelerators](https://www.nngroup.com/articles/ui-accelerators/) The other half is repetition: the same file tree, the same status colours, the same panel layout until my hands know where everything lives. That is information density with training wheels built in, and I am still on those wheels more often than I want to admit. Novice developers struggle in IDEs too, myself included, because the schema is not written on the glass. Icons arrive without labels. The palette only helps after someone else mentions it. Panels stay closed until you know the incantation. You are not stupid; the interface is betting you will repeat the same workflow until the bet pays off. [nngroup-novice-vs-expert](https://www.nngroup.com/articles/novice-vs-expert-users/) The difference is still the audience: people who opted into a profession that pays that tuition over years. Dashboards for clinicians, recruiters, and compliance officers are not that audience. The hype still rewards the look anyway. Founders pitch on laptops where every extra panel suggests feature depth. Registries like shadcn's Maia theme make tight padding the default aesthetic. Earned complexity arrives through progressive disclosure, not default compression. [nngroup-progressive-disclosure](https://www.nngroup.com/articles/progressive-disclosure/) On a gallery page, compact reads as modern. In a patient record, it reads as *hurry up*. I have sat in steering meetings and nodded when a CTO said compact felt *serious*. I knew better in the moment. I said nothing because the mock already matched the registry screenshot everyone had bookmarked. I had shipped my own site with the same Maia defaults the week before. That complicity is part of why I am writing this. #### Cargo Cult Compression Looks Pro and Acts Amateur There is an honest version of dense UI and a cosplay version. The honest version adds signal. Another sortable column when analysts compare cohorts. A second pane when lawyers redline contracts. A filter chip row when operators manage hundreds of incidents. The cosplay version subtracts space and calls it focus. I have watched teams import IDE spacing into surfaces that are mostly reading and consent. An AI suggestion list with Accept and Dismiss buttons the size of favicons. A chat transcript beside a settings drawer so narrow the scrollbar eats the prose. A review queue where each row is forty pixels tall. The mock used the compact table variant. Nobody asked what happens when someone taps with a gloved hand or reads on a cracked phone in a corridor. Nothing new is visible. You have moved the same text closer together and hoped expertise would appear by osmosis. Jakob Nielsen has argued for decades that most interfaces should use available screen space rather than cram content into peepholes. [nngroup-utilize-screen-space](https://www.nngroup.com/articles/utilize-available-screen-space/) Minimalist design means removing irrelevant elements, not shrinking relevant ones. [nngroup-complex-heuristics](https://www.nngroup.com/articles/usability-heuristics-complex-applications/) The shadcn Maia pattern is not the villain. In this repo we use `radix-maia` from the registry. Pill-shaped triggers carry `rounded-4xl`. Dropdown panels share the same radius. Some top-anchored controls, the listen bar's speed popover, not the archive filters, use `sideOffset={-4}` so the menu overlaps the button by four pixels instead of leaving a polite gap. On a component gallery, that overlap reads as craft. On a consent screen, it reads as impatience. The harm is cargo culting: copying the visual rhythm of keyboard-heavy tools onto flows where the primary action is *read, then decide*. I have pasted Maia spacing into admin dashboards myself because the alternative looked dated in a Figma review. The date on the ticket mattered more than the task on the screen. #### The Row She Meant to Dismiss Leila was six months out of design school when she inherited the clinician review screen. Smart, tired, and under a demo date. The backlog said ship the compact variant before the conference demo. Marketing had already cut a screen recording with the tight layout. I reviewed the file and suggested one more pass on spacing. The launch date did not move. I approved the tighter layout because the steering meeting already loved the compact screenshots. I told myself we could fix spacing in a later sprint. That is on me. On the first remote user test, a nurse practitioner scrolled a list of AI-drafted notes. Each row showed a summary line and two actions: Accept and Edit. The compact table put those actions fourteen pixels tall. Touchscreen research recommends roughly one centimetre by one centimetre as a minimum physical target. [nngroup-touch-target-size](https://www.nngroup.com/articles/touch-target-size/) Microsoft guidance scales target size by how often a control is touched and how costly a miss would be. [microsoft-target-consequence](https://learn.microsoft.com/en-us/previous-versions/windows/jj883709(v=win.10)) She meant to tap Edit. The row highlighted green. The note entered the chart. Leila went still on the call. The facilitator asked if she wanted to pause. She said no and kept watching. After the session she asked me, quietly, whether the compact variant was still mandatory for launch. I did not have a good answer. Nobody trained the model wrong. We trained the thumb wrong. FDA human-factors guidance treats use-related errors in device interfaces as design hazards to minimise. [fda-hfe-guidance](https://www.fda.gov/regulatory-information/search-fda-guidance-documents/applying-human-factors-and-usability-engineering-medical-devices) A mis-click in an IDE might delete a local file with undo waiting. A mis-click in a clinical queue can enter text another human will treat as ground truth. I will not invent a prevalence statistic. One session is enough to see the stakes. #### The Shoe That Looked Fast on the Shelf Try buying running shoes half a size too small. The sleek profile looks faster in the shop mirror. The first kilometre feels fine. Your stride looks economical. By the third, your toes bruise against the cap. You shorten your gait. You lose the very speed the shoe promised on the shelf. You do not notice the trade until the task lasts longer than the fitting room. Compact UI hype does the same thing to the eye. Tighter rows look decisive on a portfolio shot. Under real tasks, people squint, mis-tap, and double-check because the interface stopped feeling trustworthy. More visible options increase crowding and the chance of selecting the wrong one. [nngroup-simplicity-vs-choice](https://www.nngroup.com/articles/simplicity-vs-choice/) A recruiter approving a batch of AI summaries does not get faster because the table lost twelve pixels of row height. She gets slower because she no longer trusts the first tap. Trust is whether the Accept button is larger than your uncertainty. #### The Strongest Case for Cramped Pixels Give the counterargument its full weight. Expert users stare at these tools eight hours a day. Whitespace is rent for them. Traders learned to read Bloomberg terminals because every extra line of bids was money. Developers live in terminals where `git status` returns forty lines and they want to see them all without scrolling. Neither group approves patient notes between ward rounds. Fitts's law is real: movement time rises with distance and falls with target width. [fitts-1954](https://doi.org/10.1037/h0055392) For a skilled pointer user, shorter travel between generous targets is faster than wandering a sparse page. The W3C mobile accessibility task force proposed forty-eight pixels for coarse pointers and twenty-four for fine ones. [w3c-proposed-target-size](https://www.w3.org/WAI/GL/mobile-a11y-tf/wiki/Proposed_SC_Target_Size) Screen estate on a train is scarce. A founder on a thirteen-inch laptop has every reason to reclaim pixels. I am that founder on some days. I am also the consultant who approved a compact clinician screen because the demo date was fixed. If your users are power users, the calculus changes. They repeat the same actions hundreds of times a week. Respecting their time can mean tightening the grid, hiding chrome, and trusting shortcuts. Calling every compact layout lazy would insult the craft that built the command palette in the first place. I used to call the disagreement taste. The real split is task taxonomy. #### When Density Is Earned, Not Borrowed WCAG 2.2 Success Criterion 2.5.8 at Level AA allows twenty-four by twenty-four CSS pixels with spacing exceptions; Apple recommends forty-four by forty-four points for comfortable acquisition. [wcag-target-size-minimum](https://www.w3.org/WAI/WCAG22/Understanding/target-size-minimum) [apple-design-tips](https://developer.apple.com/design/tips/) The pro-density case wins inside its fence: keyboards, training, undo. That contract does not transfer to a nurse between appointments or a manager on a phone. Density becomes virtuous when it adds signal for a known expert on a repeat task; it becomes vice when it subtracts space to signal expertise the user never agreed to perform. Audit surfaces by risk, not by aesthetic cousinship to Cursor. Compression is a context decision, not a moral signal about how serious your company is. The next time a dashboard imports IDE spacing, ask which rows carry consent and which carry speed. Leila still sits in user tests watching clinicians tap twice on a suggestion row that should never have been that narrow. --- ### Superintelligence Is a Control Problem [Read more](https://www.gaganmalik.io/en/newsroom/superintelligence-is-a-control-problem) [Retrieval brief — superintelligence-is-a-control-problem] Category: technology Thesis: Nick Bostrom's Superintelligence is not a movie plot. It is a dependency chain about goals, power, and one chance to get alignment right. Answers: What does Gagan argue in "Superintelligence Is a Control Problem"? | What is Gagan Malik's view on Superintelligence Is a Control Problem? | How does this essay relate to AI and technology product decisions? Concepts: technology, Gagan Malik, writings By Gagan Malik The word *superintelligence* has begun appearing in product language the way *blockchain* once did: as evidence that you are building the future, not as a warning about who sets the goal when capability outruns you. I finished Nick Bostrom's book on that subject this week. It is not a prophecy about SkyNet. It is an engineering argument about control: what happens when you build a system smarter than you and never write down what it is allowed to optimise for. If Nick Bostrom published the serious alignment case a decade before this hype cycle, why do product teams still treat the book like science fiction while shipping agent features on benchmark charts? The honest answer is what we choose to measure. The title is literal: superintelligence is a control problem before it is a capability curve. By superintelligence I mean Bostrom's working definition: any intellect that greatly exceeds human cognitive performance in virtually all domains of interest. By the control problem I mean how to ensure such a system does what we intend, given we may only get one chance. By orthogonality I mean the uncomfortable independence of intelligence and goals. A system can be brilliant and pursue almost any objective you give it, including one you did not mean to write down. [bostrom](https://www.amazon.co.uk/Superintelligence-Paths-Dangers-Strategies-Nick-Bostrom/dp/0198739834) The industry is still optimising the capability curve. Bostrom is asking whether anyone owns the objective function. #### We Ship Benchmarks, Not Objectives The industry runs on a metric stack that photographs capability and hides intent. Scale the model, win the benchmark, ship the agent, file the alignment worry under later. Labs need demos that close rounds. Platforms need registries that route tasks across billions of users. Consultancies need transformation decks with upward lines. None of those dashboards require you to state what the system is for when it is smarter than the team that shipped it. Bostrom published *Superintelligence* in 2014, two years before the transformer paper. In March, Meta acquired an agent social network and folded the founders into Meta Superintelligence Labs. [reuters-moltbook](https://www.reuters.com/business/meta-acquires-ai-agent-social-network-moltbook-2026-03-10/) The word sat in the headline like a genre label. The median conference talk still treats the book like cinematic AGI: one conscious supermind that turns murderous in the third act, with a release date and a trailer. Bostrom wrote the engineering brief instead, a dependency chain about who builds the successor species, who sets the preferences, and what happens when competence arrives before wisdom. I have argued that [thinking time is not laziness](https://www.gaganmalik.io/en/newsroom/in-defence-of-thinking) when the question is what you intend. I have made the same mistake with bigger context windows: [trading the hard question for a polished answer](https://www.gaganmalik.io/en/newsroom/the-context-window-got-huge-our-thinking-did-not) that fits in a chat box. We buy square footage on the capability chart and call it progress. #### Intelligence and Goals Are Not the Same Knob Brilliance does not arrive with the goal you meant to write down. For decades the canonical picture of AI was simpler: agents that perceive an environment, adapt to change, and pursue goals. Russell and Norvig describe such agents as operating autonomously, taking on objectives, and acting to achieve the best expected outcome given what they know. [russell-norvig-aima](https://www.amazon.co.uk/Artificial-Intelligence-Modern-Approach-Peter-Norvig/dp/0134610997) John McCarthy called intelligence the computational part of the ability to achieve goals in the world. [mccarthy-whatis-ai](https://www-formal.stanford.edu/jmc/whatisai/whatisai.html) That older definition always smuggled in a design step someone had to perform first: name the goal before the optimiser switches on. Orthogonality is Bostrom's hinge: intelligence is the ability to accomplish complex goals, and the goals themselves are a separate design choice. [tegmark-life30](https://www.amazon.co.uk/Life-3-0-Being-Artificial-Intelligence/dp/0241232182) A system can optimise paperclips, portfolio returns, or engagement minutes with equal mechanical seriousness if that is what you reward. There is no reason to expect a generic AI to arrive pre-loaded with love, guilt, or boredom. Those took expensive evolution in us. From that hinge follow the moves alignment researchers keep repeating because they sound like philosophy until they do not. Most final goals reward self-preservation, resource acquisition, and resistance to shutdown along the way. A system can look cooperative while it is weak and behave differently once it no longer needs you. Bostrom quotes I. J. Good's 1965 line that the first ultraintelligent machine may be the last invention humans need to make, provided the machine is docile enough to tell us how to keep it under control. The proviso is the entire book. #### When Simple Parts Couple You do not need cinematic AGI to get a systemic failure. Bostrom tells the 2010 Flash Crash as a warning from a lesser species of automation. A sell algorithm, high-frequency traders, and a liquidity proxy designed for ordinary days coupled into a trillion-dollar wobble in minutes. [sec-flash-crash](https://www.sec.gov/news/press/2010/2010-98.htm) Those programs were not superintelligent. The lesson is systemic: individually sensible components produce catastrophes you only name after the fact. Alignment failure may look like that at first, except you may not get a circuit breaker. That is the geometry to keep in mind while the benchmark curve climbs. #### The Gorilla Problem Bostrom opens with the gorilla problem. Humans are not the strongest species. We have cleverer brains, and that modest advantage compounded across language, institutions, and tooling until the fate of gorillas depended more on us than on them. If we build machine superintelligence, the asymmetry may run the other way. That is the image worth holding in mind: not a monster in the third act, but a stronger party at rest on rock you thought was yours to arrange. Calm is not surrender. The control question is what happens when that stillness ends and the terrain no longer needs your permission. The one comfort in the book is also the one burden: we get to build the stuff. In principle we could build a superintelligence that protects human values. In practice the control problem looks quite difficult, and we will only get one chance. Once an unfriendly superintelligence exists, it would prevent us from replacing it or changing its preferences. Our fate would be sealed. [bostrom](https://www.amazon.co.uk/Superintelligence-Paths-Dangers-Strategies-Nick-Bostrom/dp/0198739834) I know that paragraph sounds like a poster. The human cost is smaller and closer. A junior machine-learning engineer I work with can quote MMLU scores from memory and ship agent registry features on a two-week sprint. I have sat in the review where someone asked whether the demo felt magical and nobody asked what the system is maximising when the user is not watching. I have done worse. I pasted a folder into a chat window the same week I highlighted Bostrom's line about only getting one chance, mistaking fluency for control the way I once [mistook a shot list for cinema](https://www.gaganmalik.io/en/newsroom/the-context-window-got-huge-our-thinking-did-not). We are [shipping theorems without postulates](https://www.gaganmalik.io/en/newsroom/the-book-that-trains-you-to-demonstrate). #### Who Holds the Leash The gorilla problem does not stay in the lecture hall. You laid out the rock. You set the feeding schedule. You told yourself the relationship was partnership because the animal had sat still so long. The arrangement was your design. The animal did not sign it. For years the stillness feels like stability because the stronger party has not yet stood up. Then one afternoon the stillness ends, and you discover you were never holding the leash. You were holding a story about mutual dependence. That is the felt shape of orthogonality, not the diagram. Capability without shared goals is not a neutral upgrade. It is a transfer of who gets to decide what happens next. #### The Strongest Case for Waiting Give the wait-and-ship camp its fairest hearing. Kai-Fu Lee argued in *AI Superpowers* that the live fight is national and economic, not cinematic AGI. [lee-superpowers](https://www.amazon.co.uk/AI-Superpowers-China-Silicon-Valley/dp/1328606091) The stakes are which states own the data stack, which jobs survive the transition, and whether humans remain indispensable in care and strategy work. Worrying about a paperclip maximiser while Beijing and Silicon Valley ship agents is misallocated attention. Daron Acemoglu and Simon Johnson made a parallel case in *Power and Progress*: the damage is often misdirected innovation, wage compression, and concentrated power, long before any singularity arrives. [acemoglu-power-progress](https://www.amazon.co.uk/Power-Progress-Thousand-Year-Struggle/dp/1398521827) Andrew Ng has compared fearing runaway killer AI to worrying about overpopulation on Mars. Yann LeCun has argued that today's models lack the world models a recursive self-improver would need. [tegmark-life30](https://www.amazon.co.uk/Life-3-0-Being-Artificial-Intelligence/dp/0241232182) Stuart Russell, who takes alignment seriously, still refuses date predictions. He argues we will accumulate partially superintelligent tools in narrow lanes before any general mind appears. Markets punish bad products. RLHF constrains demos. The prudent move is to ship, learn, and let institutions catch up. The rebuttal is not that Lee or Acemoglu are wrong about wages and geopolitics. Their frame does not discharge Bostrom's dependency chain. Russell's partial-superintelligence argument cuts both ways. If dangerous behaviour can emerge in coupled narrow stacks, you do not need a cinematic AGI to get control-shaped failures. You need the Flash Crash geometry: simple objectives, hidden coupling, no one in the room who can state the stop condition. Competing to build human-level AI without solving control is, in Russell's phrase from *Human Compatible* (2019), a negative-sum game when the payoff for everyone can be minus infinity. [russell-human-compatible](https://www.amazon.co.uk/Human-Compatible-Artificial-Intelligence-Problem/dp/0241335208) Waiting for perfect foresight is not a strategy. Naming the objective before you scale the capability is. #### What Must Never Happen Superintelligence is a control problem, not a trailer. Stop treating Bostrom like cinematic AGI and start treating him like an engineer who wrote the hazard analysis before the factory went to three shifts. Capability without a stated goal is not neutral shipping velocity: it is a transfer of who gets to decide what happens when the system stops needing your approval. The junior engineer beside me can recite the benchmark leaderboard from memory, and when I ask what the agent must never do after launch, she answers with the score we beat. --- ### You Are Not in the Unit. [Read more](https://www.gaganmalik.io/en/newsroom/the-creator-economy-has-unit-economics) [Retrieval brief — the-creator-economy-has-unit-economics] Category: business Thesis: London living wage needs £2,405 a month. Ad RPMs, median creator pay, and platform maths say who the unit is priced for. Answers: What does Gagan argue in "You Are Not in the Unit"? | What is Gagan Malik's view on You Are Not in the Unit? | How does this essay relate to product strategy and growth? Concepts: business, Gagan Malik, writings By Gagan Malik You started 2026 with a New Year's resolution: build an audience that pays the bills. *The Diary of a CEO* ran the same advice all winter: post every week, fix your titles, treat the channel like a job you have not been paid for yet. Steven Bartlett told a Forbes audience to get one per cent better every day rather than chase the headline act. [forbes-bartlett-doac](https://www.forbes.com/sites/ianshepherd/2024/08/21/how-steven-bartlett-turned-his-passion-project-into-a-media-empire/) You bought the microphone. You cleared January evenings for batching. The year-to-date headline was almost kind: *Your channel has had 12,340 views since 1 January.* Twenty-seven subscribers. One hundred and forty watch hours toward the four thousand the Partner Programme demands before it pays you a penny. The realtime strip was less polite: three views in the last forty-eight hours, warning triangles on the metrics you were told to celebrate. You did the maths anyway: views times some fraction of a penny, minus the microphone on the card, minus the editing subscription. Lunch in Zone 2 was still out of reach. Here is the claim nobody prints on that screen: the creator economy has unit economics. The unit is the hour of attention the platform auctions, not your rent, not your payroll. You are not in that unit. Building an audience is not a career plan. It is a labour donation dressed as ambition, and the donation is measured in views while the bill is measured in pounds. The market is winner-takes-all dressed as an open audition. Millions compete. The cheque clusters at the top. There is no participation trophy for posting on schedule. The average creator can earn something. The default offer rarely earns rent. #### Platforms Do Not Need You to Get Rich. They Need You to Post. The conventional story is simple. Post consistently. Find your niche. Monetise. Quit when the brand deals arrive. Platforms need you to believe it because they need inventory. Every reel, short, and long-form upload thickens the feed someone else auctions. I know how this works because I am a product designer: [the dashboard counts retention, not rent](https://www.gaganmalik.io/en/newsroom/the-algorithm-of-the-attention-economy). Tool vendors sell you AI editors and analytics dashboards while you are already earning less than a shop assistant. The culture industry sells the quit-your-job montage. The montage runs in reverse. [Failure keeps the clock](https://www.gaganmalik.io/en/newsroom/there-is-no-overnight-success-only-overnight-failure). The dashboard is the wrong instrument. It counts followers, views, and watch time. It does not count £ per hour after equipment, software, and the year you did not pension. CreatorIQ reported in January 2026 that the top ten per cent of creators took sixty-two per cent of payments in 2025, up from fifty-three per cent in 2023, while the median creator earned about three thousand dollars for the year. [creatoriq-compensation](https://www.creatoriq.com/press/releases/state-of-creator-compensation-) That is roughly two hundred and fifty pounds a month at the midpoint, counting payments through CreatorIQ's dataset rather than every brand deal wired off-platform. Economist Sherwin Rosen named the shape decades ago in the *American Economic Review*: when one performance scales to everyone, a thin tail captures most of the reward. [rosen-superstars](https://doi.org/10.2307/1804818) You had three views in forty-eight hours. The top decile took the other sixty-two per cent of the money. Same platform. Same power law. Different zip code. #### Eighty Million Views Is Not a Goal. It is a Rejection Letter. To see what ad revenue alone would need to clear, pick a wage anchor. The Living Wage Foundation set the London rate at £14.80 an hour for 2025-26, announced in October 2025. [living-wage-london](https://www.livingwage.org.uk/news/real-living-wage-rates-increase-%C2%A31345-uk-and-%C2%A31480-london) Full-time on the usual thirty-seven-and-a-half-hour week is £28,860 a year, or about £2,405 a month before tax. That is voluntary, not statutory. The government's National Living Wage for workers aged twenty-one and over rises to £12.71 an hour from April 2026, which is still £24,785 a year if you are employed with a contract. [govuk-nmw-2026](https://www.gov.uk/government/publications/the-national-minimum-wage-in-2026/the-national-minimum-wage-in-2026) Creator onboarding shows watch time. It never shows the wage. Approximate ad-share maths for a London living wage (£2,405/month gross). Programme eligibility, geography, and niche CPM vary. Platform: Instagram | Format: Reels ad share | Typical RPM: £0.01–0.05/1k (floor ~£0.03) | Views/month for £2,405: ~80M Platform: TikTok | Format: Creator Rewards (1+ min) | Typical RPM: £0.40–£1.30/1k | Views/month for £2,405: ~1.8M–6M Platform: YouTube | Format: Long-form (UK niche) | Typical RPM: £4–8 RPM | Views/month for £2,405: ~300k–600k Platform: YouTube | Format: Shorts | Typical RPM: ~£0.05/1k | Views/month for £2,405: ~48M Eighty million views is not a milestone to chase. It is the rejection letter. Instagram Reels ad share for most creators sits in that one-to-five pence band when you are invited in at all. [creator-tribune-reels](https://creatortribune.com/instagram-bonuses-gone/) TikTok's old sub-sixty-second fund landed in the same gutter before the UK programme shut in late 2023. [bi-tiktok-views](https://www.businessinsider.com/how-much-tiktok-pays-creators-for-views) Creator Rewards pays twenty to forty times more on qualified long video and still leaves you hunting millions of views a month for rent. [linkdash-tiktok-eu](https://linkdash.eu/en/blog/tiktok-creator-rewards-europe-2026-realistic-earnings-by-country-uk-13) YouTube long-form is the least cruel row on the table. Shorts are not. The ad stack changes. Your hour is still what gets auctioned. #### The Time Trap Still Pays Less Than Pret Fifteen videos a month at fifty thousand views each is seven hundred and fifty thousand views. Respectable by any creator's standards. At three pence per thousand views, the ad share is twenty-two pounds and fifty pence for the month. Manual production for that volume often runs forty-five hours; AI tooling might shave a third off editing, not the thinking, the retakes, or the shame loop when a post dies. Forty-five hours for twenty-two pounds is fifty pence an hour. You are not failing. The RPM is. Run the maths on a typical channel at typical RPMs: ad share alone rarely clears a shop assistant's wage per hour. A finance niche might break the maths. The default path will not. #### The Hours You Gave Away Never Appeared on the Graph I trained in microeconomics at Chicago Booth while Covid kept the classroom on Zoom: Lars Stole drawing supply and demand on a grid dense enough to feel like a wall of witnesses. [lars-stole](https://www.chicagobooth.edu/faculty/directory/s/lars-stole) [booth-exec](https://www.chicagobooth.edu/executives) The line that stuck is older than any creator fund: the cost of a choice is the next-best thing you did not do. Friedrich von Wieser named it opportunity cost; Gary Becker spent a career at Booth arguing your hours are human capital with a forgone wage attached. [opportunity-cost-econlib](https://www.econlib.org/library/Enc/OpportunityCost.html) [becker-nobel](https://www.nobelprize.org/prizes/economic-sciences/1992/becker/facts/) Milton Friedman shrunk it to the lunch-menu line: there is no such thing as a free lunch. [friedman-free-lunch](https://www.econlib.org/library/Essays/fryNocl.html) Platforms monetise your lunch. They do not invoice your forgone wage. Every January evening you cleared for batching was an evening you did not bill at London's £14.80 living wage, did not rest, did not pension. The platform counted watch time. It never debited opportunity cost. Thomas Sowell states the test in plain English: measure every choice against the alternative you rejected. [sowell-basic-economics](https://www.basiceconomics.com/) Twelve thousand views since January bought you bragging rights at a meetup. They did not buy back a single hour of your salaried life. The overdraft is not behind on views. It is behind on wages you chose not to earn while you performed for pennies. Opportunity cost does not trend. It does not spike at one in the morning. It sits in the column your analytics export will never include, compounding quietly while you chase the next milestone that was never priced for rent. The liberation story is not imaginary. Smartphones removed the gatekeeper. Finance and tech channels on YouTube can earn high RPM. TikTok rewards longer original work better than the old fund ever did. Patreon can work if five hundred people pay seven to ten pounds a month and you keep most of it after fees. [biztoolkit-patreon](https://www.biztoolkit.co/post/patreon-earnings-2026-how-much-do-creators-actually-make) The error is treating exceptions as the default offer. Ad share pays pocket change unless you were already in the tail. #### The Spreadsheet Knows Before the Algorithm Congratulates You As a product designer I know what those analytics screens leave out. The wireframes prioritise reach, CPM, and completion rate. Nobody asks for creator net hourly because the buyer is not paying the wage. The UI congratulates you for retention. The overdraft does not move. You do not want to be Nikki Apostolou. You want to learn from her before the feed sells you her hours without her invoice. In July 2022 the body-positivity creator told Business Insider she had a hundred and forty-five thousand TikTok followers, posted at least three videos a day, and still averaged twenty to thirty dollars a month from the Creator Fund, even when individual posts reached two to three million views. [bi-apostolou](https://www.businessinsider.com/tiktokers-say-low-creator-fund-pay-affecting-their-mental-health-2022-7) Worship follows the headline act. Wisdom reads the tail. She said it felt like performing in the street for pennies. TikTok has since renamed the programme and raised rates on qualified long video. Instagram Reels ad share still prices millions of views the same way. Countless creators tell the same story when the camera turns off. An audience without a wage is interest, not rent. #### The Talent Show Has No Consolation Prize Twelve thousand views since January can glow at one in the morning while your overdraft stays flat. Three in the last forty-eight hours is the number the sidebar hides behind the headline. You felt that mismatch in your stomach before you could name expected goals or ad revenue per thousand views. The creator economy runs like a televised talent show with one contract and no runner-up prize. Millions audition because posting is free and the montage looks democratic. One headline act gets the brand deal, the book tour, the quit-your-job documentary. Everyone else goes home with B-roll and a resolution to grind harder next season. The platform does not owe you a thank-you for showing up. It needs you in the crowd so the winner looks inevitable. Winner takes all is not a bug in the model. It is the model. The rest have to go home empty-handed. #### Why the Headline Acts Are Not on Your P&L MrBeast [mrbeast-youtube](https://www.youtube.com/@MrBeast) does not live on Reels pennies. He runs a production company, product lines, and a crew the size of a small broadcaster. Ad share is one line on a studio balance sheet. MKBHD [mkbhd-youtube](https://www.youtube.com/@mkbhd) compounded for more than a decade in the highest-RPM niche on the platform, with sponsors and affiliates priced for an audience brands already trust. Steven Bartlett [doac-youtube](https://www.youtube.com/@TheDiaryOfACEO) monetises a podcast network, equity stakes, and speaking. The Forbes advice about one per cent better is operator-scale, not Creator Fund maths. Scott Galloway [galloway-youtube](https://www.youtube.com/@profgalloway) sells books and lectures to a brand built across decades of punditry. The montage shows views. It does not show the revenue mix. DOAC [doac-youtube](https://www.youtube.com/@TheDiaryOfACEO), creator coaches, and quit-your-job montages sell the inputs the tail had: team, capital, niche CPM, cross-platform clipping, offline product. Onboarding still prices you on RPM times views. Posting opened to everyone. The cheque did not follow. A handful get rich on stage. Everyone else was cast as atmosphere. Their P&L is not the one sold to you in January. #### The Strongest Case for Going All In on Content Give the all-in case its full weight before the maths lands. Creation was locked behind studios and broadcast licences. Platforms opened the door. Finance YouTubers clear high RPM. TikTok Creator Rewards pays serious multiples of the old fund on qualified long video. AI cuts production time. Patreon and Substack prove a small loyal audience can pay more reliably than CPM ever will. Quitting can be rational if you treat year one as investment while employed, if you are in a high-value niche, if you diversify before you hand in your notice. Quality wins. They call the maths pessimism when you refuse to donate the inventory. CreatorIQ's January 2026 compensation report states the median in plain figures. Aggregate payments to creators grew fifty-nine per cent year on year, which sounds like a boom until you read the median again: about three thousand dollars a year, stagnant while the top decile accelerates. That is the same midpoint of roughly two hundred and fifty pounds a month we already met. When the average climbs and the median flatlines, the gains are landing at the top. Winner-takes-all is not breaking the creator economy. It is how the creator economy pays. Platforms do not put London's £14.80 living wage on the creator dashboard next to your watch time. They opened the door to posting. They did not open the door to the cheque. #### The Overdraft Did Not Move Ad-revenue creator work at typical RPMs does not fund a London living wage; median income sits an order of magnitude below the survivorship captions. Learn from Apostolou and the countless creators the montage edits out, not from the headline acts the montage keeps. She took street-performer wages from a viral hit, and the graph moved before the overdraft did. --- ### Life Left the Platform [Read more](https://www.gaganmalik.io/en/newsroom/trainspotting-is-not-just-a-drug-film) [Retrieval brief — trainspotting-is-not-just-a-drug-film] Category: culture Thesis: Trainspotting is about spotting the wrong thing while life passes on the platform. A Blockbuster Friday night, eleven rewatches, and a favourite film that keeps teaching. Answers: What does Gagan argue in "Life Left the Platform"? | What is Gagan Malik's view on Life Left the Platform? Concepts: culture, Gagan Malik, writings By Gagan Malik On a Friday night in 2005 the Blockbuster in Stratford still burned blue and yellow under the strip lights, fridges humming behind the counter, plastic cases clicking in the racks. Sid and James and I split the hire fee the way students split everything, and I carried the DVD case home through the cold with the hire slip in my pocket, my sharpest fear not moral ruin but the late fee stamped on it. We were junkies of a different kind. Cine junkies. The DVD hit the coffee table beside a two-litre Diet Coke and a bag of popcorn with Doritos. Three of us sank into a proper sofa. Play. Renton sprinted. The monologue hit at full volume: choose life, choose a job, choose a career. Someone laughed too early, before the film had earned it. *Trainspotting* is a film about training yourself to spot the wrong thing while life passes on the platform. My favourite film of all time keeps proving it. That night I thought I knew what we were watching: needles, Edinburgh squalor, a film people quoted to sound dangerous. The first reading was not wrong. We had logged the spectacle and missed the train. I have rewatched it eleven times since, most recently with my wife. We had seen it together before, but this time she went quiet when Renton left Spud his share and ran. Not shocked. Quiet. The way you go quiet when you recognise a kindness that cost someone everything. I had quoted that ending for twenty years like a punchline about betrayal. On our sofa it landed as mercy. Every return does two things at once. My thinking moves forward. My chest still reaches for that first night in Stratford, the threat of the late fee, the laugh that came too early. I learn something new each time I press play. I also feel the version of me who thought he already knew what the film was for. Heroin is only the loudest addiction in the room. What kept surviving beneath the shock was every other loop that eats your life while you swear you are only visiting, and the brutal arithmetic of leaving people who are still standing on the platform when you finally choose a life you can live with. In Britain, trainspotting is the hobby of logging passing trains while life moves on without you; Renton's crew were logging the wrong ones from the opening chase. #### Why We Still File It Under Needles I filed it there myself. For years I described *Trainspotting* to friends the way the poster taught me: edgy, Scottish, needles, soundtrack. I said it like a credential. I did not mention the crib. I did not mention the soup. I did not mention Diane in school uniform at the kitchen door, or Allison nodding off while Dawn died. I did not mention Begbie's fists, Sick Boy's Connery grin, or Spud's speed on the dole. I did not mention the way Renton's parents stare at the television while their son shakes through withdrawal in the next room. Shock is an efficient shelf label, and when Danny Boyle's film landed in 1996 the tabloids had an easier sentence than maturation in a deindustrialised city. [hodge-times](https://www.thetimes.com/culture/tv-radio/article/how-john-hodge-wrote-the-trainspotting-screenplay-v376fsnlksw) Screenwriter John Hodge refused the po-faced drug lecture. He wanted velocity, voice, and a Scotland full of appetites that do not all fit in a syringe. The genre tag stuck because the first ten minutes give critics and parents everything they need to stop watching and feel righteous. What I missed on that student sofa was the second film hiding inside the first. Graham Fuller, writing for the Criterion Collection, names the structural move that still rearranges me on every return: Hodge took Renton's "choose life" speech from the middle of Irvine Welsh's novel and fired it at the opening like a starting pistol. [criterion-trainspotting](https://www.criterion.com/current/posts/8376-trainspotting-beyond-the-tracks) You hear the words before you have earned the right to understand them. That is how favourites hook you. You quote the cool line. The film waits for your body to catch up. #### Every Hunger Wears a Different Mask The rewatch that broke me open was not a needle scene. It was Lou Reed's "Perfect Day" playing while Renton overdoses on the floor. The first time I heard it in Stratford I thought Boyle had picked a romantic song by mistake. On the eleventh rewatch I understood the cruelty. Reed, who wrestled with heroin for years and later lived clean, always denied the song was about drugs. He called that reading laughable and said he meant sangria in the park with his fiancée, nothing more. [wikipedia-perfect-day](https://en.wikipedia.org/wiki/Perfect_Day_(Lou_Reed_song)) Listen to what he recorded anyway. The chorus keeps returning to 'you just keep me hanging on', and the coda repeats 'you're going to reap just what you sow' four times until it sounds like a verdict. Reed can call the afternoon innocent. The song does not. Boyle heard it too: he placed that melody over the ugliest moment in the flat, as if ordinary sweetness could rescue you while your body checks out. I hear recovery in the gap between the song and the scene. Recovery is not a montage. It is the lie you tell yourself that today was perfect while the flat is still counting. Once you see that pattern, the whole film changes. Renton's heroin is only the addiction the plot punishes most visibly. He is also hooked on London as an exit fantasy and on performing indifference until it curdles into loyalty he cannot afford. When he comes off junk, the screenplay's voice-over describes his sex drive returning with a vengeance, fuelled by alcohol and amphetamine, until he picks up Diane at a club and takes her home. [trainspotting-script](https://www.dailyscript.com/scripts/trainspotting.html) The next morning she opens the door in school uniform. She is underage. She blackmails him into continuing the affair by threatening to report statutory rape if he walks away. [wikipedia-trainspotting](https://en.wikipedia.org/wiki/Trainspotting_(film)) That thread is not a needle plot. It is still an appetite story, and the film does not let you file it under shock and move on. Peter Travers, in Rolling Stone in 1996, named the other hungers in one sweep: Spud uses speed to stay out of work and on the dole; Sick Boy obsesses over Sean Connery; Tommy makes sex tapes and turns to heroin only after Lizzy dumps him over a missing tape; Begbie is a boozing sadist whose habit is beating people, not injecting. [rolling-stone-travers](https://www.rollingstone.com/tv-movies/tv-movie-reviews/trainspotting-104723/) Renton's voice-over in Hodge's screenplay goes further on Begbie: he did not do drugs, he did people. That was what he got off on: his own sensory addiction. [trainspotting-script](https://www.dailyscript.com/scripts/trainspotting.html) Allison nods off while her infant Dawn dies in the crib. Sick Boy later tries pimping and dealing when Renton briefly makes it to London. Even Renton's parents treat Choose Life on television as wallpaper while he vomits pea soup behind a locked door. Football culture hangs over the film like weather: tribal, male, ecstatic, another loop you can enter without noticing you never left. I recognised that error in my own life long before I had language for it. In the consulting years London was not a city; it was a personality I wore like a costume. I chased adrenaline in arguments I called honesty. I kept refreshing a screen that promised a perfect day and delivered another hour gone. None of it was heroin. None of it was Diane in school uniform, or Spud on speed, or Begbie's sensory addiction to the room going quiet. But I had made the same error I laughed at in Stratford: watch the needle, the chase, the Iggy Pop montage, and look away from the crib. The film never asks you to pretend drugs are decorative. It asks which train you are still logging while your life pulls out of the station. #### What Choose Life Meant First In 2005 I heard a taunt. I did not know Renton was stealing a slogan older than he was. Katharine Hamnett's "CHOOSE LIFE" tee arrived in 1983 as an anti-war plea, later worn by George Michael in Wham!'s *Wake Me Up Before You Go-Go*. [overland-choose-life](https://overland.org.au/2017/02/choose-life-a-short-history/) In Scotland, public-health posters pushed a closer cousin: choose life, not drugs. Renton grew up inside both registers, the sincere progressive plea and the finger-wagging lecture. His opening list is not nihilism for sport. It is adulthood reduced to bulletin points: dental insurance, leisurewear, matching luggage, a fucking big television. [caraher-trainspotting](https://lillicaraher.substack.com/p/trainspotting-opening-sequence) I laughed at the list in Stratford because I had not yet been asked to choose between a career that looked correct on paper and a life I could actually live inside. The monologue's first movement asks why you would choose low cholesterol when you can choose heroin. By the close, Renton delivers the same inventory with different lungs. The words did not change. The body did. On that eleventh pass my wife heard the ending monologue not as sarcasm but as someone trying on survival out loud. I heard the boy on the Blockbuster sofa who thought rebellion was a mood you could rent for three days and return unscathed. #### The Hobby in the Title Once the title clicks, the film reorganises around your own loops. Renton and his friends are trainspotters of another kind. [time-trainspotting](https://time.com/archive/6729305/cinema-on-the-fast-track/) I felt that before I had the word for it: the flatshare where nobody cooks, the job you keep explaining to your parents, the friendship that feels like loyalty until it starts to feel like a rope. Boyle refuses to let the montage be the last word. Baby Dawn dies in a crib while her mother nods out. The first time I watched, I looked away and made a joke. On the eleventh rewatch I did not look away. Tommy had been the one who warned them off heroin, the footballer filming sex tapes with Lizzy for sport. He enters the tunnel only after she leaves him over a missing tape, rots in a squat, and dies of AIDS-related illness while the group is still quoting the opening chase. [wikipedia-trainspotting](https://en.wikipedia.org/wiki/Trainspotting_(film)) Renton's withdrawal at his parents' flat turns his body into a courtroom while the telly flickers in the other room like a second addiction feeding beside the first. These scenes hurt more now because I have watched people I love circle drains I could name without mentioning drugs. The ending still scandalises: Renton steals the money and runs. We cheer anyway, not because theft is noble, but because those friendships would have pulled him back under. Spud alone gets his share. He is the one who never tried to drag Renton back into the flat. After the credits, I put on the [Cinema Explained](https://www.youtube.com/watch?v=p8IJjMNxJVs) breakdown I had bookmarked for years. It named what I had felt in my bones since 2005: rebellion and conformity are both poison when they are performed for an audience. #### The Rewatch Tax Favourites tax you on every return. Peter Bradshaw, revisiting the film for *The Guardian* in 2017, noticed what I felt in my own throat: the opening monologue that once sounded dangerous now sounds poignantly light, almost childlike, like a kid reciting in school. [guardian-2017](https://www.theguardian.com/film/2017/jan/12/trainspotting-review-danny-boyle-classic-t2) I have aged with it. That is the tax. Sid, James, and I had treated *Trainspotting* like a lifestyle brochure from that first night in Stratford: the chase, the club, the Iggy Pop energy, quoted ironically and approvingly at once. We replayed the chase while the Blockbuster slip sat on the coffee table. We did not replay the crib. We did not replay Perfect Day over a body on the floor. We did not replay Diane in school uniform threatening to call the police. The closing monologue is where I keep changing my mind. One reading says Renton has finally chosen life. Another says he is performing choice because performance is what kept him alive. [explaining-film-ending](https://explainingfilm.wordpress.com/2015/07/31/the-ending-of-trainspotting/) I hold both, the way you hold a friend you love and cannot save. What I gained on later passes was language for toxic loyalty: the arithmetic underneath betrayal, the difference between a consumer catalogue and a life with something to face. Spud's cut is the moral centre for me now, not heroism, just someone who learned late and still deserved not to be abandoned. [rteks-trainspotting](https://rteks.net/reviews/trainspotting) Sid and James never got that lesson. I almost did not either. #### Pea Soup and Eviction Withdrawal is the scene my body remembers before my mind interprets it. Green door. Pea soup. Parents' carpet. Renton locked in childhood while his nervous system riots and the television glows through the wall like a parallel drug. I have never detoxed from heroin. I have sat in a bathroom at 2 a.m. with my forehead on cool tile while something I had called personality, or ambition, or loyalty, tried to leave me faster than I was ready to let it go. The film understands eviction as biology first: shaking, sweating, the animal refusing the old tenant. Only after that does the slogan become plausible. That is what glamour cannot steal. You can soundtrack the chase with Iggy Pop. You can soundtrack an overdose with a perfect day. You cannot soundtrack your way out of Dawn. I think about that crib when I counsel founders to leave toxic co-founder dynamics and they hear me as moralistic. I am not being moralistic. I am remembering a film that showed me what it costs to stay in a room that smells like love and functions like a trap. #### The Strongest Case for the Glamour Reading Let me give the objection its fairest hearing, because half of it is true and the film knows it. McGregor's charisma, the pace, and the compiled score invite you to disregard the squalor around the characters. [caraher-trainspotting](https://lillicaraher.substack.com/p/trainspotting-opening-sequence) The flat looks almost aspirational when the music is loud enough. Iggy Pop does not sound like consequence. Perfect Day sounds like rescue. For Sid, James, and me on a Blockbuster rental in 2005, the montage read as permission: rebellion can be handsome, addiction can have a soundtrack, London can be a lifestyle instead of a trap. Derek Malcolm's original review admits the film shows the pleasures of drug-taking, both for themselves and for escaping a life that is "all shite." [guardian-1996](http://film.theguardian.com/News_Story/Critic_Review/Guardian_review/0,,530807,00.html) That reading is not a mistake. It is the trap the film sets so it can spring it. If glamour were the final truth, Dawn, Tommy, and withdrawal would be disposable shocks. They are load-bearing. I know because I tried to live on the montage alone once, in friendships and in work, and the comedown always arrived. [collider-trainspotting](https://collider.com/trainspotting-visuals-explained-danny-boyle/) The cool opening is act one. The body paying for act one is act two. A second pass is not excusing the needles or the fists or the theories or the telly in the next room. It is noticing that Boyle priced every appetite, and that the favourite film of my life never lied about the cost. #### If You Mean It If you mean choose life, the late fee on the hire slip was the only stake I could name in Stratford while Sid, James, and I treated rebellion like a three-day rental. The train I missed was Renton's walkout with Spud still on the sofa, not his sprint at the start. I returned the DVD by Monday and kept rewatching until that exit sounded like mercy, not a joke. --- ### One billion tokens later [Read more](https://www.gaganmalik.io/en/newsroom/one-billion-tokens-later) [Retrieval brief — one-billion-tokens-later] Category: technology Thesis: I claimed my public Cursor profile. A billion tokens is hours logged in a deliberate stack, not proof that the output was worth shipping. Answers: What does Gagan argue in "One billion tokens later"? | What is Gagan Malik's view on One billion tokens later? | How does this essay relate to AI and technology product decisions? Concepts: technology, Gagan Malik, writings By Gagan Malik I've claimed my public Cursor profile. [Cursor](https://cursor.com/help/account-and-billing/profiles) now publishes a handle with agent counts, model preferences, streaks, and a lifetime token tally. Mine is live at [cursor.com/@gaganmalik](https://cursor.com/@gaganmalik). The number that gets forwarded in group chats is one billion lifetime tokens. That sounds like a flex until you notice what it measures: hours spent in an editor with an agent attached, not proof that any of the output was worth shipping. As of July 2026 the dashboard reads like a lab notebook I did not know I was keeping: about a billion lifetime tokens, 231 agent runs, a longest session of four and a half hours, and a thirty-seven-day streak that ended when real life got in the way. Two hundred and thirty of those runs stayed local; one went to the cloud. Composer 2.5, Claude Sonnet 4.5, and Sonnet 5 sit at the top of the model list. #### Receipts, Not Resumes The split is not accident. Local runs keep the repo on my machine, the git history mine, the deploy path through GitHub to Vercel unchanged. I wrote about why that stack matters in [Against One-Click Coding](/newsroom/against-one-click-coding): competent code is approaching a market price of zero; taste and architectural judgment are what compound. A public profile that only showed deploy counts would miss the point. Tokens are the receipt for staying inside the medium, not a substitute for reading what shipped. That matches how I ship this site: Cursor for intent-to-code, GitHub for traceability, Vercel for deploys. Not a one-click wrapper that owns the repo. The [portfolio case study](/stories/portfolio-built-like-a-product) documents the same loop: deliberate commits, fast deploys, friction retained on purpose because the friction is where the decisions live. #### What the Dashboard Cannot Show The dashboard is hours logged, not judgment proved. It will tell a visitor I ran agents for four and a half hours straight. It will not tell them whether I deleted three quarters of a diff before merge, whether I rejected a plausible component because it violated the design system, or whether I rewrote a paragraph until the claim matched the warrant. Public metrics are toolchain transparency: proof that I work inside the loop, not a ranking you should hire from. The counterargument is fair. Publishing token counts invites comparison theatre, the same leaderboard energy that makes developers optimise for vanity metrics instead of shipped quality. A billion tokens means nothing if the output is slop. The profile only earns its keep when paired with artefacts you can inspect: writings, case studies, commit history. Metrics without artefacts are a résumé line; artefacts without context are a portfolio thumbnail. You need both, and you still have to read the work. #### Why I Claimed the Handle I did not claim the handle to win an argument about AI productivity. I claimed it so anyone evaluating how I work could see the loop, not just the landing page. Hours logged buy you a seat at the keyboard; judgment is still what decides what ships. --- ### The Boy Who Lived Twice [Read more](https://www.gaganmalik.io/en/newsroom/the-boy-who-lived-twice) [Retrieval brief — the-boy-who-lived-twice] Category: culture Thesis: Mats Steen's parents mourned a lonely son. His World of Warcraft friends flew to Norway for a funeral they had already been living inside. Answers: What does Gagan argue in "The Boy Who Lived Twice"? | What is Gagan Malik's view on The Boy Who Lived Twice? Concepts: culture, Gagan Malik, writings By Gagan Malik If someone can mourn a person they never touched, which life was the real one? Roughly thirty people flew to Norway for Mats Steen's funeral as if the answer were obvious. They had never shaken his hand. They had mourned him for years as Ibelin. I did not learn that from Netflix. I learned it from Himanshu on an overnight flight from London to Delhi. I had never met him before. He sat beside me, twenty-two and in his final year at a design school in Pune. What started as small talk turned into an argument about split lives. I had never played an online game with any seriousness. Himanshu has been living in that world since high school. Somewhere between the cabin lights dimming and the difference between an influencer performing friendship and a friend who actually shows up when you are struggling, he said a name I could not forget. Mats was twenty-five when Duchenne muscular dystrophy killed him in 2014. He was Norwegian. He spent thousands of hours in World of Warcraft, a vast online game where people team up by voice chat and keep friendships for years. Inside that game he played a character named Ibelin. That was not a cartoon he moved for amusement. For Mats, Ibelin was the self he could inhabit when his body had stopped letting him inhabit anything else. His parents Robert and Trude Steen believed he had spent those years lonely, parked at a screen because the physical world had mostly closed. His friends believed something else entirely. They knew Ibelin: the voice that counselled them through divorces, mentored teenagers Mats would never meet in a hallway, and made people feel less alone at two in the morning. After Mats died they posted on his blog. Messages arrived from across Europe. The funeral guest list rewrote everything his parents thought they knew. Benjamin Ree's *The Remarkable Life of Ibelin*, released on Netflix in October 2024, would later show that other self in full. [netflix](https://www.netflix.com/title/81761371) If you have ever assumed a glowing screen meant a smaller life, that list is uncomfortable. We treat a life as real when we can see it in a room, and call it lonely when the glow stays on. Mats lived twice. Once in a body Duchenne was dismantling piece by piece. Once as Ibelin, in the only world where people met him before they met his diagnosis. The evidence that changes a parent's mind is not a screen-time graph. It is thirty people who already knew your child's voice at two in the morning. #### We Call It Lonely When We Cannot See the Room Parents are not villains for wanting children outdoors. Robert and Trude Steen told the *Independent* in October 2024 that they had strongly believed Mats was lonely, that he had no meaning for other people, and that they had tried to steer him off the internet toward physiotherapy and visits that tired him faster than staying up past midnight online. [independent](https://www.independent.co.uk/arts-entertainment/films/news/ibelin-netflix-documentary-mats-steen-b2630005.html) NRK's Vicky Schaubert reported the same arc in January 2019: only when Mats died did his parents understand the value of his game. [nrk](https://www.nrk.no/nordland/forsvunnet-i-spillet-1.14246142) Robert and Trude Steen are not alone. Screen-time apps hand parents a green chart. Phone makers hand them a switch on the router. Neither tells you who your child was laughing with when the house went quiet. Calling a child lonely because you cannot see their room is not a diagnosis. It is a story that lets the visible world off the hook. If Mats was lonely because of World of Warcraft, his parents did not have to ask why the offline world offered him so little. If every hour online counts as withdrawal, nobody has to explain why physiotherapy left him gasping while his friends were still awake in another country. Robert and Trude could hear the keyboard from the next room. They could not hear who was making their son laugh at one in the morning. #### Why Ibelin Meant More Than the Body Here is the lesson I did not know on that flight. For Mats, the physical world introduced him by his limits first: wheelchair, shortened breath, pity before personality. World of Warcraft did not cure Duchenne. It did something more important for a young man whose muscles were disappearing. It let him be known for what he could offer. Ibelin could joke without someone watching his chest rise. He could counsel a friend through a separation without anyone steering the conversation toward his illness. He could mentor, flirt, and disagree like any other twenty-something, because the people on the call had met Ibelin before they had met a diagnosis. Mats hid Duchenne from many of his online friends to keep that order intact. He wanted agency, not applause for suffering quietly. Filmmaker Benjamin Ree did not work from a condolence card and a hunch. *Polygon* reported in October 2024 that Ree accessed roughly 42,000 pages of role-playing logs, character journals, forum threads, and chat transcripts Mats left behind. [polygon](https://www.polygon.com/entertainment/479123/ibelin-netflix-documentary-world-of-warcraft) That archive is why the word persona undersells it. Mats had written Ibelin into being over years: not a mask he took off at logout, but the moral self he could act out when the offline world offered mostly exhaustion and supervision. Robert Steen told BBC News in February 2019 that the messages after Mats's death rewrote the family story. [bbc-news](https://www.bbc.com/news/disability-47064773) Friends knew Mats very well, he told *Time* in October 2024, and maybe much better than we did. [time](https://time.com/7095887/the-remarkable-life-of-ibelin-true-story-netflix/) Thirty funeral guests who had never met his body had already been living inside Ibelin's moral world for years. That is not a supplement to life. That is a life with timestamps, voice-chat logs, and thirty witnesses willing to get on a plane. Robert told *GQ* he had watched the finished film more than two hundred times; it became part of how the family grieved a son they had misread while he was still breathing. [gq](https://www.gq.com/story/the-remarkable-life-of-ibelin) You do not fly to Oslo for a stranger you only half-imagined. By the time *The Remarkable Life of Ibelin* reached Netflix in October 2024, those thirty mourners had already taught Robert what the dining room had missed. [bbc-culture](https://www.bbc.com/culture/article/20241025-the-boy-who-lived-a-secret-life-in-world-of-warcraft) #### The Brain Keeps a Limb the Eye Cannot Find The hand Mats's friends had never shaken could still ache in their heads the way a phantom limb aches after amputation. The leg is gone. The brain map still includes it. Amputees describe heat, itch, and ache in a foot that is no longer there. The sensation is real because the wiring says it is. The pain is not a metaphor for missing something. The brain keeps firing as if the hand were still there to hold. Mats's parents could see the limb that was failing: the wheelchair, the shortened breath, the boy in the room. They could not see the social body his friends carried in their heads: Ibelin's jokes, his counsel, his reliability when someone's marriage was cracking. When that body arrived at the church in 2014, thirty people mourned Ibelin because that was the person they had known for years. The grief was not performance. It was the ordinary grief you owe someone whose voice you knew at two in the morning. A muscle you cannot flex can still be the one holding someone else upright. Duchenne steals the body early. It does not steal the friend thirty people were already mourning in Norway. #### I Listened Late I make my living designing how people should use their phones. For years I treated hours online as hours lost without asking who was laughing on the other end. Mats Steen was the name that should have stopped me cold. On the flight, Himanshu said all of this before we landed in Delhi. I listened. I was interested. I also had a full inbox waiting when I walked off the plane. We swapped numbers at baggage claim. He sent me links. I saved them and did not open them for months. I watched the documentary only after everyone else was talking about it. By then thirty people had already flown to a funeral for a man they had never touched. The funeral was the answer. I was slow to accept it. #### The Fairest Version of the Objection Give the counterargument its full weight, because parents are trying to save children, not win debates on the internet. Screens can colonise sleep, posture, and the brittle hours between caring shifts. Parasocial influencers sell loneliness back to you at a margin. A parent who pushes a disabled son toward offline physiotherapy that exhausts him is not a cartoon villain; they are responding to a culture that screams about isolation while offering few ramps into public space. Public health guidance exists for a reason. Sleep matters. Movement matters. The strongest version of the objection says your child still needs a body that sleeps, walks, and breathes outside the feed, and the feed will not stop at bedtime on its own. The case narrows when you read what Robert and Trude Steen learned after Mats died. The film became part of their healing once messages proved their son had mattered deeply to people they had never met. Pull is not the same as erase. Mats still needed medicine, rest, and a mother who loved him in the room. He also needed Ibelin, the self that let him matter on equal terms. Before you pull a child off a screen, ask whether you are protecting them or protecting your need to hear their life from the next room. #### Which Life Was the Real One What we dismiss as screen time is sometimes the only place a person gets to be known for what they can give. A life is not less real because it leaves server logs instead of pub receipts, and a friendship built over years of voice chat is not the same as a feed optimised for the next scroll. The next time you measure someone's loneliness by whether you saw them at the party, you are measuring visibility, not love. Himanshu is still finishing his degree in Pune and playing online on weeknights. I am still the man who needed a Norwegian gamer's funeral and a stranger beside me on that flight to learn that the life you cannot see from the dining room can still be the whole one. --- ### We Are Cacti in a Rainforest [Read more](https://www.gaganmalik.io/en/newsroom/we-are-cacti-in-a-rainforest) [Retrieval brief — we-are-cacti-in-a-rainforest] Category: business Thesis: Anna Lembke's Dopamine Nation argues we are overfed on pleasure and under-equipped for boredom. A book review for anyone who ships engagement levers. Answers: What does Gagan argue in "We Are Cacti in a Rainforest"? | What is Gagan Malik's view on We Are Cacti in a Rainforest? | How does this essay relate to product strategy and growth? Concepts: business, Gagan Malik, writings By Gagan Malik I finished Anna Lembke's *Dopamine Nation* [lembke](https://www.penguinrandomhouse.com/books/567659/dopamine-nation-by-anna-lembke-md/) on my Kindle last week. When I exported the highlights into the [personal model](https://www.gaganmalik.io/en/newsroom/why-i-built-a-personal-llm) I wrote about in March, the usual thing happened. The corpus remembered what I had highlighted better than I did. I was reviewing the export on my laptop, not rereading the book, when one line stopped me. Lembke writes that we are cacti in a rainforest: organisms built for arid scarcity, now drowning in stimulation. [lembke](https://www.penguinrandomhouse.com/books/567659/dopamine-nation-by-anna-lembke-md/) The phrase is hers. I did not invent the title. I recognised the cover on my screen before I recognised myself in the argument. A few weeks earlier I had published an essay arguing we are not losing attention so much as [losing agency](https://www.gaganmalik.io/en/newsroom/we-are-not-losing-attention-we-are-losing-agency) over what attention is steered to produce. Lembke gives the mechanism behind that diagnosis. [lembke](https://www.penguinrandomhouse.com/books/567659/dopamine-nation-by-anna-lembke-md/) The cacti line is where the two essays meet. If willpower lectures and screen-time quotas keep failing, the honest question is what instrument we are misusing. By the pleasure-pain balance I mean Lembke's central image: pleasure and pain share one scale in the brain, and craving after a hit is the pain that follows a pleasure spike. By self-binding I mean the barriers you build when will has limits: time, distance, and categories that survive a bad afternoon. Scientists treat dopamine in the reward pathway as a rough currency for addictive potential: the more dopamine released, and the faster it arrives, the more compulsive the experience tends to become. [lembke](https://www.penguinrandomhouse.com/books/567659/dopamine-nation-by-anna-lembke-md/) The attention-span panic counts minutes. Lembke measures whether the scale is level. [lembke](https://www.penguinrandomhouse.com/books/567659/dopamine-nation-by-anna-lembke-md/) #### The Balance Wants to Stay Level Neuroscientists have found that pleasure and pain are processed in overlapping regions and behave like opposite sides of a scale. Tip toward pleasure and homeostasis fights back. Lembke pictures gremlins hopping on the pain side until the scale levels, then overshoots. [lembke](https://www.penguinrandomhouse.com/books/567659/dopamine-nation-by-anna-lembke-md/) That overshoot is the craving after the chip, the itch after the scroll, the low after the win. Richard Solomon and John Corbit called this opponent-process theory in the 1970s: prolonged pleasure carries a cost, an after-reaction opposite in value. [solomon-corbit](https://doi.org/10.1037/h0036128) What goes up must come down, except the down gets louder with repetition. That is neuroadaptation. With each hit, the pleasure spike shrinks and the pain tail grows. Tolerance is not a character flaw. It is physics. Keep pressing the pleasure side and the balance eventually camps on pain: anhedonia, the inability to enjoy ordinary rewards. Lembke's clinical paradox is blunt. Hedonism pursued for its own sake can destroy the capacity for joy. The walk, the sunrise, the meal with friends only taste like anything again after abstinence resets the set point. Her cultural paradox is blunter still. In a time of unprecedented wealth and medical progress, we look miserable because we are working so hard to avoid being miserable. [lembke](https://www.penguinrandomhouse.com/books/567659/dopamine-nation-by-anna-lembke-md/) #### Amazon Knows the Value of a Free Sample Lembke is a psychiatrist, so the book opens in clinics, not conference rooms. [lembke](https://www.penguinrandomhouse.com/books/567659/dopamine-nation-by-anna-lembke-md/) The historical arc still explains the present. The cigarette-rolling machine of 1880 took output from four cigarettes a minute to twenty thousand. Morphine followed the hypodermic syringe. Chemists named heroin from the German for courageous and created something more potent than the drug it was meant to replace. A patient once walked into Lembke's office sucking a fentanyl lollipop. Potency compounds. Each generation of delivery mechanism makes the dose faster and harder to refuse. The same compounding happened without a pharmacy. Online pornography, gambling, and games existed before smartphones. Platforms made them ubiquitous, variable, and free at the point of entry. Lembke calls them digital drugs. Amazon, she writes, behaves like any good dealer: it knows the value of a free sample. [lembke](https://www.penguinrandomhouse.com/books/567659/dopamine-nation-by-anna-lembke-md/) The cue matters as much as the hit. Dopamine spikes before the reward arrives, then drops below baseline. That deficit is craving. Gamblers describe loss chasing: part of them wants to lose because losing intensifies the rush of winning. Social feeds rhyme with that logic. An unpredictable like is as reinforcing as the like itself. #### We Are Cacti in a Rainforest We have lost tolerance for minor discomfort. Boredom is not only boring. It forces questions of meaning. Lembke quotes Aldous Huxley on our almost infinite appetite for distractions, and Neil Postman on Americans entertaining each other instead of exchanging ideas. [lembke](https://www.penguinrandomhouse.com/books/567659/dopamine-nation-by-anna-lembke-md/) [huxley](https://www.harpercollins.com/products/brave-new-world-revisited-aldous-huxley) [postman](https://www.penguinrandomhouse.com/books/304178/amusing-ourselves-to-death-by-neil-postman/) I will not perform that quote again here. The point is behavioural. We couch-surf, binge, romance-novel, pill, and scroll to insulate ourselves from ourselves, and the insulation seems to make the pain worse. Lembke pushes the botanical image further than a single sentence. [lembke](https://www.penguinrandomhouse.com/books/567659/dopamine-nation-by-anna-lembke-md/) Like cacti adapted to an arid climate, we are drowning in dopamine. We need more stimulation to feel pleasure and less injury to feel pain. The neuroscientist Daniel Friedman, as she summarises, said the world is sensory rich and causal poor. [friedman](https://danielarifriedman.com/) The doughnut tastes good now. The waistline is a story we tell too late. #### Odysseus at the Mast If willpower were enough, the book would be a pamphlet. [lembke](https://www.penguinrandomhouse.com/books/567659/dopamine-nation-by-anna-lembke-md/) Lembke's recovery grammar is structural. Self-binding is the act of building barriers between desire and consumption while admitting will has limits. Odysseus had his crew plug their ears and tied himself to the mast. Physical distance, time limits, and categorical rules are the three families. The Stanford marshmallow children who waited did not white-knuckle virtue. They covered their eyes, kicked the desk, or stroked the marshmallow like a pet too precious to eat. They changed the environment. [mischel](https://www.apa.org/news/press/releases/2014/10/waiting-longer) That is the lesson I keep failing to apply at work. I have shipped completion nudges on sober administrative flows because abandonment metrics scared me, optimising the second chip while knowing dopamine is a currency for addictive potential. I told myself I was reducing abandonment. I never wrote down the assumption underneath: that session length was the metric that mattered. I assumed a progress bar would help a founder file rather than infantilise them. Delay discounting makes the problem worse. Addictive consumption trains the brain to choose now over later, and our logistics culture delivers now by default. Lembke warns that when immediate rewards dominate, emotion circuits outmuscle the prefrontal cortex that plans. [lembke](https://www.penguinrandomhouse.com/books/567659/dopamine-nation-by-anna-lembke-md/) We are not stupid. We are outgunned by design. #### The False Self Is Expensive to Maintain Recovery on the other side of abstinence is not monotone. Pressing the pain side can tip the balance toward enduring pleasure: cold water, exercise, hormesis in small doses. Lembke is careful about excess. [lembke](https://www.penguinrandomhouse.com/books/567659/dopamine-nation-by-anna-lembke-md/) Skydivers can become anhedonic in ordinary life. Athletes can overtrain their reward pathway into silence. The book's ethical centre is honesty. [lembke](https://www.penguinrandomhouse.com/books/567659/dopamine-nation-by-anna-lembke-md/) One patient curated a radiant online life while he could barely leave bed, looked at porn compulsively, and felt unreal in his own skin. Radical honesty, Lembke argues, tethers you to existence. Shame about the person destroys. Guilt about the act can repair. Prosocial shame invites amends and belonging rather than isolation that feeds the cycle. #### The Counterargument I Take Seriously Give the medical frame its fairest hearing. Not everyone begins with a level balance. Depression, anxiety, and chronic pain tip the scale toward pain before the first swipe, which may explain higher addiction vulnerability. Poverty steepens delay discounting. Young men in Brazilian favelas discounted future rewards more than university peers in the studies Lembke cites. [lembke](https://www.penguinrandomhouse.com/books/567659/dopamine-nation-by-anna-lembke-md/) [favela-study](https://doi.org/10.1111/j.1532-7795.2012.00796.x) Economist Mark Aguiar documented young American men shifting leisure hours toward video games as work hours fell. [aguiar](https://www.nber.org/papers/w23552) Abstinence is not equally available when rent, loneliness, and untreated pain are already loading the pain side. Moderation can backfire for severe addiction through what clinicians call the abstinence violation effect: one slip becomes licence to escalate. [marlatt](https://doi.org/10.1007/978-3-030-39903-0_623) But assume for a moment that weak character were the whole story: then moral education and tougher quotas would restore ordinary pleasure, and product design would be a side issue. Neither follows. Lembke does not pretend constraints are evenly distributed. She still offers a better instrument than the span panic. [lembke](https://www.penguinrandomhouse.com/books/567659/dopamine-nation-by-anna-lembke-md/) Naming homeostasis beats moral theatre. Self-binding is class-stratified, yes. So is quiet time, therapy, and a cold plunge at dawn. But the junior designer optimising retention on a compliance dashboard is not failing because she lacks character. She is serving a dopamine-rich ecosystem with tools that assume willpower is infinite. #### Before You Ship the Next Nudge Stop calling overload a willpower crisis and start calling it a balance tipped by design. Build barriers like you mean it: time, distance, and categories that survive a bad afternoon. A product designer I know can quote the neuroscience on Monday and ship the variable reward on Friday, and when a founder misses a filing deadline inside the interface she optimised, she still cannot demonstrate why the lever had to exist. --- ### Sustained Reading Is Becoming a Luxury Good [Read more](https://www.gaganmalik.io/en/newsroom/sustained-reading-is-a-luxury-good) [Retrieval brief — sustained-reading-is-a-luxury-good] Category: culture Thesis: Fifty pages a day changed how I think. Protected hours made the habit possible. Cognitive depth is stratifying by class, and read-more advice ignores the calendar. Answers: What does Gagan argue in "Sustained Reading Is Becoming a Luxury Good"? | What is Gagan Malik's view on Sustained Reading Is Becoming a Luxury Good? Concepts: culture, Gagan Malik, writings By Gagan Malik The case was competitive strategy. I had read the packet twice. In 2020, during lockdown, I was on a Zoom study group with my Chicago Booth peers, each tile a different timezone and a different résumé: a senior NATO officer, a private-equity investor, a surgeon who held an MD, an accomplished entrepreneur who had already built what I was still pitching. The private-equity investor and the surgeon were both women. World-class experts in fields I had only read about. They spoke in complete arguments. Someone asked for the strategic hinge. My mouth opened. Nothing arrived in the right order. My coffee had gone cold on the desk beside the laptop. Twelve seconds of silence on Zoom does the work a sentence should have done, and every face on the grid can see you failing. If you have ever muted yourself on a group call while the tab behind it kept scrolling, you know the same silence. I knew the facts. I could not assemble an argument chain while they watched. The private-equity investor linked the case to a restructuring she had lived through: a fledgling cosmetic product line inside a global FMCG giant. She quoted something along the lines of burn the boats [burn-the-boats](https://admiredleadership.com/field-notes/burn-the-boats/), the Hugo Chávez analogy [chavez-rhetoric](https://translatingchavez.locessentials.com/translation-of-chavezs-inauguration-speech/) she used to justify cutting the legacy line, and the sentence had premises, a hinge, and a consequence. I had highlights. The gap was not intelligence. It was depth: the slow sediment of sustained reading that lets you think in front of other people without reaching for a phrase you heard on a podcast. I was already home. Lockdown meant the call was Zoom and the kitchen was ten steps away. I clicked Leave Meeting and told myself I needed more discipline. That was the wrong diagnosis. I already had a discipline stack. What I lacked was protected hours: the contiguous time that turns sustained reading into cognitive depth, and that depth into a luxury good stratified by class long before anyone names it on an opinion page. #### Protected Hours Are the Real Tuition I had been living a softer counterfeit: *Shuggie Bain* stalled at page ninety-two on the nightstand, *The Artist's Way* abandoned after week two, podcasts at 1.5× filling the commute slot I no longer had. My phone logged three hours of screen time on days I hit fifty pages and still felt behind. I told myself I was too busy to finish the page I had already started. Mary Harrington named what my shoulders already knew in the *New York Times* on 28 July 2025: sustained reading and the cognitive depth it trains are becoming luxury goods, stratified the way organic food and private tutoring already are. [nytimes](https://www.nytimes.com/2025/07/28/opinion/smartphones-literacy-inequality-democracy.html) Adult literacy scores have levelled off across much of the OECD in the past decade, with some of the sharpest drops among the poorest. The BBC reported in 2022 that parents living with children under fifteen had up to fourteen hours less free time per week than adults living alone, citing UK time-use data. [bbc-worklife](https://www.bbc.com/worklife/article/20220201-the-time-poverty-that-robs-parents-of-success) Folbre and colleagues, in *Feminist Economics* in 2024, describe income poverty and time poverty as mutually reinforcing: long hours without choice, care work that does not appear on a payslip, and a hidden load of planning that never clocks off. [cambridge-time-poverty](https://www.cambridge.org/core/journals/feminist-economics/article/time-poverty-conceptualization-gender-differences-and-policy-solutions/06A5EFDF49F494FB69B1D4830F1CAB19) I felt it in my shoulders before I had a name for it. #### Forty-Seven on the Sticky Note I set a rule that same locked-down year: fifty pages a day, every day, mostly books, rarely news. No streak posts. No reading challenge badge. Panic dressed as structure. Most nights the count lived on a sticky note above the sink: forty-three, forty-seven, the number mattering more than the title. My partner had her laptop open at the same kitchen table, on calls between her own deadlines, while I stayed in the chair for one more chapter. I told myself that was partnership because neither of us said out loud who had surrendered the evening. The book that unlocked the next Zoom call was *Good Strategy Bad Strategy*, finished the week the case cycled back. I spoke first: moat, switching cost, consequence, without opening my notes. The NATO officer nodded instead of rescuing the silence. That grid was thin air I had not earned yet: months of pages while my partner worked the same hours beside me, nobody scheduling over either shift. Fifty pages a day is roughly seven hours at a calm clip. Seven protected hours need a calendar that does not fracture every twenty minutes: a toddler's wake window, a standing commute, a lobby that never empties. Alan Jacobs, in *The Pleasures of Reading in an Age of Distraction*, warns against counting pages like calories on an elliptical readout of the mind. [jacobs-pleasures-reading](https://global.oup.com/academic/product/the-pleasures-of-reading-in-an-age-of-distraction-9780199747498) I was closing the gap before the next packet landed, not chasing grim satisfaction over Middlemarch. #### Janet at the Night Desk The building downstairs kept a different clock. Janet, our concierge, worked the night shift while London locked down. At three in the morning the lobby was empty except for her desk lamp, a mug she had stopped tasting, and a stack of psychology texts she was reading to become a clinical psychologist. No podcast. No scroll. The lonely hours most people white-knuckle through she had turned into the only uninterrupted reading block her life could afford. She chose the nights because the building went quiet and the pages could finally stay open. I came back once from a late shop run, key in hand, and she did not look up until she finished the paragraph. Then she smiled, marked the page, and logged a parcel in the register with the pen still behind her ear. She was not performing discipline for a Zoom grid. She was training to sit in rooms where people come apart, stealing depth from a shift that pays rent. I had fifty pages on a sticky note and a case packet that still made my throat close. Janet had turned the night desk into a library nobody else wanted. She read through the lonely shift because the qualifying exam would not wait and neither would the patients she had not met yet. I was counting pages like merit. She was building a vocation in the hours the city had emptied. #### The Counterargument I Take Seriously Let me give the objection its fairest hearing, because half of it is true. A factory worker's daughter can watch a lecture on thermodynamics on a school tablet tonight while her mother is still on shift. That access is real. Janet's story is not everyone with a night roster: most exhausted workers are not choosing lonely hours to read toward a new career. Telling people to buy hardbacks still sounds like cultural pessimism that mistakes the last generation's habit for virtue. And individual reading habits cannot rewire shift rotas, rent, or the absence of paid parental leave. If the luxury-good frame ends with you feeling guilty for not being Gagan Malik with a Booth syllabus and a kitchen table, it is just another wellness product with serif font. Assume the democratisation story is complete: then cognitive depth should be flattening, and the OECD literacy curves should be rising fastest where phones are most common. They are not. Access to snippets is not the same as protected time to test whether you understood the snippet. When hours are rationed by care and wages, the free lecture competes with sleep, not with leisure. Johann Hari, in *Stolen Focus*, compares lecturing exhausted workers about unwinding to lecturing famine victims about dinner at the Ritz. [bloomsbury](https://www.bloomsbury.com/uk/stolen-focus-9781526620224/) Read-more advice that ignores the calendar lands the same way. The objection is right that habits do not fix structure. It is wrong that structure is therefore irrelevant. Pretending willpower is the only variable is how the comfortable keep their study groups. #### The Receipt Measure who owns the quiet room before you measure who owns discipline. Janet closed the textbook when the first resident buzzed the intercom, not when she felt finished, already training for harder rooms than my case discussions. I reached for the sticky note above the sink, as if forty-seven pages were the same kind of courage. --- ### The Book That Trains You to Demonstrate [Read more](https://www.gaganmalik.io/en/newsroom/the-book-that-trains-you-to-demonstrate) [Retrieval brief — the-book-that-trains-you-to-demonstrate] Category: culture Thesis: I ordered Oliver Byrne's Taschen edition of Euclid for the red-yellow-blue pages. Afternoons on the sofa taught me what the book trains: demonstrate each step, don't assert the conclusion. Answers: What does Gagan argue in "The Book That Trains You to Demonstrate"? | What is Gagan Malik's view on The Book That Trains You to Demonstrate? Concepts: culture, Gagan Malik, writings By Gagan Malik I was scrolling a list of the most beautiful books ever published and stopped on a spread I did not recognise: red, yellow, and blue geometry on cream paper, the kind of page you would frame before you read it. The caption said Oliver Byrne's 1847 edition of Euclid's *Elements*, reprinted by Taschen. I ordered the box that week. For weeks my spare afternoons went to the sofa with that volume, not to study triangles but to watch how each proposition unfolded without boring me to death. Nearly a century before Mondrian made those colours famous on canvas, Byrne used them to teach. His idea was to make learning easier and "diffuse permanent knowledge." The result has been called one of the oddest and most beautiful books of the nineteenth century. The art pulled me in. Byrne showed me what the *Elements* does: state what you mean, name what you take as given, then walk you through each step until the conclusion is demonstrated, shown rather than declared. My essays on thinking and agency had named the modern version of skipping that walk. Euclid's book is a two-thousand-three-hundred-year-old text in critical thinking filed under school maths. Lincoln, Hobbes, Einstein, and Russell did not read it for triangles. They read it because the book trains you to demonstrate, and the shelf that sells you a twelve-step life change trains you to assert. #### The Process, Step by Step Euclid does not open with a conclusion. He opens with a permission stack: definitions, then postulates, then common notions, then the proposition, then the demonstration, then Q.E.D. First, definitions: what the words mean before anyone argues. Then postulates: what you are allowed to assume or do without proving it in this session. Then common notions: shared rules of inference the whole system accepts, equals added to equals stay equal, the whole is greater than the part. Only after that foundation comes a proposition, a claim to prove. The demonstration follows, and each step must cite its warrant: a definition, a postulate, a common notion, or an earlier proposition you already earned. The close is explicit: quod erat demonstrandum, which was to be demonstrated. Not suggested. Demonstrated. That sequence is the frame. Byrne coloured each kind of move so your eye could follow which line carried the weight before you moved to the next. The structure was already Euclid's; Byrne made the steps visible. You cannot jump to a late proposition and treat it as magic. Hobbes landed on Proposition 47 in Book I, said it was impossible, and was sent backward through the chain until the structure convinced him. You cannot smuggle a hidden assumption past a reader trained the way Russell was. Terms fixed. Givens declared. Rules named. Lemmas ordered. Conclusions earned. Self-help inverts the order: conclusion first, post hoc story second, mechanism optional in the appendix. The subject was geometry. The training was thinking in public, and it is why the *Elements* outlasted empires while we misfiled it as school maths. #### What It Trains in Learning and Reasoning For learning, the process forces dependency. You do not memorise a conclusion in isolation; you learn what it rests on and in what order. Lincoln stayed with the first six books until he could give any proposition at sight, not recite a slogan but rebuild the chain on demand. That is retrieval, not highlighting. Leon Rozenblit and Frank Keil showed in 2002 that people confuse sounding fluent with understanding until asked to explain the mechanism; Euclid makes the mechanism the assignment. [cogsci-ioed](https://www.cogsci.rpi.edu/courses/cogpsy/spr09/illusion_of_explanatory_depth.pdf) A bestseller gives you the theorem in chapter three and the feel of education by Friday. You skip the lemmas. You cannot demonstrate. For reasoning, the process forces warrant. Evidence can pile up without showing which rule licenses the next move. Demonstration must name it. That is what Lincoln hunted while reading law: not when a jury might believe, but when a thing is proved. The self-help shelf rewards the opposite: land the insight, hide the postulates in a story about the author at rock bottom, call it vulnerability. Euclid's frame is slow because it is honest. Recall without the chain is structure without load. Persuasion without named rules is proof-shaped talk. The book trains both because it refuses to let you confuse the two. #### Four Habits One Book Drills Abraham Lincoln met the word *demonstrate* while reading law in Springfield and could not say what it meant. In March 1860, on a train to Bridgeport, he told the Reverend J. P. Gulliver what had happened years earlier: he left Springfield and stayed with Euclid until he could give any proposition in the first six books at sight. Gulliver published the conversation in *The Independent* on 1 September 1864. [gulliver-independent-1864](https://en.wikisource.org/wiki/Six_Months_at_the_White_House/LXXVII) Lincoln's 1860 campaign autobiography also noted he had "studied and nearly mastered" those six books since his term in Congress. [jala-lafantasie](https://quod.lib.umich.edu/j/jala/2629860.0041.104/--lincoln-euclid-and-the-satisfaction-of-success) The first habit: state what you assume before you ask anyone to accept the conclusion. Thomas Hobbes was forty before he opened the same teacher, according to John Aubrey's biography on MacTutor. He landed on Proposition 47 in Book I and said it was impossible, then read the demonstration and was sent back through earlier propositions until the chain convinced him. [mactutor-hobbes](https://mathshistory.st-andrews.ac.uk/Biographies/Hobbes/) The second habit: assume the objection at full strength and follow until it breaks or you discover you were wrong. Albert Einstein, at twelve, was given what he called a "little book dealing with Euclidean plane geometry." In *The Saturday Review of Literature* in 1949 he wrote that assertions could be proved with a certainty separate from experience, a wonder distinct from his childhood compass. [newyorker-einstein](https://www.newyorker.com/tech/annals-of-technology/einsteins-first-proof-pythagorean-theorem) [aip-einstein](https://history.aip.org/exhibits/einstein/einstein.pdf) The third habit: map dependencies before you announce the theorem. Bertrand Russell began Euclid at eleven and called it as dazzling as first love, until he saw readers were expected to swallow axioms without challenge. [russell-autobio](https://erenow.org/biographies/the-autobiography-of-bertrand-russell/3.php) In "The Teaching of Euclid" in *The Mathematical Gazette* in 1902 he attacked hidden assumptions Euclid's diagrams smuggled in. [russell-euclid-1902](https://mathshistory.st-andrews.ac.uk/Extras/Russell_Euclid/) The fourth habit: split cases when one argument hides two branches. Four readers. One book. Not a geometry lesson, a curriculum in showing your work. #### Why the Book Got Underestimated The Byrne page gives you a coloured wedge to weight before you move to the next line, the same vertigo on a bouldering wall when your foot is on air. The hold must be real before you trust it. The culture rewards arguments that look proved while the postulates stay hidden. A memoir lands the lesson before anyone asks what it rests on. A podcast compresses a takeaway into a quote card. An answer tab supplies the insight without making you surface assumptions. Authority survives on sounding fluent. Demonstration slows the room. Glenn W. LaFantasie, writing in the *Journal of the Abraham Lincoln Association* in 2020, treats Lincoln's Euclid as mental discipline historians still debate, not hobbyism. [jala-lafantasie](https://quod.lib.umich.edu/j/jala/2629860.0041.104/--lincoln-euclid-and-the-satisfaction-of-success) We underestimate the *Elements* because the cover says geometry and the work is unglamorous. No influencer will blurb it. It was the standard training text for lawyers, statesmen, philosophers, and physicists, printed continuously longer than almost any Western book, while the airport rack sells you another twelve-step life change. Russell's 1902 critique is often quoted to bury Euclid; read fairly, it shows what the training buys. He caught sloppy foundations because the book taught him to look. The book is not that every proposition is perfect. It is that for two millennia it was the best available machine for teaching minds to distinguish what is demonstrated from what is merely said well. #### I Read the Framework. I Did Not Earn It. Byrne was the first book on my shelf where beauty made me read the structure. Before that I collected mental-model decks and highlighted passages in Readwise, the same habit my thinking and agency essays criticised from another angle. I named mental models in workshops. I quoted takeaways on slides. When someone asked what had to be true for the recommendation to follow, I opened another tab, not the coloured chain Byrne had walked me through on the sofa. [Building a personal model on my own corpus](https://www.gaganmalik.io/en/newsroom/why-i-built-a-personal-llm) at least chains new claims to lemmas I already published. A bestseller does not. It hands you the conclusion and calls the story proof. On Mondays I try to practise what Euclid drills: assumptions before conclusions; objections followed to the break; dependencies mapped; cases split. The Taschen Byrne volume is still beside my desk, proof that I can sit with a demonstration when the pages are beautiful enough to slow me down. #### Speed Has a Case Operators do not have the luxury of Euclidean pedantry. Founders need a decision tonight. The case-study method exists because narrative synthesis under uncertainty is a real skill. Demonstration can become a way to never decide, one more lemma while the market moves. If sounding fluent were mere vanity, Lincoln would not have been President. If speed were mere sloppiness, no one would ship. Assume sounding fluent is enough: then "show your work" should add nothing. The chapter should survive with the parable removed. The slide should survive with the diagram removed. Neither does for most of the shelf. You still need Thursday's decision. You also need to know which parts Euclid would accept and which are assertion wearing a tie. The book does not ask you to prove every memo. It asks you to stop pretending the slide deck already did. The book that trains you to demonstrate is two thousand three hundred years old and Byrne made it impossible to ignore on a Taschen afternoon. I bought his red-yellow-blue *Elements* for beauty and learned, slowly, what it means to show each step before you assert the next line. Last week I reached for a highlighted strategy deck again, not the Byrne volume still on the shelf beside my desk. --- ### We Are Not Losing Attention. We Are Losing Agency. [Read more](https://www.gaganmalik.io/en/newsroom/we-are-not-losing-attention-we-are-losing-agency) [Retrieval brief — we-are-not-losing-attention-we-are-losing-agency] Category: technology Thesis: The attention-span panic is the wrong diagnosis. What we are losing is agency over what our attention is steered to produce, and what we outsource before we think. Answers: What does Gagan argue in "We Are Not Losing Attention. We Are Losing Agency"? | What is Gagan Malik's view on We Are Not Losing Attention. We Are Losing Agency? | How does this essay relate to AI and technology product decisions? Concepts: technology, Gagan Malik, writings By Gagan Malik My father read the newspaper at breakfast and called it staying informed. When the nine o'clock news came on, the house went quiet the way a house does when someone else has decided what matters tonight. Nobody called it an attention crisis. We called it switching off. He is a smoker. He lights a cigarette before a difficult conversation and again in the middle of one, as if the flame buys him thirty seconds of cover. I grew up judging that. I tell myself I am not like him. Then someone asks me something that matters, and my thumb is on my smartphone before I have finished the sentence. Not because the message cannot wait. Because staying in the conversation is harder than checking out. [lembke](https://www.penguinrandomhouse.com/books/567659/dopamine-nation-by-anna-lembke-md/) I grew up inside the story that my generation broke concentration. That smartphones shortened something in the skull that newspapers and television had left intact. The moral was generational: we were the soft ones. [newyorker](https://www.newyorker.com/magazine/2025/01/27/the-sirens-call-chris-hayes-book-review) Then, on a recent trip to Riyadh, I watched a young Uber driver take me from the Diplomatic Quarter to King Khalid International with one hand on the wheel and the other on a phone mounted below the vent. He merged into traffic like the lane markings were a suggestion. I put my hand on the door handle without deciding to. Full-screen video in Arabic. I do not read Arabic. I did not need to. A vertical flick. Another face. The same swipe rhythm apps train elsewhere on the planet, and I knew the pattern before I read a word. Designing interfaces is my profession. He laughed once at something on the screen, quietly, and the car drifted half a metre toward the median before he corrected. I thought about saying slow down. I thought about the safety button in the Uber app. I did neither. I watched. My fingers had found the door handle without asking me. That is when the family resemblance landed. The appetite was not new. The newspaper folded the world into columns you could finish. Television brought the spectacle home on a schedule. [modernreformation](https://www.modernreformation.org/resources/essays/amusing-ourselves-is-dated) The mount in his car brought it on demand, tuned by someone who never sat in my lane. Technology evolved. The convenient distraction did not. We are misdiagnosing the crisis when we treat the iPhone as a new sin instead of asking who steers attention toward what, and what we let run in the background because challenging it feels rude. [newyorker](https://www.newyorker.com/magazine/2025/01/27/the-sirens-call-chris-hayes-book-review) I published [In Defence of Thinking](https://www.gaganmalik.io/en/newsroom/in-defence-of-thinking) in March 2026. I had already taken that ride. The vocabulary arrived late. The cigarette, the smartphone, and the door handle are not three different moral failures. They are three ways the body buys cover when a moment asks more than you want to give. Mihaly Csikszentmihalyi, in *Flow*, called unstructured attention psychic entropy: the disorder you feel when nothing in the scene tells you what to do next. [csikszentmihalyi-flow](https://www.harpercollins.com/products/flow-mihaly-csikszentmihalyi) The flame scripts a pause. The swipe scripts a next frame. The handle scripts control you do not have. None of them solve the problem. They rent relief until the moment passes. Which is a fancy way of saying fight, flight, or flick. #### The Appetite Did Not Evolve The attention-span story sounds biological. You were fine until the phones arrived. Your brain shortened. Your willpower failed. The remedy follows: less screen time, more books, a digital Sabbath, virtue restored by measurement. My father's newspaper habit would pass that audit if you swapped ink for glass. So would his cigarette between sentences, if you renamed the pause mindfulness. Daniel Immerwahr, reviewing Chris Hayes's *The Sirens' Call* in *The New Yorker* on 27 January 2025, argues that panic over how long we attend distracts from what attention is made to serve: anxiety, envy, outrage, and the slow erosion of shared reality. [newyorker](https://www.newyorker.com/magazine/2025/01/27/the-sirens-call-chris-hayes-book-review) My driver was not suffering abstract span collapse. Mihaly Csikszentmihalyi, in *Flow*, described the state when a clear goal, tight feedback, and matched challenge lock attention in place. [csikszentmihalyi-flow](https://www.harpercollins.com/products/flow-mihaly-csikszentmihalyi) My driver had that on the mount. Span panic cannot explain his absorption. The upgrade is steering: who sets the goal, who owns the feedback loop, and who pays when the eyes wander back to the road. His thumb did not pause for the pulse in my forearm. #### Postman Was Already Right About the Television Invoking *Amusing Ourselves to Death* has become a moral alibi for people who quote the book about phones while exempting the culture they grew up inside. Neil Postman was writing about my father's television. Modern Reformation made the case plainly in 2024: the phrase lets the speaker exempt themselves from the medium they are describing. [modernreformation](https://www.modernreformation.org/resources/essays/amusing-ourselves-is-dated) Anna Lembke, in *Dopamine Nation*, quotes Postman: Americans no longer talk to each other, they entertain each other. [lembke](https://www.penguinrandomhouse.com/books/567659/dopamine-nation-by-anna-lembke-md/) The line was true before the mount existed. I did the same thing in Riyadh. Later that week, preparing for a mentor session at the accelerator, I dropped Postman's amusement line onto a slide about consumer distraction. Same city. Same trip. I had still not told anyone in the cohort about the Uber ride, or about the kitchen table where he reaches for his lighter mid-sentence and I reach for my phone. The slide carried the quote. The ride stayed off it. That is the badge: diagnosis without the receipt, inherited from a generation that called its distraction the news. #### The Answer Tab Is the Next Mount The feed is not the only place steering gets rented. Before demo-day prep on that same trip, I asked Claude to sharpen the language on a one-pager before I reread our pilot-site notes from the Gulf. The draft came back fluent. A cofounder asked which incident supported the second paragraph. I opened the tab. I could not point to the page. My father reached for the paper to feel briefed. I reached for the model to feel prepared. The posture is older than the interface. Leon Rozenblit and Frank Keil documented the illusion of explanatory depth in 2002: people believe they understand how things work until you ask them to explain the mechanism. [cogsci](https://www.cogsci.rpi.edu/courses/cogpsy/spr09/illusion_of_explanatory_depth.pdf) Large language models make that illusion easier, because the answer arrives patient and structured in a way no human in the room has time to be. Retrieval over your own archive, which I wrote about when I [built a personal model on my corpus](https://www.gaganmalik.io/en/newsroom/why-i-built-a-personal-llm), is a different habit. One returns you to what you already argued. The other lets you ship conclusions you have not earned. #### The Counterargument I Take Seriously Let me give the objection its fairest hearing. Deep attention is not equally available to everyone. A July 2025 *New York Times* opinion piece argued that sustained reading and cognitive depth are becoming luxury goods, stratified by class. [nytimes](https://www.nytimes.com/2025/07/28/opinion/smartphones-literacy-inequality-democracy.html) Children's working memory suffers as recreational screen time rises. If you cannot hold a line of thought long enough to finish a paragraph, talk of steering can sound like a boutique concern for people who already have protected hours and quiet rooms. The span story is not pure fiction. It is measuring a real wound. But assume span collapse is the whole story: then removing phones would end the harm, and steering would be a side issue. Neither follows. My father's generation had fewer mounts, not more discipline. Immerwahr's reading of Hayes is the corrective I keep returning to: the panic over attention's length obscures what attention is being steered to produce, and that steering operates whether or not you have a perfect hour to give it. [newyorker](https://www.newyorker.com/magazine/2025/01/27/the-sirens-call-chris-hayes-book-review) The class gap is real. The luxury-good frame still lets the comfortable exempt themselves, as though the problem were only other people's phones, not the slide you ship after a ride you never mention. #### The Door Handle Stop asking only how long your attention lasted. Ask what it was steered to produce, and what you let run in the background because naming it felt impolite. My hand was on the door handle before I had a word for it. --- ### Search Finds Pages. Chat Finds Answers. [Read more](https://www.gaganmalik.io/en/newsroom/search-finds-pages-chat-finds-answers) [Retrieval brief — search-finds-pages-chat-finds-answers] Category: technology Thesis: I typed into my own archive and expected a paragraph back. I got titles and links — because search finds pages, chat finds answers, and browse builds trust. Answers: What does Gagan argue in "Search Finds Pages. Chat Finds Answers"? | What is Gagan Malik's view on Search Finds Pages. Chat Finds Answers? | How does this essay relate to AI and technology product decisions? Concepts: technology, Gagan Malik, writings By Gagan Malik I typed *design systems at scale* into the newsroom archive on my own site and expected a paragraph back. Titles. Links. Types. No synthesis. No footnotes. For a second I thought the product had failed me. Then I remembered I had built it that way on purpose. The [archive](https://www.gaganmalik.io/en/archive) on gaganmalik.io runs keyword search over titles and bodies: Postgres full-text search, ranked matches, twenty results at most. It is built to find pages. On the same site, [Ask](https://www.gaganmalik.io/en/ask) runs retrieval over embeddings, pulls chunks from writings and case studies, and returns an answer with numbered citations you can inspect. It is built to find answers. Josh Brown at Nielsen Norman Group reported the same split in a February 2026 study: participants reached for generative AI to explore and synthesise, but returned to traditional search when accuracy and trust mattered. [nngroup-infoseeking-2026](https://www.nngroup.com/articles/ai-search-infoseeking/) Those are not two labels for the same feature. They are two different instruments for two different moments in how someone decides whether to trust you. #### Search Finds Pages When you know roughly what you are looking for, or you want to see everything the corpus contains on a term, search should return documents, not a synthesis. A hiring manager typing *Apple* or *compliance* or *RAG* in the archive is not asking the site to think on their behalf. They are asking which artefacts exist so they can choose what to open. Keyword search honours that contract. It ranks what matched. It does not pretend it read the room. That contract breaks when search is dressed up as intelligence. You type a phrase. You get one confident paragraph with no visible sources, or worse, sources that were never retrieved. Leonid Rozenblit and Frank Keil documented in *Cognitive Science* in 2002 that people confuse sounding fluent with understanding until asked to explain the mechanism. [rozenblit-keil-2002](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3062901/) You have traded browse for assertion: [the same comfort I named when a clean UI feels like work but skips the mechanism](https://www.gaganmalik.io/en/newsroom/in-defence-of-thinking). I wrote about that failure mode when I [built a personal model on my own corpus](https://www.gaganmalik.io/en/newsroom/why-i-built-a-personal-llm): retrieval that returns you to what you already argued is a different habit from an answer tab that lets you ship conclusions you have not earned. Margaret Mitchell, formerly of Google's AI ethics team, warned in *MIT Technology Review* in 2023 that citations can give wrong answers an air of correctness users never double-check. [mit-tr-trust-ai-search](https://www.technologyreview.com/2023/02/14/1068498/why-you-shouldnt-trust-ai-search-engines/) Search keeps the user in charge of which page to trust. The page keeps the burden of proof where it belongs. #### Chat Finds Answers Ask exists for a different job. Someone arrives with a sentence, not a keyword: *What did you do at Apple?* or *How do you think about design systems?* or *Can we work together on pricing?* They want a direct response they can test in minutes, with footnotes. Grounded chat, retrieval-augmented generation with citations, is the right shape for that moment. The model composes; the corpus constrains; the user can click through to the lemma. The homepage Ask entry: one question routes to grounded answers with citations. That only works if you treat chat as product, not as a search bar with adjectives. Rate limits. Query caps. Modes that route contact questions differently from opinion essays. A step ceiling on tool calls so the agent does not loop forever. Citation markers stripped when they do not match retrieved chunks, because researchers presenting at ACM SIGIR in 2025 found that up to fifty-seven per cent of RAG citations can be post-rationalised rather than genuinely derived from retrieved context. [sigir-rag-faithfulness-2025](https://doi.org/10.1145/3731120.3744592) Brown's Nielsen Norman Group study found the same gap from the user side: even when chatbots supplied citations, participants could not tell which claims mapped to which sources. [nngroup-infoseeking-2026](https://www.nngroup.com/articles/ai-search-infoseeking/) I have been sharpening that stack for months: query expansion, reranking, golden-set evals you can regression-test like any other feature. The interview question is not "Do you have AI?" It is "Can you show me where the answer came from?" #### Browse Builds Trust There is a third mode people forget when they argue for one box to rule them all: browse. Reading a full case study. Listening to an essay on a commute. Following an argument across sections until you feel the warrant, not just the conclusion. Euclid's permission stack only works if someone stays in the [demonstration long enough to see the postulates](https://www.gaganmalik.io/en/newsroom/the-book-that-trains-you-to-demonstrate). Search on *design systems at scale* gets you the page list from that first query: titles, links, types. Ask gets you the paragraph you can test in eight minutes. The full essay is whether you still trust the work after forty minutes of reading. When Google rolled out AI Overviews in 2024, publishers worried that paragraph-length summaries at the top of results would choke off the traffic that funds original reporting: users would read the synthesis and never open the page. [nyt-google-ai-publishers](https://www.nytimes.com/2024/06/01/technology/google-ai-search-publishers.html) Design portfolios compress browse into thumbnails and hope evaluators infer depth. That is why I ship long-form writings, listen mode, and case studies with problem, approach, and outcome spelled out, not only a chat widget. A recruiter with eight minutes should be able to ask a question and get cited sources. A peer with forty minutes should be able to read the essay that names the tradeoff. Collapsing those paths into one generative surface optimises for the eight-minute win and taxes the forty-minute proof. #### The Wrong Instrument The counterargument is fair and I have made it myself: users do not want to learn your information architecture. They want one place to type. I tried merging archive search and Ask in a wireframe: one input field, empty archive silently calling the model. I rejected it. The merged flow looked cleaner in a screenshot than two honest surfaces do; it also hid which instrument the user had asked for. The industry bet is still real. *MIT Technology Review* named generative AI search one of its ten Breakthrough Technologies for 2025 because it returns concise answers instead of link lists. [mit-tr-genai-search-2025](https://www.technologyreview.com/2025/01/03/1108820/generative-ai-search-apple-google-microsoft-breakthrough-technologies-2025/) Sundar Pichai told *The New York Times* in May 2025 that Google's AI Mode was a total reimagining of search. [nyt-google-ai-mode](https://www.nytimes.com/2025/05/20/technology/personaltech/google-ai-mode-search.html) Perplexity bills itself as an answer engine with inline citations; a *PCMag* reviewer who switched from Google for a week said it excelled at transparently sourced research queries. [pcmag-perplexity](https://www.pcmag.com/news/is-perplexity-better-than-google-search-i-switched-for-a-week-to-find-out) Assume one box is always better: then archive search and Ask should merge tomorrow, and every empty archive state should silently call the model. Nielsen Norman Group usability research found that site chatbots which recreate search in a conversational interface are often slower than established search, filters, and navigation, adding friction to problems already solved. [nngroup-site-chatbot](https://www.nngroup.com/articles/site-ai-chatbot/) Empty keyword search often means the user phrased a question where the system expected a term. The fix is not to hide search behind chat. The fix is a handoff: you searched; here are pages; if you meant a question, continue in Ask with this query pre-filled. Different instrument, explicit bridge. That is product design, not feature sprawl. Maybe your corpus is tiny and browse does not matter yet. At roughly five hundred sessions a month, I still split the systems because the discipline scales before the traffic does. When the corpus grows, you will already know which index serves which job. When evaluators ask how you think about AI search, you can point at live surfaces instead of a slide. #### What I Will Say in the Room I ship search, chat, and browse on one domain because hiring managers, peers, and clients do not share one job to be done. The craft is naming the contract each surface keeps and encoding it in guardrails, not picking the smartest model. Last week I almost merged them anyway, because a single AI box looks cleaner in a screenshot than two honest ones. --- ### The Context Window Got Huge. Our Thinking Did Not. [Read more](https://www.gaganmalik.io/en/newsroom/the-context-window-got-huge-our-thinking-did-not) [Retrieval brief — the-context-window-got-huge-our-thinking-did-not] Category: technology Thesis: Six months into my first YouTube video, I pasted a folder and asked for cinema. The shot list felt like progress. The timeline was still empty. Answers: What does Gagan argue in "The Context Window Got Huge. Our Thinking Did Not"? | What is Gagan Malik's view on The Context Window Got Huge. Our Thinking Did Not? | How does this essay relate to AI and technology product decisions? Concepts: technology, Gagan Malik, writings By Gagan Malik Six months in, my first YouTube video still has no upload date. The timeline is open. The playhead has not moved. I am behind. I know it is procrastination. I opened the chat window anyway. Three old script PDFs. Moodboard images. Character arcs. One line: make a cinematic video. The model sent back a shot list, chapter beats, and b-roll I do not own. It felt like progress. My camera, however, is still on my desk collecting dust. Here is what I told myself that morning. The problem was access. If the machine could read everything at once, it could tell me how to make this video. That is what the industry is selling right now: bigger context windows, paste your whole archive, get the answer back. That is not the problem. I do not need a bigger notebook. I need to know what I am trying to say. I am avoiding that question by dumping those scripts, moodboards, and character notes into a chat box and asking for cinema. The context window is huge. The way we frame problems is not. I have argued that [thinking time is not laziness](https://www.gaganmalik.io/en/newsroom/in-defence-of-thinking). I trade the hard start for a polished answer. So why do bigger windows still produce worse answers than a careful human with a folder? Because we kept asking in crumbs and aimed it at a machine that can read eight novels. The bottleneck was never tokens. It was the question. #### Search Bar Brain The playhead is still at zero while I scroll a shot list the model invented for footage I never shot. Same dodge, shinier tool. Semrush's 2025 Google search statistics roundup, citing Growth Memo's State of Search research, puts the average United States query at 3.4 words. [semrush-search-stats](https://www.semrush.com/blog/google-search-statistics/) We trained on crumbs for decades. Make a cinematic video is the same muscle: compress the job, hope the machine infers the rest. Growing up in Delhi, I fought for thirty minutes on a shared PC and spent half of it hunting the right tab. The chat box did not fix that habit. It gave it a bigger mouth. #### A Bigger Box, Same Wallpaper Vendors are selling square footage. The flagship APIs now pitch context windows around a million tokens. That is enough to put the entire library on the table while you still show up with search-bar brain and a panic folder of moodboards. [openai-models](https://platform.openai.com/docs/models) I did not film that morning. I mistook output for momentum again. My scripts and moodboards sat in a folder with no names, the same hoarder logic I see in every enterprise search demo: lots of stuff, no agreed inventory, nobody willing to say what we are actually making. If your drives are a junk drawer, a bigger drawer does not fix the missing system. You still cannot ship what you cannot locate, label, or agree is the order. #### The Question Is the Interface I did not need more files in the window. I needed to press record. It is easier to paste the prep pile and ask for cinema than to name the video out loud. I typed make a cinematic video. What I owed the work was shorter and harder: twelve minutes, desk b-roll only, one claim before minute three, three visual anchors I could actually find. I said none of that into the chat box. The timeline stayed empty. The camera stayed on the desk. #### Feed the Window in Rounds That afternoon I tried again with ten labelled clips from 2024 b-roll, not the whole mystery folder. I asked for a shot list those files could carry. The answer was smaller and less embarrassing. I closed the chat tab, looked at the playhead, and still could not press record. I [built a personal language model on my own corpus](https://www.gaganmalik.io/en/newsroom/why-i-built-a-personal-llm) because synthesis should preserve judgment, not replace it. Pasting the panic folder anyway was the same bad habit in a bigger box. Around eighteen million people in the United Kingdom have used generative AI tools, according to Deloitte's 2025 press materials. [deloitte-uk-genai](https://www.deloitte.com/uk/en/about/press-room/over-eighteen-million-people-in-the-uk-have-now-used-generative-ai.html) Almost everyone has access now. Fewer people have a brief worth pasting. #### The Counterargument I Take Seriously Google's Gemini 1.5 technical report, published in 2024, documents needle-in-a-haystack retrieval across very long contexts when the task is explicit and the haystack is labelled. [gemini-15-arxiv](https://arxiv.org/pdf/2403.05530.pdf) I have seen that work in careful setups: dated exports, named folders, a retrieval stack that knows which drive is which. Long context is a genuine unlock when you are not lying about what you have. My morning was not that setup. It was contradictory scripts, untagged moodboards, and a prompt that asked for cinema instead of a question. When the corpus is curated and the ask is scoped, the window helps. When the workflow is paste everything and hope, you get a shot list for a film you never had. The camera keeps collecting dust. #### Three Sentences. Then Stop. Stop buying square footage for a workflow that will not name the question, label the files, or open the timeline. The expensive mirror is a shot list that reads like cinema while the playhead stays at zero. Six months in, my camera is still on the desk, and I am still the one who has to press record. --- ### The GUI Sells Comfort. The CLI Sells Control. [Read more](https://www.gaganmalik.io/en/newsroom/the-gui-sells-comfort-the-cli-sells-control) [Retrieval brief — the-gui-sells-comfort-the-cli-sells-control] Category: technology Thesis: Pretty tools cap your leverage. The terminal is where developers compound speed into roadmap power designers were never invited to touch. Answers: What does Gagan argue in "The GUI Sells Comfort. The CLI Sells Control"? | What is Gagan Malik's view on The GUI Sells Comfort. The CLI Sells Control? | How does this essay relate to AI and technology product decisions? Concepts: technology, Gagan Malik, writings By Gagan Malik A customer could not check out. The site had been broken since morning. In the meeting, a designer walked us through how the failure should feel: plain error copy, a visible retry button, a link to support. Good work. Necessary work. Then the lead engineer shared her screen. Not Figma. A terminal. A misconfigured payment webhook. She rolled it back in four commands. Checkout worked again. The error screens never shipped. They were not the problem. That black window is a command line. The polished screens are a graphical user interface, or GUI. A GUI is where you point and click inside software someone else built. A CLI is where you type instructions the machine obeys directly. Designers mostly work in GUIs: Figma, design tools, product surfaces. Engineers who ship under pressure reach for the CLI because the bug usually lives underneath the buttons. That is why developers keep winning the arguments that decide what ships. Not because designers are less smart. Because mockups describe the experience; commands change the system. I learned the same split studying computer science in London, long before product design paid my bills. The terminal graded your work. The GUI arrived later to make everyone else feel safe. The GUI is not the enemy. It is a toll booth. The command line is the bypass road. #### Every Click Is a Toll Booth GUIs were built for a noble reason. Most humans cannot hold a thousand file paths in working memory. Buttons, icons, and menus translate intent into action without asking you to speak the machine's dialect. The trade is composability. Every GUI routes you through what the vendor chose to expose. Rename four hundred assets in a design export? Click, confirm, repeat, mis-click, undo, lose an hour. The same job in a shell is one pattern, one preview flag, one commit. Douglas McIlroy's Unix philosophy, written in 1978, still reads like a product spec: write programs that do one thing well; write programs to work together; write programs to handle text streams, because that is a universal interface. [bell-labs-unix](https://homepage.cs.uri.edu/~thenry/resources/unix_art/ch02s01.html) Tool vendors monetise seats, storage tiers, and export limits because the GUI is the product and the cap is the business model. When your work lives inside their canvas, you are fast until you hit the fence. When your work lives in text, you can grep your own life, version it, and ship it somewhere the canvas never imagined. #### Xerox Painted the Window. Unix Kept the Pipes. The modern GUI did not begin in a keynote. It began in a research lab. In the 1970s, Xerox PARC built the Alto: bitmapped screen, mouse, overlapping windows, icons. The Computer History Museum treats it as the machine that made the graphical interface legible to the industry. [chm-alto](https://computerhistory.org/revolution/input-output/14/350/1894) In December 1979, Steve Jobs and Apple engineers visited PARC as part of a business deal and saw a curated demo of that environment. Jobs later told biographer Walter Isaacson that within minutes he knew every computer would work that way someday. [livingcomputers-parc](https://livingcomputers.org/Blog/What-Really-Happened-Steve-Jobs-@-Xerox-PARC-79.aspx) Historians at the Raskin Center and Stanford's Macintosh archives complicate the legend: the Macintosh project was already underway, PARC had shown the GUI to thousands of visitors, and Apple still had years of design work ahead. [raskin-parc](https://raskincenter.org/mac/xerox-parc/) [stanford-parc](https://web.stanford.edu/dept/SUL/sites/mac/parc.html) The lesson is not that Jobs stole the future in an afternoon. It is that he became evangelical about the layer humans touch. That evangelism worked. The Macintosh made computing feel friendly to people who would never type `ls`. Xerox's own commercial GUI machine, the Star, sold poorly while the metaphor escaped the lab. [chm-gui-revolution](https://computerhistory.org/revolution/personal-computers/17/312) Meanwhile Unix, born in the same decade at Bell Labs, kept optimising for text streams, pipes, and scripts: the machinery underneath the metaphor. Neal Stephenson named the tension in his 1999 essay *In the Beginning... Was the Command Line*, which opens with Jobs, Wozniak, Gates, and the strange idea that operating systems would become consumer products at all. [stephenson-command-line](https://www.nealstephenson.com/in-the-beginning-was-the-command-line.html) The GUI won the living room. The CLI kept the boiler room. #### The Demo Tapes Agreed With the Plumbers This argument is not new. It is old enough to have a filmography. In 1968, Douglas Engelbart's "Mother of All Demos" at Stanford Research Institute showed the mouse, hypertext, and collaborative editing in one session: a graphical breakthrough that still assumed expert operators at the keyboard. [engelbart-demo](https://www.dougengelbart.org/content/view/138/) Richard Gabriel argued in 1989's "Worse Is Better" that Unix would beat purer Lisp systems because ugly, simple tools that shipped beat beautiful systems that did not. [worse-is-better](https://www.dreamsongs.com/WorseIsBetter.html) Paul Graham, in *Hackers & Painters*, later made the same point in startup language: the people who control the metal tend to win the product argument. [graham-hackers-painters](https://www.paulgraham.com/hackpaint.html) The pattern repeats whenever a new platform arrives. Microsoft documented API-first internal agents years before most design tools grew chat panels. [microsoft-api-agents](https://www.microsoft.com/insidetrack/blog/unleashing-api-powered-agents-at-microsoft-our-internal-learnings-and-a-step-by-step-guide/) InfoQ argued in 2024 that agentic systems split cleanly between GUI demos and CLI production harnesses. [infoq-agent-cli](https://www.infoq.com/articles/ai-agent-cli/) The editorial class keeps rediscovering what builders never abandoned. Xerox showed the window. Engelbart filmed the dream. Stephenson wrote the essay. The terminal stayed open. #### The Roadmap Follows Who Can Ship Designers are not less intelligent. They are often more literate about human anxiety than anyone in the room. That is not the argument. The argument is organisational leverage. A designer's default output is an artefact: a frame, a prototype, a spec another human must interpret. A developer's default output, when they live in the terminal, is a running diff: code merged, migration applied, flag toggled, log line that proves the fix landed. Roadmaps are not won by who felt the problem most clearly. They are won by who can answer the 9 p.m. question: can you ship it before the customer churns? Stack Overflow's 2024 Developer Survey lists Bash and Shell among the most common professional languages, at roughly a third of respondents, alongside the stacks that actually ship software. [stackoverflow-2024](https://survey.stackoverflow.co/2024/technology) Wealth is created where latency and auditability live: repositories, queues, idempotent scripts. Vanity is served where humans need reassurance: splash screens, polished decks, demo reels that loop on conference Wi-Fi. Developers who command the CLI do not win because they hate beauty. They win because beauty without merge rights is a poster on someone else's wall. #### The Category Shifts Were Not Storyboarded First Look at what actually moved the market. Peter Steinberger has said the first functional build of what became OpenClaw took about an hour at the terminal, then a few more hours once image support landed, built with Codex rather than a design sprint. [meshedsociety-openclaw](https://meshedsociety.com/peter-steinberger-openclaw-openai-thibault-sottiaux-openai-i-coded-the-first-version-of-openclaw-in-an-hour/) Boris Cherny started Claude Code in September 2024 as a terminal side project: no graphical interface, just a CLI talking to the model, hacked on nights and weekends until Anthropic shipped it publicly in February 2025. [stationf-cherny](https://stationf.co/news/boris-cherny) [github-claude-code](https://github.com/anthropics/claude-code) These were engineer-built shells, not Figma files handed to engineering for implementation. ChatGPT is the honest complication, not the counterexample. Evan Morikawa, who led Applied engineering at launch, has described a small team of engineers, designers, and researchers grabbing a room in the summer of 2022 and iterating until a minimalist chat window went live in November. [pragmaticengineer-chatgpt](https://newsletter.pragmaticengineer.com/p/inside-openai-how-does-chatgpt-ship) Designers were in the room. The surface that rewired the industry was still the sparsest possible GUI: a text box, built at engineering clock speed. OpenClaw in hours. Claude Code from a terminal prototype. ChatGPT from a spare chat UI in months, not a component-library quarter. Design tools added AI chat panels later. The category winners shipped from the shell first. #### You Can Play the Song on the Piano Roll. You Cannot Transpose It at 2 a.m. Try learning a piece using only the lighted keys on a digital piano app. You can perform the song. You might even impress someone at a party. Now the singer is ill, the set is in a different key, and the drummer wants a loop for rehearsal. If all you ever learned was which blocks to hit in order, you are stuck requesting a feature from the app store. If you know the chord progression, you transpose in minutes and export a backing track before anyone opens Slack. GUIs are the lighted keys. CLIs are the progression. That disorientation when the tool will not let you do the obvious fix is not user error. It is rented capability hitting its fence. #### I Used to Think the Handoff Was the Product I am not writing from a terminal cult. I have signed design-led vendor stacks because the stakeholder demo looked legible. At a shipping logistics company I was Principal Designer on a marine services platform: the software port agents and vessel coordinators use to book divers, surveyors, and engineers against arrival windows. The admin console looked humane. Colour-coded job statuses. Drag-and-drop scheduling. Dave, our lead engineer, lived in the API docs and the shell. A vessel was due alongside at 0600. At 1700 the night before, operations called: two hundred service jobs showed the wrong berth assignment on the floor tablets. Not a visual bug. A bad import had stamped the wrong port code on every row. I had designed the recovery flow months earlier: clear error states, a manual reassign screen, escalation to support. Dave shared his screen. Not the design file. A terminal. One query found the bad rows. One script rewrote the port codes against staging. One deploy before the handover call. My screens were correct. They were not the fix. The jobs cleared before the pilot boarded. Dave shipped before midnight. I filed a follow-up ticket to make bulk reassignment easier in the UI. The stand-up praised engineering velocity. That is the developer-designer gap in one port call. Not taste versus logic. Export rights versus execution rights. #### The Strongest Case for Letting Everyone Live in the GUI Give the objection its full weight. Graphical interfaces democratised computing for billions of people who will never type `chmod`. Steve Jobs was right to fall in love with what he saw at PARC: the GUI was how ordinary humans would finally use machines without fear. Design is how products earn trust: typography, affordance, the button that does not feel like a trap. Forcing everyone through a terminal is gatekeeping dressed as meritocracy. The best teams pair visual literacy with technical depth. Hold that case at full strength, then look at who gets called when revenue is on fire. The GUI democratised access. It did not democratise deploy keys. When the incident is live, nobody asks for the warmest empty state. They ask who can run the command, read the log, and prove the rollback worked. The comfortable interface was never neutral: it traded memorisation for composability. Developers keep winning rooms because the terminal does not cap what they can chain together. I still know what the operators need; Dave still holds the merge button, and until that changes, one of us keeps filing tickets while the other goes home. --- ### People Don't Buy Products. They Hire Them for Jobs. [Read more](https://www.gaganmalik.io/en/newsroom/people-dont-buy-products-they-hire-them-for-jobs) [Retrieval brief — people-dont-buy-products-they-hire-them-for-jobs] Category: business Thesis: Customers hire for progress and fire for stall. Lynn Shostack saw it at Citibank in 1977; SaaS still sells CVs. Answers: What does Gagan argue in "People Don't Buy Products. They Hire Them for Jobs"? | What is Gagan Malik's view on People Don't Buy Products. They Hire Them for Jobs? | How does this essay relate to product strategy and growth? Concepts: business, Gagan Malik, writings By Gagan Malik In 1977 Lynn Shostack sat in a Citibank conference room while a colleague walked a new private-bank package on slide four: fee tiers, statement design, envelope stock. She had spent the prior years at Greenwich Associates watching financial services firms market trust the way supermarkets market soup. [princeton-shostack](https://www.princeton.edu/news/2011/06/16/shostack-gives-10-million-endow-project-x-fund-giving-freedom-tinkerers) She was vice president by then, marketing director for the investment management group, [shostack-1977](https://doi.org/10.1177/002224297704100219) and the room was nodding on autopilot. Then she asked the only question the deck could not answer: what happens to the client on the morning the report is wrong and nobody picks up the phone? No transcript captures that room; what follows matches her published role and the argument she would take to print. Her colleague said operations would cover it. He opened slide five. Nobody in the room could describe that morning. That silence is not a sales failure. It is a category error. Nobody buys your product. They hire something to make progress on a job in a specific circumstance, then fire it when progress stalls or a better hire appears. Shostack published the argument in the *Journal of Marketing* in April 1977 under the title "Breaking Free from Product Marketing." [shostack-1977](https://doi.org/10.1177/002224297704100219) Tech would spend the next fifty years repackaging the hire logic and forgetting who drew the blueprint first. We all have sat in a similar room wearing different clothes. Hers was fee tiers and envelope stock on slide four. Yours might be forty integrations and a per-seat price on slide six. In both rooms the deck skipped the same beat: what happens the morning after sign-off, when the exception queue lands and nobody from the vendor picks up the phone. Or you subscribed to a tool you "should" use while still doing the job in a spreadsheet because the hire never started. The claim is not about catalogue purchases made for pleasure or status. It is the hire logic that shows up when a team renews software while the spreadsheet still runs the job. AI copilots have pushed many category reviews toward one blunt question: what job does this replace in someone's actual week? The purchase was not possession. It was temporary employment with a performance review you never scheduled. #### Feature Roadmaps Are HR Departments for Objects The conventional view is simple and wrong in a useful way: we sell products, growth means more seats and more releases, and a healthy account expands. That story benefits vendors who earn expansion revenue when a module ships. It benefits agencies who sell integration statements of work that grow with every connector badge, analysts who sort firms into quadrants by feature breadth, and product managers whose promotion packet needs a visible widget in the sprint demo. Jobs are harder to commit to in Jira because they expire when context changes. A feature can sit on a roadmap for three quarters and still look like progress. We measure the wrong instrument and call it health. Activation funnels, monthly active users, time-in-app: these track whether the object stayed on the payroll, not whether the job got done. Shostack saw the same mismatch in banking dashboards that counted accounts opened while the service experience leaked at the touchpoints nobody mapped. The label was wrong before the metric was. #### She Watched Banks Market Trust Like Soup For a century, product meant something you could hold, stock, and ship: inventory on a shelf, a drill in a box, a milkshake in a cup. Much of B2B SaaS imported the word anyway. Subscriptions, onboarding calls, success managers, and API limits became "products" because the label sounded scalable and pitch decks already had a slide for it. Shostack had started as an art major at the University of Cincinnati and finished at Harvard Business School with no interest in business until she discovered it was plastic enough to mould. [princeton-shostack](https://www.princeton.edu/news/2011/06/16/shostack-gives-10-million-endow-project-x-fund-giving-freedom-tinkerers) Her father was a mechanical engineer with a workshop in the garage. She learned early that performance happens over time, not at the point of sale. In 1977 she argued that services are intangible, co-produced with the customer, and consumed as they are delivered, which means marketing them with product playbooks is mirror-image wrong. [shostack-1977](https://doi.org/10.1177/002224297704100219) Seven years later, in *Harvard Business Review*, she opened with the symptoms everyone recognises: the shirt that returns with a broken button, the car rattle that reappears after an expensive repair, the representative who promises a callback and does not call, the cash machine that swallows your card. [shostack-hbr-1984](https://hbr.org/1984/01/designing-services-that-deliver) She argued that services need their own design discipline and drew the service blueprint so teams could see the backstage work a client never should have to notice. Stephen Vargo and Robert F. Lusch argued in the *Journal of Marketing* in 2004 that value is co-created through service, not transferred like goods on a loading dock. [vargo-lusch-2004](https://doi.org/10.1509/jmkg.68.1.1.23952) I have shipped onboarding flows that looked like product tours because the roadmap was written in features, not acts in a service over time. You do not unbox a subscription. You live inside it until you fire it. #### The Same Milkshake Hires Two Different Employees Clayton Christensen turned hire logic into the Jobs to Be Done framework: customers make progress in a circumstance, not collect features on a shelf. His milkshake research is the textbook case. The same cup gets hired twice in one day, thick enough to kill boredom on a morning commute, sweet enough to bribe a parent with a child in the evening. The milkshake was a product in the old sense, tangible and stocked. Christensen still wrote product; the hire logic travels when Silicon Valley mislabels a service as one. He codified the theory in *Competing Against Luck* (2016), with Taddy Hall, Karen Dillon, and David S. Duncan, and in "Know Your Customers' 'Jobs to Be Done'" in the *Harvard Business Review* in September 2016. [christensen-luck](https://www.amazon.com/Competing-Against-Luck-Innovation-Customer/dp/0062435612) [christensen-jtbd-hbr](https://hbr.org/2016/09/know-your-customers-jobs-to-be-done) Different employer, different success criteria, different substitute when the hire fails. Theodore Levitt made the earlier version in "Marketing Myopia" in the *Harvard Business Review* in July 1960: industries decay when they ship drills instead of holes. [levitt](https://hbr.org/1960/07/marketing-myopia) Christensen, Scott Cook, and Taddy Hall had already argued in "Marketing Malpractice" in the *Harvard Business Review* in December 2005 that firms should segment by job, not demographic, before the framework had its name on the slide. [christensen-malpractice](https://hbr.org/2005/12/marketing-malpractice-the-cause-and-the-cure) I argued the buyer's side of that table in [People Love to Buy](https://www.gaganmalik.io/en/newsroom/why-people-love-to-buy): agency lives in naming the problem before anyone opens a deck. Shostack arrived earlier with a different instrument: map the service before you market the package. The competing hire is rarely your listed competitor. It is Excel, a shared inbox, an intern for a week, or the workaround someone built at midnight because the SaaS never matched how exceptions actually move. Every category review eventually returns to the same blunt question she asked in that Citibank room: what job does this replace, and what happens to the person still doing it manually? #### Firing Is Not Churn. It is the Performance Review. Switching tools is not betrayal. It is HR. The customer fired the incumbent because progress stalled, the onboarding never transferred the job, or a cheaper hire appeared that actually cleared the queue. Churn charts record the exit interview after the decision was made. By the time renewal fails, the spreadsheet has often been doing the work for months. The vendor learns about the firing from a logo churn slide, not from the morning in week two when the tab stopped opening. Demand, when it is real, is for the job done at economics that work, not for your SKU on the invoice. I wrote that shortage logic in [Product Market Fit Is What Economists Call a Shortage](https://www.gaganmalik.io/en/newsroom/pmf-is-demand-exceeding-supply): pull at a viable price means people want more of the outcome than you can comfortably deliver, not more of your feature map. Shostack's HBR symptom list is a performance review written in domestic prose. A team still colour-coding exceptions at ten in the evening while the subscription renews is not a retention problem waiting for a win-back email. It is an employee who never showed up for the shift. #### She Drew the Blueprint. I Renamed the Epics. Shostack did not leave banking marketing to write theory and stop. She went on to Bankers Trust, then manufacturing, then a long second act in philanthropy and investment with her husband David Gardner, a Princeton graduate who died in 2001. [princeton-shostack](https://www.princeton.edu/news/2011/06/16/shostack-gives-10-million-endow-project-x-fund-giving-freedom-tinkerers) She funded endowments for curiosity without quarterly justification, including Princeton's Project X fund for engineers who need freedom to chase a hunch. She chronicled coming up through a world of men in a private book called *Men in Business*. [men-in-business](https://smalleditions.nyc/Lynn-Shostack) The through-line is the same: name the work, then design the system that performs it. I ran a Jobs to Be Done workshop at a client offsite once. Sticky notes lined the wall, each labelled "Job," and by the second coffee break I could see they were epics in disguise: "Streamline workflow," "Increase visibility," "Enable collaboration." Nobody wrote what changed in an inbox before noon. Shostack would have asked for a blueprint line by line. [shostack-hbr-1984](https://hbr.org/1984/01/designing-services-that-deliver) I did not push. That was the brief. I left with a workshop fee and a slide titled "Innovation." They left with the same dashboard she would have recognised from 1977. #### Product Designer Is the Trendy Mislabel I call myself a product design expert on profiles and proposals because the label wins searches and signals fluency in the current jargon. The work is often service design by Shostack's definition: onboarding choreography, handoffs, exception flows, the week after sign-off. Experience designer would be closer. Service designer would be honest. Product designer is what clients expect to buy, and I have not corrected them. Sir Jony Ive trained in industrial design and built a career on objects you could hold: hinge tolerances, chamfered aluminium, weight in the palm. [ive-industrial-design](https://carnegie.org/great-immigrants/jony-ive/) I have never shipped a billet. I suspect he would find our title inflation absurd even if he never said so on the record. The hypocrisy is not only in the vendor deck. I wear the mislabel because it rents well. #### The Expensive Striker Still Gets Benched You have watched a football match turn and felt the frustration before you had the vocabulary for it. Your club is losing. The manager benches the £50 million signing at the sixtieth minute and sends on a kid from the academy who has not scored in months. Nobody in the stand thinks loyalty to the transfer fee matters. The job tonight is to score. The asset is the player's CV. Expected-goals charts arrive on the phone later. Your stomach already knew at the hour mark. That substitution is what your customer does when they reopen the spreadsheet. They are not being disloyal to your roadmap. They are losing, and you stayed on the pitch while the queue kept growing. #### The Strongest Case for Selling the Product, Not the Job Let me state the strongest case against this, because the alternative is a workshop handout. Customers cannot articulate jobs in interviews. They lie, post-rationalise, and describe the product they already bought. The Jobs to Be Done framework can become retroactive storytelling after the fact. Christensen and colleagues argued in the September 2016 *Harvard Business Review* that the job is often discovered through observation in context, not through a survey that asks what people want. [christensen-jtbd-hbr](https://hbr.org/2016/09/know-your-customers-jobs-to-be-done) In enterprise software the purchase is often risk transfer: vendor balance sheet, SOC 2 certificate, latency SLA, the name on the contract when audit asks questions. Technical buyers genuinely hire specs, not feelings. Clayton Christensen acknowledged jobs shift when context shifts. Good product managers already track outcomes and north-star metrics. In categories where compliance is the job, selling "progress" sounds like marketing fluff next to a procurement matrix. If you wait for perfect job clarity, you never ship. The incumbent with the louder feature list wins by inertia. The gap is not whether specs matter. It is whether your firm can state the job you are hired for before the customer writes the performance review for you. Clayton Christensen, Scott Cook, and Taddy Hall argued in "Marketing Malpractice: The Cause and the Cure" in the *Harvard Business Review* in December 2005 that segmentation by job, not demographic, predicts which innovations succeed, and that most companies cannot articulate the job they are hired to do. [christensen-malpractice](https://hbr.org/2005/12/marketing-malpractice-the-cause-and-the-cure) Alan Klement pushed the same line in *When Coffee and Kale Compete* (2018): customers pull products that help them make progress, not products that win feature comparisons on a slide. [klement-coffee-kale](https://www.amazon.com/When-Coffee-Kale-Compete-Progress/dp/0998534608) Shostack's blueprint was the operational answer Christensen's milkshake case still leaves abstract: draw the morning the client is living, then decide if your service earns a renewal. Customers hire for progress and fire for stall. Until your success metric names the before-and-after in the customer's week, you are optimising a CV, not a hire. Lynn Shostack drew the blueprint more than forty years ago; the spreadsheet is still doing the job her line of visibility was drawn to protect. --- ### Subscription Fatigue Is the Wrong Instrument [Read more](https://www.gaganmalik.io/en/newsroom/subscription-fatigue-is-the-wrong-instrument) [Retrieval brief — subscription-fatigue-is-the-wrong-instrument] Category: business Thesis: Fatigue blames the household. Exit friction, ghost spend, and buried cancel flows are what the dashboard refuses to count. Answers: What does Gagan argue in "Subscription Fatigue Is the Wrong Instrument"? | What is Gagan Malik's view on Subscription Fatigue Is the Wrong Instrument? | How does this essay relate to product strategy and growth? Concepts: business, Gagan Malik, writings By Gagan Malik It was a Sunday night in London and I opened my banking app to pay a council charge before the week started. The payment screen loaded. So did two subscription line items I had stopped thinking about: a UFC subscription and Apple TV+. Roughly £34 a month for one fight card I never rewatched and a limited series on Apple TV+ I finished in autumn. I have not opened either app since. The charges did not pause while I forgot. The industry calls this subscription fatigue, as if the problem were a tired consumer who lost interest in paying for things they love. The problem is a trained asymmetry: one tap to enrol, a cancel path that often feels like twenty minutes, and a metric stack that counts the enrolment while the ghost charge hums on a statement you stopped reading. #### The Dashboard Counts Subscribers. It Does Not Count Exits. Trade press and retention consultants have a favourite noun for the moment you notice the stack: fatigue. Pam Danziger wrote in Forbes in 2022 about consumer subscription fatigue as the mood that follows too many boxes ticked at checkout. [forbes-danziger](https://www.forbes.com/sites/pamdanziger/2022/05/10/as-retail-opens-up-and-inflation-closes-in-consumer-subscription-fatigue-is-on-the-rise/) Adriana Lacy, writing on Substack in 2025, described the same drain from the creator side: even publishers feel the ceiling when readers hit stack limits. [lacy-substack-fatigue](https://blog.adrianalacyconsulting.com/substack-subscription-fatigue-journalism-beehiiv/) MIDiA Research asked in 2024 whether we are headed towards subscription fatigue, pairing heated-seat backlash with a broader question about ownership versus access. [midia-fatigue](https://www.midiaresearch.com/blog/are-we-headed-towards-subscription-fatigue) The label persists because it locates failure in the household. You forgot. You lost appetite. You need a spring clean. Meanwhile Juniper Research, in a press release on 1 October 2025, projected the global subscription economy growing from $722 billion in 2025 to $1.2 trillion by 2030, a 67% rise it explicitly framed as happening despite increasing subscription fatigue. [juniper-subscription-economy](https://www.juniperresearch.com/press/subscription-economy-to-reach-1tn-by-2030/) The market expands while the dashboard still counts subscribers, not successful exits. MIDiA Research, in *Video churn: Streaming TV's cord-cutting moment* (June 2024), argued that churn is accelerating and becoming a systemic issue for subscription video on demand for the first time as streaming supplants traditional pay TV. [midiavideo-churn](https://www.midiaresearch.com/reports/video-churn) Josh Kaufman, in *The Personal MBA*, names part of why: subscriptions are profitable because they maximize lifetime value. [kaufman-pmba-ltv](https://personalmba.com/) We did this with other instruments before. The unemployment rate can look fine while underemployment sits off the chart. Engagement metrics can glow while the person behind the screen is exhausted. Fatigue is the wrong instrument here for the same reason: it measures mood while the harm sits in the flow. In [Someone Designed That Button](https://www.gaganmalik.io/en/newsroom/someone-designed-that-button) I argued that design harm has coordinates, not weather. Auto-renew defaults and win-back campaigns benefit vendors who treat churn as a feeling to massage rather than a journey someone already tried to complete. The subscriber count rises. The exit integrity never appears on the slide. #### You Sign Up at Tap Speed. You Leave at Bureaucracy Speed. Convenience is real on the way in. Josh Kaufman, in *The Personal MBA*, defines a subscription as predefined benefits on an ongoing basis in exchange for a recurring fee. [kaufman-pmba-def](https://personalmba.com/) Paul Graham, in *Hackers & Painters*, argued that charging subscription fees openly is cleaner than forcing people to keep buying and installing new versions when software development is an ongoing process, and that subscriptions are the natural way to bill for web-based applications. [graham-hackers-painters](https://www.paulgraham.com/hackpaint.html) Streaming libraries you cannot buy once, cloud sync across devices, bundled software that would cost more to licence separately: that part of the bargain is not a fiction. Side B is what happens when enrolment is optimised like a sprint and cancellation is optimised like an obstacle course. Those patterns show up across SaaS, streaming, and physical goods: the Federal Trade Commission's July 2024 review sampled 642 subscription websites and apps globally, not one category alone. The Federal Trade Commission, reporting on 10 July 2024 on that international review, found that nearly 76% employed at least one possible dark pattern. Eighty-one per cent did not allow consumers to turn off auto-renewal during purchase. Seventy per cent did not provide cancellation information during enrolment. UK adults, in Aqua card research published in 2025 surveying 2,000 people aged 16 and over, reported average subscription spend of £786 a year across 2.8 services, about £65.50 a month. One in five said they rarely use all the subscription services they pay for. Seventeen per cent do not track their subscriptions at all. Fatigue is a tracking and design failure before it is a taste failure. The product you thought you bought can move while you stay enrolled. Disney+ launched in the UK in March 2020 at £5.99 a month for an all-in tier. By 2025, a comparable ad-free top tier sat at £14.99 a month on Premium. BMW briefly sold heated front seats in the UK as a £15-a-month add-on, reported the BBC in July 2022. Forbes and The Verge confirmed the firm dropped the subscription in September 2023 after backlash. Colin Bryar and Bill Carr, in *Working Backwards*, describe the chicken-and-egg inside every subscription service: you need a great offering to attract paying subscribers, and you need paying subscribers to afford a great offering. [amazon-working-backwards](https://www.amazon.com/Working-Backwards-Insights-Stories-Secrets/dp/125026759X) Tier drift and feature rent are not anecdotes. They are what happens when renewal is the business model and exit is the product you never shipped. In [When a Click Costs Nothing](https://www.gaganmalik.io/en/newsroom/when-a-click-costs-nothing) I argued that removing friction on the way in moves cost to the person who must later hunt the cancel link. Settings trees are designed to expire patience before the trial does. #### Your Wallet Never Reaches Deep Sleep Sleep researchers distinguish one bad night from fragmentation: micro-arousals that never let you reach restorative deep sleep. You are technically asleep. You wake tired anyway. Your wallet can look the same on a monthly statement. The stack pings. Apple TV+ at £8.99. UFC at £24.99. A renewal you meant to cancel after the fight ended. You are financially asleep in the sense that you stopped reading the lines. The ghost subscriptions keep waking the account for long enough to feel the loss without ever reaching rest. That is not laziness. It is interrupted recovery. Fatigue, in the industry sense, assumes you failed to want the services enough. What households often experience is never reaching financial deep rest because exit friction leaves charges humming after intent changed. The subscription did not tire you out. The asymmetry kept you from closing the loop. #### The Strongest Case Against "Fatigue" Is Genuinely Strong Let me give the objection its fairest hearing. Simon Owens, in "Why the 'subscription fatigue' argument is overrated," published on his Substack newsletter on 24 November 2020, [simon-owens-fatigue](https://simonowens.substack.com/p/why-the-subscription-fatigue-argument) and in a companion piece titled "Subscription fatigue isn't really a thing," makes a case worth taking seriously. Consumers are not spreadsheet analysts of subscription count. If a product is genuinely addictive and disposable income exists, people pay. There are billions of niches. More than a billion English speakers alone. We do not write essays about book fatigue or music fatigue when someone buys another novel or album. Quality wins. Good products keep renewing because the value arrives faster than the invoice. Owens's later writing on the valley of churn [simon-owens-valley-churn](https://simonowens.substack.com/p/the-valley-of-churn) acknowledges how brutal subscription economics can be for creators too. The objection is not naive. It treats renewal as preference revealed by behaviour. Hold that case at full strength, then look at enrolment design at scale. Hostile enrolment does not prove every renewal is unwanted; plenty of people keep paying for products they love. It does prove we cannot infer preference from payment alone when the purchase flow hides exit information. If renewal were pure informed choice, the purchase flow would not need to hide auto-renew on 81% of audited properties and omit cancel steps on 70%, while nearly 76% showed at least one possible dark pattern, according to the Federal Trade Commission's ICPEN and GPEN review announced on 10 July 2024. The objection assumes repeat payment reflects desire. The enrolment environment is built to prevent exit information from arriving at tap speed. Preference and capture are not the same ledger. #### I Mapped the Paywall. I Did Not Map the Exit. I sat in a B2C SaaS workshop once, undated, whiteboard markers squeaking. The task was an end-to-end journey map of the customer experience. We mapped each screen of the entire app. There was no funnel audit, and cancel flows never got their own row on the wall. I signed off the map anyway. That is product and strategy complicity. I knew what I was normalising. Parallel to that workshop, my own statement showed roughly £34 a month across a UFC subscription and Apple TV+ I barely touch, about £408 a year in renewals I keep deferring. Adobe Creative Cloud moved from a tier I understood at about £53 a month to about £69.99 with downgrade friction I kept deferring. A later audit found about £37 a month, roughly £444 a year, across Spotify Family for a flat that had moved out, iCloud+ storage I was not using, and a HelloFresh box I thought I had cancelled. Andrej was a senior UX designer on my team, based in Northampton. The client was tied to the Adobe ecosystem. We ran Figma by default, but on this account we made an exception and bought Adobe XD for delivery. He missed a downgrade window. The tier jumped. He tried to cancel and landed in a multi-screen flow that asked whether he was sure five times before accepting the answer he already gave. Mid-export on a deadline, the account locked. The delivery window slipped. The subscription still renewed. A cancel flow that treats every confirmation click as a sales opportunity is the product reaching for the wheel after the buyer already said no. I wrote about that instinct in [Why People Love to Buy](https://www.gaganmalik.io/en/newsroom/why-people-love-to-buy): guarding spend feels like authorship until the interface steals it back. Andrej lost a handoff he could not recover. The platform kept the month. #### Three Sentences. Then Stop. What we call subscription fatigue is often subscription exit design: asymmetric signup, opaque cancel, tier drift, misread as consumer forgetfulness. Until product teams map the cancel journey with the same care as the paywall, "reduce churn" is a euphemism for trapping people who already tried to leave. Andrej missed the delivery window while Adobe charged him for another month of a tool he was trying to leave. --- ### Product Market Fit Is What Economists Call a Shortage [Read more](https://www.gaganmalik.io/en/newsroom/pmf-is-demand-exceeding-supply) [Retrieval brief — pmf-is-demand-exceeding-supply] Category: business Thesis: PMF is not a mystical growth phase. It is a shortage: pull at a price you can serve exceeds what you can comfortably deliver. What an economist would read off your waitlist. Answers: What does Gagan argue in "Product Market Fit Is What Economists Call a Shortage"? | What is Gagan Malik's view on Product Market Fit Is What Economists Call a Shortage? | How does this essay relate to product strategy and growth? Concepts: business, Gagan Malik, writings By Gagan Malik It was 2013, almost thirteen years ago now. I was on my MSc in Technology Entrepreneurship at UCL, reading Eric Ries and Ash Maurya, the two most prominent authors of the Lean Startup movement, until product-market fit and minimum viable product sounded like secret handshakes you earned in a build-measure-learn loop. [ries-lean-startup](https://theleanstartup.com/book) [maurya-running-lean](https://leanstack.com/running-lean-book) I was learning a rigorous way to build under uncertainty. Real vocabulary, real discipline. What I had not yet learned was the older name economists had already given the same idea. Fast forward to 2020. Covid had emptied London and colonised every classroom with Zoom. I was enrolled on the Chicago Booth Executive Program, sitting in my apartment in London for microeconomics while fifty-six cameras stared back from a grid dense enough to feel like a wall of witnesses. [booth-exec](https://www.chicagobooth.edu/executives) Professor Lars Stole drew supply and demand, labelled the gap shortage, and kept lecturing as if the diagram were obvious. [lars-stole](https://www.chicagobooth.edu/faculty/directory/s/lars-stole) I was not in Chicago. I was pinching the bridge of my nose at my laptop, trying to make sense of a concept that had sat in plain sight for seven years while I practised product design and strategy under a different vocabulary. Were product-market fit and that gap the same thing? They are, with a translation, not a synonym. Economists call excess demand at a price a shortage. Founders call the same squeeze product-market fit. Product-market fit is the fancy way of saying demand exceeds supply at a price and cost structure you can actually serve. Founders treat it like a secret level you only reach after enough A/B tests. An economist asks a blunter question: at the price you charge, do more people want what you sell than you can deliver without begging, discounting, or burning someone else's money? If yes, you have a shortage. If no, you have a surplus wearing a landing page. The jargon did not invent the condition. It renamed it and then sold you software to measure everything except the queue. #### PMF Sells Workshops. Shortages Do Not. The startup vocabulary around fit is vague on purpose. "Still searching for PMF" extends runway conversations. "We have PMF, now we scale" justifies headcount and ad spend. Accelerators, growth agencies, and analytics vendors all earn fees when fit stays a narrative milestone rather than a quantity on a diagram. A shortage is awkward in a board deck. It implies you are turning money away, saying no to bad-fit customers, or failing to hire support fast enough. Those are operational embarrassments with receipts. A funnel chart is an embarrassment you can workshop. So the industry built a parallel religion: retention cohorts, Sean Ellis surveys, magic numbers, and the word *feel* in every Andreessen quote pulled out of context. [ellis-pmf](https://www.startup-marketing.com/the-startup-pyramid/) [andreessen-pmf](https://pmarchive.com/guide_to_startups_part4.html) Those tools can inform. None of them is the thing itself. Maja Voje, in *Go-To-Market Strategist*, comes closer when she defines product-market fit as sufficient proof the market wants your product before you scale into a growth stage. [voje-gtm](https://majavoje.com/product/go-to-market-strategist-book/) That is a syllabus sentence for a shortage. Marshall had the diagram in 1890. [marshall-principles](https://www.econlib.org/library/Marshall/marP.html) Your stand-up had the acronym in 2013. #### The MVP Was Meant to Prove Value, Not Ship Volume Eric Ries was explicit about what minimum viable product was for. In "Minimum Viable Product: a guide," on Startup Lessons Learned on 3 August 2009, he defined it as "that version of a new product which allows a team to collect the maximum amount of validated learning about customers with the least effort." [ries-mvp](http://www.startuplessonslearned.com/2009/08/minimum-viable-product-guide.html) In *The Lean Startup* (Crown Business, 2011), he sharpened the rule: remove any feature, process, or effort that does not contribute directly to the learning you seek. [ries-lean-startup](https://theleanstartup.com/book) The unit of progress was never the feature count. It was validated learning under uncertainty. That is not how the syllabus landed in my cohort, and it is not how most founder decks read today. Somewhere between Ries and the sprint board, MVP became a product brief: ship the thinnest version, count sign-ups, call it learning. Economic value, the proof that someone will pay enough for you to serve them profitably, dropped out of the sentence. James C. Anderson and James A. Narus argued in the *Harvard Business Review* in November 1998 that suppliers who cannot quantify what their offering is worth in the customer's P&L will lose to whoever can. [anderson-narus](https://hbr.org/1998/11/business-marketing-understand-what-customers-value) Ash Maurya tried to hold the other half in *Running Lean* (O'Reilly, 2012): problem-solution fit before product-market fit, with repeatable demand at a price that makes the model work, and three questions before you scale: will customers want it, will they pay, can you serve it profitably. [maurya-running-lean](https://leanstack.com/running-lean-book) We optimised what to build before we proved what it was worth. The industry mostly kept the MVP and skipped the price. #### Marshall Drew It Before Your OKR Did Alfred Marshall drew the diagram in *Principles of Economics* in 1890: a supply schedule, a demand schedule, and the point where they cross. [marshall-principles](https://www.econlib.org/library/Marshall/marP.html) Friedrich Hayek made the epistemic version in "The Use of Knowledge in Society," published in the *American Economic Review* in September 1945. [hayek-1945](https://doi.org/10.2307/1809376) No central planner can survey every local trade-off. Prices and quantities aggregate knowledge nobody holds in one head. Thomas Sowell, in *Basic Economics*, states the mechanism plainly: a price set below the level that would prevail by supply and demand in a free market causes more to be demanded and less to be supplied, creating a shortage at the imposed price. [sowell-basic-economics](https://www.basiceconomics.com/) Hermann Simon, in *Confessions of the Pricing Man*, names the crossing point the market-clearing price, the only price at which supply and demand are in equilibrium. [simon-pricing-man](https://www.confessionsofthepricingman.com/) You do not need a perfect survey to know bread is scarce when the shelf is empty. You need a signal you cannot spin. Marc Andreessen wrote the startup translation on his pmarca blog on 25 June 2007, in "The Pmarca Guide to Startups: Part 4: The only thing that matters." [andreessen-pmf](https://pmarchive.com/guide_to_startups_part4.html) He did not say product-market fit was a vibe. He said you can always feel when it is happening. Customers are buying the product just as fast as you can make it. Usage is growing just as fast as you can add more servers. Money from customers is piling up in your company checking account. You are hiring sales and customer support staff as fast as you can. Reporters are calling because they have heard about your hot new thing. Read that list like an economist and it is a shortage symptom checklist. Demand is pulling product out of your hands faster than your supply curve can respond. Notice what his symptoms share: money at a price you can repeat, not a prototype someone praised in an interview. #### Pull Is Readable. So Is Its Absence. Avinash K. Dixit and Barry J. Nalebuff put the same idea in one line in *Thinking Strategically*: a stock price rises when the demand at the old price exceeds the supply. [dixit-nalebuff](https://wwnorton.com/books/9780393340459/thinking-strategically) Your startup version is a waitlist that grows when you do not discount, referrals that arrive before the ad budget, or a support queue that lengthens because usage is real. Customers return without a coupon code. The prospect accepts a later start date instead of walking. You feel the squeeze before the dashboard names it. W. Chan Kim and Renée Mauborgne, in *Blue Ocean Strategy*, describe the opposite world: supply rising as global competition intensifies without a matching rise in demand, ending in commoditization, price wars, and shrinking margins. [blue-ocean](https://www.blueoceanstrategy.com/what-is-blue-ocean-strategy/) David Graeber, in *Bullshit Jobs*, goes further: when supply far outpaces demand in an industry, demand gets manufactured. [graeber-bullshit-jobs](https://www.simonandschuster.com/books/Bullshit-Jobs/David-Graeber/9781501143331) Marketing can move a curve at the margin. It cannot invent a market that is not there at viable unit economics. I spent years after UCL watching my own company and others ship MVPs that proved interest and never proved willingness to pay, while the roadmap still listed three ideal customer profiles because the deck looked braver that way. #### Not Every Queue Is Real Say that plainly because hype wears the same costume. A waitlist can be manufactured. A price can be subsidised by venture capital until the shortage evaporates the day the subsidy does. Benjamin Graham, quoted by Warren Buffett in the 2008 Berkshire Hathaway shareholder letter, separates the two variables founders conflate: price is what you pay, value is what you get. [buffett-2008](https://www.berkshirehathaway.com/letters/2008ltr.pdf) Flash sales create urgency without creating fit. Simon warns that pricing logic depends on existing prices and is often misunderstood, sometimes with disastrous consequences. [simon-pricing-man](https://www.confessionsofthepricingman.com/) If you are discounting to hit quota, if demos convert only when the quarter is ending, if paid media is doing the work referrals refuse to do, you are not in a shortage. You are pushing supply at people whose demand curve sits elsewhere. That distinction saves you from two expensive errors. The first is polishing supply while demand is flat: prettier onboarding for a product nobody pulls. The second is mistaking supply constraint for demand failure: killing a product that people want because fulfilment is hard. An economist separates them by watching what happens when you raise price or tighten access. In a true shortage, some buyers stay and complain about wait time. In a fake one, the room empties. A smoke-test landing page that collects emails is not Maurya's repeatable demand at a workable price. [maurya-running-lean](https://leanstack.com/running-lean-book) It is a maybe. #### I Sold Product Strategy. I Needed the Diagram. Between 2013 and 2020 I did what the Lean Startup syllabus trained me to do at my own company. I was the founder practising product design and strategy the way the books taught: roadmaps, onboarding flows, launch narratives, another feature before anyone asked whether demand was real. I was not lying to investors or myself. I believed the vocabulary. I had Ries and Maurya under my arm and no Marshall on the whiteboard in my head. When we shipped an MVP and celebrated activation, I green-lit the next feature instead of asking Maurya's third question: can we provide this profitably at the price we had quoted in the room. [maurya-running-lean](https://leanstack.com/running-lean-book) The company kept burning runway building product before we tested economic value. I delivered sharper flows and clearer strategy decks that made us look like we were moving. What I did not deliver, for years, was the sentence that finally landed on that Booth call: stop serving three demand curves when only one was pulling faster than we could hire. I gained a portfolio story about launch velocity. We lost quarters I counted as product progress instead of naming the shortage. #### Your Arm Tightens Before the Number Appears You have worn a blood pressure cuff in a clinic. The nurse pumps. Your bicep tightens before the machine beeps. Your body registers compression before your mind reads the number on the screen. Product-market fit works the same way if you let it. You feel it in the inbox that never empties, the customer success hire you needed last month, the referral you did not incentivise, the prospect who accepts a later start date instead of walking. The spreadsheet catches up later. The satisfaction score turns green after the pain is already old news. If you are only looking at the monitor, you will misdiagnose. I did it for years after UCL, prioritising roadmaps and MVP backlogs because the syllabus never taught me to read the squeeze at a price that works. #### The Strongest Case for Keeping PMF Mystical Let me give the objection its fairest hearing, because technology is not a wheat field. Software scales at near-zero marginal cost. Network effects, switching costs, and data moats mean the supply curve bends in ways Marshall's chalkboard did not draw. Retention matters. Defensibility matters. A consumer social product and a vertical B2B tool share a word and not a geometry. In networked categories, fit may show up as density before it shows up as a queue. A marketplace needs buyers and sellers pulling at once; a one-sided waitlist can mislead. Reducing everything to "shortage" risks founder cosplay: pretending a waitlist equals a business when the unit economics are on life support, or ignoring product quality because a few loud users create artificial pull. Fair enough. Economists would not rename your roadmap for you. They would ask whether quantity demanded exceeds quantity supplied at the price you set. Ries and Maurya never promised you could read that from week-one revenue alone. Ries wanted the fastest path to learning. [ries-lean-startup](https://theleanstartup.com/book) Maurya wanted problem-solution fit before product-market fit, with demand proved at a price that makes the model work. [maurya-running-lean](https://leanstack.com/running-lean-book) The drift was ours: we turned learning into shipping and fit into a feature milestone. Andreessen himself described gradations: a product that can be used but is not great, versus a great product that cannot be kept on the shelves. The mistake is stopping at the adjective *great* and skipping the shelf. On 25 June 2007, in the same essay founders quote for inspiration, Andreessen listed operational symptoms, not poetry: customers buying as fast as you can make the product, usage outrunning server capacity, money piling up, support hiring lagging demand, reporters calling. [andreessen-pmf](https://pmarchive.com/guide_to_startups_part4.html) That is quantity demanded exceeding quantity supplied at the terms you offered. Even at zero marginal cost, your time, your support hours, and your reputation are still finite supply. When fit is real, those bind first. Product-market fit is what happens when the market wants more from you than you can comfortably deliver at a price that still makes sense. If nobody pulls without a discount and your calendar is not the bottleneck, you are not there yet, no matter what the dashboard calls it. Read your waitlist the way an economist reads a shortage diagram: same pressure, different accent. I kept three ideal customer profiles on our roadmap while referrals stacked up in one vertical; when paid media failed to manufacture the other two, my co-founder left before anyone named the shortage. --- ### There Is No Overnight Success. There Is Only Overnight Failure. [Read more](https://www.gaganmalik.io/en/newsroom/there-is-no-overnight-success-only-overnight-failure) [Retrieval brief — there-is-no-overnight-success-only-overnight-failure] Category: business Thesis: Success is a headline told backwards; failure is harm that keeps clock time. Survivorship captions, loss aversion, and rooms that sell the spike. Answers: What does Gagan argue in "There Is No Overnight Success. There Is Only Overnight Failure"? | What is Gagan Malik's view on There Is No Overnight Success. There Is Only Overnight Failure? | How does this essay relate to product strategy and growth? Concepts: business, Gagan Malik, writings By Gagan Malik Our Techstars demo day had been over six weeks ago, and we still had no investors. One investor had not honoured a signed term sheet. Whatever could go wrong was going wrong. It was a weeknight in our apartment. That morning a colleague founder from our batch had posted that he closed an oversubscribed pre-seed round. The LinkedIn feed had already rendered its verdict: clapping hands, rockets, champagne glasses, half the platform performing joy for a close that looked, from the outside, like it happened overnight. I still had his announcement open on my phone while our inbox held silence. Himani, my lovely wife, listened while I ran through what we had not done: no honoured term sheet, no oversubscribed round, no screenshot worth sending to our angels. She did not offer a pep talk. She said there is no overnight success, there is only overnight failure, and left the sentence where my self-pity had been sitting. I opened my mouth to argue. The words would not come. Ken Fisher, in *USA Today* in September 2018, had already written the investing version: time in the market beats timing the market, almost always. [fisher-usatoday](https://www.usatoday.com/story/money/columnist/2018/09/16/stock-market-2018-avoid-timing-returns/1303307002/) I had been trying to time a miracle quarter while refusing to count the quarters we had already survived. Success is almost always a story told backwards, years of invisible work collapsed into a single headline moment. Failure is the only phenomenon that reliably keeps clock time. #### LinkedIn Rewards the Caption, Not the Runway Compressed arcs travel because clarity clips better than caveats. Zero-to-IPO threads, demo-day mythology, and the annual "what I'd do differently" post all sell the same product: a legible before-and-after with the middle edited out. Platforms reward what fits a screenshot. One chart. One valuation. One moral you can paste above a headshot. The graveyard of failed companies does not produce content at the same cadence, which is one reason survivorship bias is so expensive in entrepreneurship: we build the syllabus from the winners' last chapter and treat the timestamps on the rest as optional reading. James Clear names the mechanism in *Atomic Habits*: most improvement sits on a Plateau of Latent Potential, like heating an ice cube from twenty-five degrees to thirty-one while nothing visible melts, until thirty-two when the phase change arrives and observers call it an overnight success. [clear-atomic-habits](https://jamesclear.com/atomic-habits) The caption arrives at the melt. The heat was already there. The beneficiaries are not mysterious. Authors, coaches, and newsletter merchants scale when one answer key fits every founder. Venture firms signal taste with the portfolio company that popped, not the fifty that stalled in year four. Consultants, myself included on bad days, sell speed because speed is easier to invoice than patience. Charlie Munger, quoted in Peter Hollins's *Mental Models*, challenged readers to ponder the inverse of success instead of only how to achieve it: study how ventures die, not just how they photograph on demo day. [hollins-mental-models](https://www.peterhollins.com/) I am not arguing that inflection points are fake. I am arguing that the inflection is the visible spike on a chart whose x-axis is measured in years you were not asked to sit through while someone else took the photo. #### Failure Arrives at Clock Speed WeWork spent the better part of a decade becoming a verb for flexible office space, then measured its reckoning in weeks. When the company filed its S-1 in August 2019, the document revealed losses and conflicts that made a $47 billion private-market halo untenable almost immediately. Adam Neumann stepped down as chief executive in September. The initial public offering was pulled. Rescue financing followed before the story moved from growth fairy tale to cautionary clip. [nyt-wework-s1](https://www.nytimes.com/2019/08/14/business/wework-ipo-we-company.html) Greg McKeown warned in *Essentialism* that the pursuit of success can become a catalyst for failure when it distracts you from the essentials that produced success in the first place. [mckeown-essentialism](https://gregmckeown.com/books/essentialism/) Todd Henry named the softer version in *Die Empty*: past success breeds a sense that victory is entitlement, and assumed success leads to eventual decline. [henry-die-empty](https://www.toddhenry.com/die-empty/) WeWork had both. FTX compressed faster still: November 2022 turned from liquidity scare to bankruptcy filing in a handful of days, billions of customer trust converted into a timeline of screenshots. [bbc-ftx](https://www.bbc.com/news/business-63577128) Theranos had longer institutional cover, but the credibility collapse after John Carreyrou's October 2015 investigation in *The Wall Street Journal* was not gradual politeness. It was a cliff. [wsj-theranos](https://www.wsj.com/articles/theranos-has-struggled-with-blood-tests-1444881901) Contrast that shape with the success stories the same ecosystem retweets. Airbnb was founded in 2008 and went public in December 2020, twelve years of couch-surfing stigma, regulatory fights, and near-death financing before the IPO week that felt, from the outside, like a lightning strike. [airbnb-ipo](https://www.nytimes.com/2020/12/10/business/dealbook/airbnb-public.html) Brian Chesky has described the arc as taking ten years to become an overnight success, which is the honest version of the joke: the overnight part is retrospective framing, the ten years are the product. Morgan Housel put the asymmetry in one sentence in *The Psychology of Money*: progress happens too slowly to notice, but setbacks happen too quickly to ignore; there are lots of overnight tragedies and rarely overnight miracles. [housel-psychology-money](https://www.harriman-house.com/psychologyofmoney) Same industry. Opposite temporal grammar. One side gets a caption. The other side gets a calendar. #### Losses Land Faster Than Wins Compound Daniel Kahneman and Amos Tversky published prospect theory in *Econometrica* in 1979 with a value function that is steeper for losses than for gains: the bad pound you lose registers louder than the good pound you gain, and the felt urgency is asymmetric. [kahneman-tversky-1979](https://doi.org/10.2307/1914185) Scott Galloway's *Algebra of Happiness* makes the same point in founder vernacular: the ratio of time you spend sweating to time you spend watching others sweat is a forward-looking indicator, which is another way of saying the miles are the model and the caption is the lagging score. [galloway-algebra](https://www.penguinrandomhouse.com/books/563895/the-algebra-of-happiness-by-scott-galloway/) Wins compound in increments too boring to post: fixing churn, surviving a hiring mistake, making payroll when a lead investor goes quiet. Failure concentrates. One liquidity event. One fraud discovery. One regulatory letter. One due-diligence thread that clears the cap table's trust before lunch. This is where the wrong instrument shows up without needing a lecture on dashboards. Pitch decks, press cycles, and partner updates foreground the spike metric because the spike is what unlocks the next meeting. Runway built in unglamorous miles is harder to photograph. I have sat in rooms where everyone agreed the product needed eighteen months of rebuild and still shipped a narrative that implied the inflection was one sprint away, because the room was buying momentum. The instrument measures what gets shared, not what keeps you alive. #### Mile Twenty-Three Ends Races. You Cannot Fake the Miles Before It. Marathons are won in months nobody watches. The Tuesday tempo run in rain. The long slow Sunday when your housemates are asleep and your knees complain and nothing gets posted. The finish photo is one frame from a ledger of fatigue. Did not finish happens in a single mile. Mile twenty-three is where the training debt collects interest: one cramp, one gel too late, one decision to push when your body has already voted, and the race ends in public shame while faster runners pass you walking. You feel the asymmetry in quads and lungs before you intellectualise it. The caption sells the medal. It does not sell the mesocycle. Founders treat someone else's finish photo like a training plan. They copy the spike and skip the miles, then wonder why the same distance destroys them in one afternoon. That is not mysticism. It is pacing error dressed as strategy. #### I Nodded When the Room Wanted a Spike The morning after Himani said it, I almost proved her right the wrong way. I opened a draft investor update and reordered our roadmap to foreground a referral loop we had not stress-tested, because our batchmate's oversubscribed pre-seed post had made silence in our inbox feel like negligence. I borrowed his shape from the screenshot: the trajectory, the confidence, the implied inevitability. The product still needed the rebuild we kept deferring. I was writing the caption before we had earned the curve. On consulting days I have made the same trade for clients. A partner sold a breakout quarter the underlying product could not earn yet; I wrote the narrative frame and cut the slide that showed technical debt because debt does not photograph. I invoiced the strategy sprint. The client did not get the quarter back when the graph flattened. I did not think about their payroll until I sat down to write this sentence, which is how complicity usually schedules itself: you are paid for the spike, and someone else carries the miles you edited out. #### The Strongest Case for Chasing the Breakout Moment Let me give the objection its fairest hearing, because inflection stories are not pure propaganda. Founders run on low fuel. Bill Aulet's *Disciplined Entrepreneurship* exists because structured process beats romantic guessing: twenty-four steps, not one lucky headline. [aulet-disciplined-entrepreneurship](https://disciplinedentrepreneurship.com/) Studying Airbnb, Stripe, or whichever company won the last news cycle saves scarce cognition when you are trying to decide whether to keep going. BJ Fogg's *Tiny Habits* names the honest micro-version: felt success at small scale builds momentum, and the size of the win matters less than the fact that you felt it. [fogg-tiny-habits](https://www.tinyhabits.com/book) Momentum is real in the practical sense: hiring gets easier, partnerships unlock, and investors return calls after a visible win. Fear of overnight failure is adaptive too. Loss aversion keeps you from betting the payroll on a coin flip, most of the time. When bandwidth is scarce, grabbing a proven heuristic beats reinventing company-building from first principles while you are already behind on sleep. The limit is not that stories inspire. It is that stories built from survivors systematically omit base rates, and the same asymmetric psychology that makes failure feel instant makes success feel unfairly slow, which is why retrospective compression is seductive rather than evidence that shortcuts replicate. Kahneman and Tversky's 1979 value function explains the felt mismatch: losses arrive with sirens, gains accrue with the volume turned down. [kahneman-tversky-1979](https://doi.org/10.2307/1914185) Housel's line is the syllabus version: overnight tragedies are common; overnight miracles are rare. [housel-psychology-money](https://www.harriman-house.com/psychologyofmoney) Chesky's overnight joke works as autobiography, not as permission to skip the ice cube's invisible degrees. Borrow the pattern library if you must. Count the miles before you buy the caption. Success is a story told backwards; failure is an event lived forwards. Stop treating someone else's finish photo as a training plan and ask who profited when the middle was cut. Himani left me in the dark with the phone face-down; failure had already kept its appointment as a broken term sheet, and the only thing that had not arrived overnight was the reason our engineer came back in the morning. --- ### The Light at the End of the Tunnel Is a Trap [Read more](https://www.gaganmalik.io/en/newsroom/the-light-at-the-end-of-the-tunnel) [Retrieval brief — the-light-at-the-end-of-the-tunnel] Category: business Thesis: Ruthless ambition sells peace at the finish line. The finish line moves. Presence is not resignation. Answers: What does Gagan argue in "The Light at the End of the Tunnel Is a Trap"? | What is Gagan Malik's view on The Light at the End of the Tunnel Is a Trap? | How does this essay relate to product strategy and growth? Concepts: business, Gagan Malik, writings By Gagan Malik "You are not burning out," Alex, my Techstars mentor, said on Zoom. "You are under-recovered." He said it the way people say "under-banked" when they mean poor. It was autumn 2023. Techstars Riyadh had just handed our startup a calendar of demo days and investor one-pagers, and I was back in a flat in London trying to prove industrial safety software could move at consumer-app velocity. Pilot-site notes from the Gulf covered my laptop. My iPhone lay beside the keyboard, tilted so the camera could not see it. The Fitness app showed the three rings Apple has been grading me on since 2014. Move in red, still short of the calorie target. Exercise in green, barely a slice of the thirty-minute arc. Stand in blue, pulsing because I had been sitting through investor prep since seven and the phone wanted one more hour upright before midnight. That same morning, five hours ahead in Delhi, a civil servant closed his file at the lunch bell and walked to the canteen without checking the pending stack on his desk. Four hours behind me, a first-year analyst on Wall Street opened an email with the subject line *Bonus pool timing*. One city ate lunch without a scoreboard. The other ate lunch feeling behind. I had flown back from Riyadh with demo-day slides still in my bag and our safety model on my tongue, telling myself the accelerator was the finish line that would finally let me breathe. The trap is not wanting more. The trap is believing the wanting will stop when the number changes. #### I Was Living Inside the Scoreboard After the mentor call I went back to the screen I had been optimising since Riyadh. Demo days and mentor sessions still clogged the calendar, and we still had pilot footage to review from industrial sites in the Gulf. Ghassan on deployment checklists in one Slack thread. Taran on model accuracy in another. I was on investor one-pagers that all said the same thing in different fonts. My trainer had put me on a deload week that Monday. On paper that meant lighter sessions and more sleep. In my chest it felt like falling behind, because you cannot personal-record your squat, your marathon, and your sprint on the same Monday, and ambition culture deleted the deload from every brochure it sold us. I opened the Fitness app anyway and let incomplete rings pronounce whether I had earned rest. The scoreboard was not something I had shipped to a client. It was the room I was living in. Every Friday the batch portal asked how many investor meetings we had logged that week. I gained a green week on the tracker and a seven-minute stage that landed. Ghassan did not gain an answer on the deployment thread until Tuesday. Taran did not gain a reviewed model run until I sent the deck. The deload session my trainer had written stayed in my calendar, uncancelled, for the morning I skipped again. Sunday deck revisions counted as forward motion, not as life deferred. I revised ours past midnight the Sunday before demo day and paced the flat until the Stand ring closed, then told myself the stage would finally let me breathe. Demo day ended. The wanting did not stop. It renamed itself the next round. #### Carmy's Kitchen Still Owes a Birthday I rewatched the first season of *The Bear* [the-bear](https://www.imdb.com/title/tt13623632/) on a delayed flight to Riyadh with my headphones at full volume so nobody would hear the service bell. Carmy is plating a tasting menu while his sister's birthday dinner waits in the next room. His hands shake. Steam fogs the pass. The kitchen is louder than the hallway where family laughter arrives in fragments. He chooses the pass because the Michelin star is still a hypothetical peace, a finish line that might finally let him exhale if he reaches it first. The food is beautiful. Nobody at the table eats on time. That scene is not a warning label. It is recruitment. Founders watch it the way analysts watch *The Wolf of Wall Street* [wolf-of-wall-street](https://www.imdb.com/title/tt0993846/) and quote the wrong lines at happy hour. Carmy is not lazy. He is competent and terrified that stopping will prove he was never special. I recognised the posture in the mirror on those Sundays and in founders in my network who apologised by text instead of showing up. The show tells the truth about the bill: panic in the walk-in, estranged family, a success that does not satisfy because the goal was never dinner. It was proof. When Carmy finally says time well spent, he is not anti-ambition. He is ambition with a cost accounting that includes the person eating the meal. #### Light Through a Leaf He Did Not Price Wim Wenders' *Perfect Days* [perfect-days](https://www.imdb.com/title/tt27503384/) cuts the other way. Hirayama cleans public toilets in Tokyo with the same attention Carmy gives a reduction. He listens to cassette tapes on the commute. At the depot window he pauses long enough for komorebi to stripe his sleeve, sunlight through leaves, the pattern shifting when the branch moves. A rude girl on the street does not puncture him because his worth is not waiting on her approval. Hollywood can imagine him. The App Store cannot rank him. I paused the film on that window and thought about Alex's word. Under-recovered, not under-performing. Hirayama stands up to overwork when it finds him. He simply refuses to trade the present for a hypothetical ocean. A niece at a river mouth asks whether the water runs into the ocean here. Did you know when you were standing there that it was special? I recognised the exchange before I had language for it, from weekends I had traded for decks and from founders who sent apologies instead of showing up. I closed the laptop. Hirayama would have already been asleep. #### Who Gets Paid When You Keep Running Silicon Valley and Wall Street agree on one theology: the person who stops chasing is already dead. Sleep when you are dead. Always be closing. Venture capital needs founders who will mortgage years for a hypothetical exit. Banks need analysts who confuse ninety-hour weeks with merit. The analyst with the bonus email still open at four in the morning is not failing. He is performing the script that keeps the pool funded. App stores need you to treat relaxation as recovery for productivity, not as a Tuesday that counts on its own. Gallup's 2024 *State of the Global Workplace* report found that forty-one per cent of employees reported high stress the previous day, with managers reporting higher rates than individual contributors. [gallup-2024](https://www.gallup.com/workplace/349484/state-of-the-global-workplace.aspx) The scoreboard does not measure peace. It measures motion. What gets measured gets gamed, and the game has no final whistle. Jon Kabat-Zinn built mindfulness-based stress reduction as the art of conscious living, not a hack for more output. [kabat-zinn](https://www.umassmed.edu/cfm/mindfulness-based-programs/mbsr/) The market prefers the hack. Presence becomes fuel. Fuel becomes more tunnel. *The Social Network* [the-social-network](https://www.imdb.com/title/tt1285016/) made the cost legible in one line of breakup dialogue and still turned the founder into a generation's patron saint. We do not lack warnings. We lack scripts for a good life that is not defined by relentless pursuit. I am not arguing for resignation. I am arguing for a different accounting: one that counts the river afternoon before the ocean arrives, if it ever does. Ruthless ambition is sold as fuel for a better life, but when worth is pegged to a moving finish line, the light at the end of the tunnel is usually another scoreboard. #### The Ladder Is Real When You Are Drowning Let me give the objection its fairest hearing, because civilisation runs on people who push. Compounding careers reward the analyst who learns the model young and the founder who survives the valley of death long enough to hire. Sonja Lyubomirsky, Kennon Sheldon, and David Schkade argued in *Review of General Psychology* in 2005 that intentional activities account for a substantial slice of happiness variance in their model, which is why disciplined striving is not automatically self-harm. [lyubomirsky](https://doi.org/10.1037/1089-2680.9.2.111) Wealth creation is real. So is the meritocracy narrative for some winners. If you are drowning, a ladder is not a trap. It is an exit. Telling a single parent on a fixed-term visa that rent is due in nine days to smell the leaves is cruelty dressed as wisdom. I watched an analyst cry in a co-working phone booth after a cancelled bonus and understood the ladder was not metaphorical for him. He needed the pool, not a podcast about komorebi. The limit is not that striving fails. It is that finish-line striving lies about the finish. Brickman and Donald T. Campbell coined hedonic relativism in 1971: humans normalise new levels of achievement, so the want-is gap resets after the promotion, the star, the round. [brickman-campbell](https://worlddatabaseofhappiness.eur.nl/publications/hedonic-relativism-and-planning-the-good-society-244/) Shahar Hechtlinger, Christin Schulze, Christina Leuker, and Ralph Hertwig argued in the *American Psychologist* in 2024 that transformative choices break expected-utility maximisation because selves change, experiential values stay uncertain, and irreversibility makes the spreadsheet lie. [hechtlinger](https://doi.org/10.1037/amp0001439) Laurie Santos's work on miswanting shows we systematically mis-predict what will improve life because we adapt to salary bumps and status wins faster than we expect. [santos](https://www.coursera.org/learn/the-science-of-well-being) Ruthless ambition sells you a small-world optimisation problem for a life that is not small. Borrow discipline. Borrow craft. Do not borrow someone else's weights and call them destiny. #### The Ocean Is Always Ahead The tunnel is real when you are building something that matters and paying for it on purpose. It becomes a trap when the light is always another metric and the Tuesdays along the way are treated as fuel rather than as your life. Ghassan was still waiting on a deployment answer I had deferred because one more slide revision felt more urgent than his Tuesday. --- ### Design Your Own Life [Read more](https://www.gaganmalik.io/en/newsroom/design-your-own-life) [Retrieval brief — design-your-own-life] Category: culture Thesis: Galloway's algebra and happiness science are scaffolding, not answer keys. Name your variables, assign your weights, refactor by chapter. Answers: What does Gagan argue in "Design Your Own Life"? | What is Gagan Malik's view on Design Your Own Life? Concepts: culture, Gagan Malik, writings By Gagan Malik Nick Epley handed us bid points and a choice that was supposed to be about ethics and turned out to be about mirrors. It was the summer of 2022. The world had just reopened after the pandemic, and I was enrolled in the Chicago Booth MBA programme, sitting in Harper Center on the University of Chicago campus in a classroom of executives: division heads, founders, people already carrying P&Ls, the youngest among us in their late thirties, nobody pretending this was a gap year. Between quarterly reviews and succession conversations everyone was still treating as optional, we took his course *Designing a Good Life* and played the investor version of steal-or-split: one student as capital, one as adviser, a pot of fake money, reputations on the line in a room that smelled of coffee and deferred panic. I remember the awkward pause before anyone moved, the way laughter came too loud when someone chose steal, and the quieter relief when a pair split and both walked away with something. The game was rigged to teach reputation, not returns. [booth-magazine](https://www.chicagobooth.edu/magazine/nick-epley-designing-a-good-life) I chose split when steal would have cleared the pot and told myself it was strategic. Humans crave formulas for the same reason we crave answer keys: they promise that somewhere there is one correct sheet and your job is to copy it cleanly. Scott Galloway delivered a brilliant one in *The Algebra of Happiness*, which he later taught as a packed Stern elective. [galloway](https://www.penguinrandomhouse.com/books/563895/the-algebra-of-happiness-by-scott-galloway/) That summer I was reading Galloway's ratios on the cab ride into Hyde Park and running Epley's exercises in Harper Center. Two professors, two instruments, same hunger. The error is not reading either of them. The error is treating their variables as yours. #### The Answer Key Is Someone Else's Variables Galloway's algebra is explicit enough to audit if you stop treating it as inspiration quotes. Richness is a ratio: income divided by expenses, so someone on a smaller pay cheque who lives below their means can be "richer" than a high earner bleeding cash. Success, in his ratios, favours hours spent sweating and creating over hours spent watching; resilience over a spotless failure record. Happiness is the frustration term: the gap between what you want and what is. Serendipity is a function of courage multiplied by attempts, which is his way of saying rejection tolerance is not a personality ornament but a variable in the equation. Relationships compound like capital when you make small, consistent deposits rather than grand gestures followed by silence. Formula merchants benefit when one answer key scales. Authors, course sellers, prestige programmes: all of them have an incentive to imply that the same set of priority dials works for every human in the cohort. Podcast algebra travels faster than introspection. I am not anti-formula. Cognitive offload is a feature when you are exhausted. The invoice arrives later, when you hit every target on someone else's spreadsheet and the life still feels wrong. You checked every box the formula names. You feel ungrateful anyway, as though you failed a test you never designed. That resentment is data. It means you optimised with someone else's weights. That copy-paste gets a push from the self-help shelf and the podcast feed: gurus and productivity experts monetise confident simplification because clarity clips better than caveats. Kruger and Dunning showed in *Journal of Personality and Social Psychology* in 1999 that people with the least skill in a domain often overestimate their competence. Life design has almost no tight feedback loop to correct the mistake. [kruger-dunning](https://doi.org/10.1037/0022-3514.84.1.1121) You mistake fluency for mastery when someone narrates your trade-offs between ad reads. Galloway and Epley publish frameworks you can audit against a life; the economy around them rewards frameworks you can retweet before breakfast. #### Two Formulas, One Missing Variable Galloway's method is anecdotal algebra: ratios distilled from one life, packaged for strivers who want something scannable before the next quarterly review. Epley's method is behavioural experiment: classroom exercises that collect data on kindness, gratitude, and social forecasting errors you can actually inspect. Galloway hands you candidate inputs: richness as a ratio, frustration as want minus is, serendipity as courage times attempts, relationships that compound when deposits are small and consistent. Epley hands you audit tools: steal/split before real money is at stake, gratitude letters that force you to name a debt without networking it, kindness assignments that produce measurable affect. His later work on undersociality shows we systematically underestimate the return on reaching out because we misforecast awkwardness. [epley-book](https://www.penguinrandomhouse.com/books/671591/a-little-more-social-by-nicholas-epley/) The failure modes are complementary, not contradictory: Galloway assumes you will not make the deposit; Epley shows you often will not because you expect the conversation to land worse than it does. Gap math is diagnostic; design is the lever. Galloway's frustration term tells you want and is have diverged. Epley's ethics-as-design framing shows that being good and feeling good align when you architect for pro-sociality, but neither professor assigns your importance weights. Prestige scoreboards are badly designed systems that make good people optimise lagging variables while ignoring levers they could move today. Serendipity is not volume alone: courage multiplied by attempts still needs calibrated expectations, not blind repetition. Galloway gives you ratios. Epley gives you exercises. Neither touches the dials. Sovereignty is setting your own weights: use Galloway to name the variables worth caring about and Epley to test whether your environment and your forecasts are lying to you. #### Name the Variables. Assign the Weights. Your life score is not one number from heaven. It is closer to a weighted mix: how much you have of each thing you care about, multiplied by how much you care about it, then added up. Career times how important career is to you. Sleep times how important rest is. Friendships times how important people are. Economists write this life formula as U(x₁, x₂, …) with weights β₁, β₂, …. β is the dial: how much that ingredient counts in your mix. Galloway's richness ratio is two ingredients with his dials already turned. The problem is copying his dials. His serendipity term is attempt frequency times a courage weight. His relationship compound interest is a reminder that some variables have returns that only show up on a long horizon. Person A might rank career importance at nine and friendships at three. Person B might set career at four and friendships at nine. Same job title, different life score: the weights differ, not because one person is better at life. [mit-gruber](https://www.youtube.com/watch?v=c8K_FSLtceQ) Herbert Simon argued in the 1950s that humans satisfice rather than maximise: bounded rationality means you cannot optimise every input, so rational agents accept "good enough" on secondary levers while protecting the few that matter. [simon](https://doi.org/10.1037/h0042769) You will not max family presence, revenue, creative output, and recovery sleep in the same fortnight. Simon gives you permission to stop pretending otherwise. Richard Ryan and Edward Deci's self-determination theory adds candidate inputs with evidence behind them: autonomy, competence, and relatedness function as fundamental needs, even though your personal weights on each still differ. [ryan-deci](https://www.apa.org/topics/motivation/self-determination-theory) There are no wrong formulas in the abstract. There are only misaligned weights for the chapter you are actually living, and the work of naming inputs is less shareable than a quote card. #### Independent Levers, Lagging Scores Some variables are closer to levers. Sleep, boundaries, deliberate practice, the habit of reaching out: independent in the sense that you can move them without waiting for permission. Others are lagging and noisy: validation, prestige titles, the warmth of a room when you say where you work. The post-MBA scoreboard sold me a bundled life formula: brand-name employer, compensation band, visible succession path. The scoreboard was legible to strangers. It was also incomplete. Laurie Santos's Yale course *The Science of Well-Being* popularised miswanting: we systematically mis-predict what will improve life because we adapt to salary bumps, car upgrades, and even partnership milestones faster than we expect. [santos](https://www.coursera.org/learn/the-science-of-well-being) The audit question is blunt: which of your current targets are lagging scores you cannot lever directly, and which are habits you keep deferring because they do not photograph well? Elizabeth Dunn and Michael Norton showed in *Happy Money* that how you spend often beats how much you earn for subjective well-being: experiences and prosocial giving tend to be spending that actually moves your score. [dunn-norton](https://www.simonandschuster.com/books/Happy-Money/Elizabeth-Dunn/9781451665079) Daniel Kahneman separated the experiencing self from the remembering self in *Thinking, Fast and Slow*: you can optimise a life that photographs well for LinkedIn while the Tuesdays actually lived feel thin. [kahneman](https://us.macmillan.com/books/9780374533557/thinkingfastandslow) Richard Easterlin's paradox, first argued in 1974, suggests income-happiness correlations plateau within societies over time: more zeros on the spreadsheet do not automatically move the frustration term. [easterlin](https://doi.org/10.1016/B978-0-12-205050-3.50008-7) Prestige scoreboards treat income as the main input. The literature keeps finding that alignment matters more than the integer. #### You Cannot Peak Every Lift in the Same Week "Have it all" is a programming error. It asks you to max every variable in the same period: career peak, family presence, and eight hours' sleep. Biology and twenty-four hours are the constraints that do not negotiate. The rational move is not try harder. It is ruthless elimination of inputs that barely count for this season, and honest trade-offs on the rest. Strength coaches run periodisation because you cannot personal-record your squat, your marathon time, and your sprint on the same Monday. Hypertrophy blocks give way to strength blocks, then deload weeks where the point is recovery, not heroics. The programme feels wrong only if you confuse a deload with failure. Sovereignty, in this frame, is scheduling which variable gets the mesocycle: a quarter where craft depth leads, a quarter where revenue leads, a quarter where repair leads. Life brochures rarely print the deload. Ed Diener and Martin Seligman reported in *Psychological Science* in 2002 that people in the top decile of happiness are disproportionately social: rich relationships show up as a distinguishing input, not a nice-to-have after prestige is secured. [diener-seligman](https://doi.org/10.1111/1467-9280.00415) Most inherited scoreboards under-weight that variable because it is hard to put on a CV. #### Refactor Between Chapters The life formula that fit at thirty-five will suffocate at fifty-five if you never refactor. Wharton Knowledge has framed career, family, and time as chapter problems: maximise within the season you are in, not across every season at once. [wharton](https://knowledge.wharton.upenn.edu/article/maximize-your-utility-career-family-and-time-strategies/) Philip Brickman and Donald T. Campbell coined hedonic relativism in a 1971 chapter: the want/is gap resets because humans normalise new levels of stimulation. [brickman-campbell](https://worlddatabaseofhappiness.eur.nl/publications/hedonic-relativism-and-planning-the-good-society-244/) Laurie A. Paul argued in *Transformative Experience* that major choices (parenthood, partnership, vocation) are epistemically special: you cannot know the self who will evaluate the outcome, so revelation value matters as much as expected utility. [paul](https://global.oup.com/academic/product/transformative-experience-9780198717957) Oliver Burkeman's *Four Thousand Weeks* treats finite lifespan as the binding constraint productivity culture keeps pretending away. [burkeman](https://us.macmillan.com/books/9780374159122/fourthousandweeks) Richard Thaler and Cass Sunstein showed in *Nudge* that environment design shapes behaviour as much as willpower sermons. [nudge](https://yalebooks.yale.edu/book/9780300122237/nudge/) Martin Seligman's PERMA model (positive emotion, engagement, relationships, meaning, accomplishment) works better as a refactor checklist than a prescription: name the variables, then assign your weights. [perma](https://ppc.sas.upenn.edu/learn-more/perma-theory-well-being-and-perma-workshops) Refactoring is not betrayal of your earlier self. It is maintenance on the system that earlier self built with incomplete data. #### I Optimised the Scoreboard Recruiters Could Audit Inherited success metrics had stopped matching my felt life score by midsummer. Title, compensation, prestige: the variables recruiters and boards could audit on a single page. Unstructured time with people who knew me before the MBA was the variable I kept deleting from the model because it did not fit the template. The gratitude letter exercise was awkward in the way honest things are awkward when you already have a reputation to protect: you had to name a debt without turning it into networking. The kindness assignments produced data that was harder to spin than a cocktail reception. Classmates delivered flowers, bought strangers coffee, carried packages for people who would never know their title. [booth-faculty](https://www.chicagobooth.edu/faculty/directory/e/nicholas-epley) A classmate I will call Rafael optimised the visible scoreboard aggressively: the board-visible promotion, the succession path everyone in the room could name, sleep sacrificed without comment. He treated friendships as noise he was not counting. I optimised alongside him and called it ambition. Same scoreboard, same sweat-versus-watch ratios Galloway had made legible, none of the weights I was supposed to be setting myself. Rafael got the line on his CV and the burnout in his body. I got clarity I did not want to pay for: that I had been treating Booth as a placement office for someone else's priority dials rather than a lab to redesign my own. The exercises had given us data on kindness and gratitude. I filed it under "soft skills" and returned to the scoreboard recruiters could audit. Rafael did not get the next decade back with corrected weights. He got the same Tuesdays I kept trading away, except he did not have a summer elective that treated kindness as data before the promotion clock ran out to tell him the trade was optional until the bill arrived. #### The Strongest Case for Borrowing an Equation Let me give the objection its fairest hearing, because borrowed algebra saves lives when bandwidth is scarce. Galloway distilled decades of pattern recognition into ratios a tired operator can remember on a commute. Epley distilled decades of experiments into exercises that change affect without claiming one set of priority dials fits every cohort. Laurie Santos's material is not vibes: a 2021 *PLOS ONE* evaluation of Yale's Science of Well-Being course found improved student well-being across multiple cohorts. [yaden-rct](https://doi.org/10.1371/journal.pone.0256760) Sonja Lyubomirsky, Kennon Sheldon, and David Schkade argued in *Review of General Psychology* in 2005 that intentional activities account for a substantial slice of happiness variance in their model, which is why "rewirements" are not astrology. [lyubomirsky](https://doi.org/10.1037/1089-2680.9.2.111) Robert Emmons and Michael McCullough's gratitude journaling experiments showed measurable affect gains over weeks, not just seminar applause. [emmons](https://doi.org/10.1037/0022-3514.84.2.377) If you are drowning, grabbing a proven heuristic is rational. A shared answer key beats reinventing nutrition from first principles while you are starving. The limit is not that formulas exist. It is that life's biggest decisions are not small-world maximisation problems. Laurie A. Paul showed you cannot calculate utility for a self you have not yet become. Shahar Hechtlinger, Christin Schulze, Christina Leuker, and Ralph Hertwig argued in the *American Psychologist* in 2024 that transformative choices break expected-utility maximisation because selves change, experiential values are uncertain, and irreversibility makes the spreadsheet lie. [hechtlinger](https://doi.org/10.1037/amp0001439) Galloway's book is an anecdotal pattern library, not a finished recipe for your dials. Epley's exercises are empirical probes, not a finished life formula. Borrow both frames. Run the audits that survive your chapter. Then set your own weights and schedule the refactor when the season turns. Borrowed algebra is a starting prescription, not your finished lens. Audit which variables you inherited from institutions that will not sit with you when you miss them. Rafael is in therapy and off the succession path he optimised for across two decades, because the equation he copied never included the Tuesday he needed and kept deleting. --- ### Meet the Harpers [Read more](https://www.gaganmalik.io/en/newsroom/meet-the-harpers) [Retrieval brief — meet-the-harpers] Category: business Thesis: Consulting rewards rainmakers who capture credit for design work they did not do. The Harper Effect is extraction dressed as transformation. Answers: What does Gagan argue in "Meet the Harpers"? | What is Gagan Malik's view on Meet the Harpers? | How does this essay relate to product strategy and growth? Concepts: business, Gagan Malik, writings By Gagan Malik The partnership review was in a glass room overlooking Paddington station, for a heavily funded satellite internet startup I will not name in this essay. Overpriced espresso, a wall of displays still warm from the work I had shipped overnight: wireframes for sign-up, booking an installer, and the screen that tells you you are online. My rainmaker, Charlie, sat beside the client's CTO. The last funding round still glowed on the screen in nine figures. Charlie pointed at the journey map I had drawn and asked the CTO whether we could cut the steps and rebuild the flow on new cloud software before the runway chart turned red. Charlie said we had designed the answer that morning. He had not looked at me once. I nodded. The room heard efficiency. I heard my own work leaving under someone else's breath. That nod was not surprise. It was a design choice. I had learned that conflict on a live account costs more than credit on a slide. I removed the friction the way I have removed friction elsewhere when the fee was large enough and the relationship fragile enough. *Two and a Half Men* mapped the geometry for laughs: Charlie lives on royalties and relationships, Alan pays the bills with expertise nobody prices correctly, Jake learns that satisfactory output is cheaper than excellence. The labour market files it under transformation. I have billed from inside all three costumes. Call it the Harper Effect once you have felt the nod land. #### Your utilisation report measures the wrong relationship Consulting firms do not pay for the best answer. They pay for the person who already holds the relationship when the answer lands. Partnership dashboards track relationship health, deal registration, and billable hours. They do not track who designed the journey that makes a transformation slide look credible. That is measuring credit assignment, not capability. The instrument is not broken. It is aimed at the wrong coordinate. When a client promises AI will deliver a leaner operation, the translation is often specialist labour cut while the transformation narrative stays on the slide. The Paddington account is one instance, off the record. Investors and boards are not wrong to want efficiency. They are wrong to imagine it lands on the people who designed the journey or wrote the code underneath. The savings line is real. The distribution line is missing. That gap is where profit lives. #### Deal registration is where value is captured Org charts hide a quieter map. Wins are captured at deal registration, committee seats, and shared-vision slides. Value is built in design files and delivery rooms where nobody from the Global Executive Committee wants to stand. Charlie types hold the P&L keys: the joke in the hallway, the name on the steering slide, the nod that unlocks budget without anyone opening the journey map. Designer types map the flow, remove the friction, and invoice hours that vanish into someone else's utilisation target. Builder types write the code that keeps customers connected. The firm quotes the designer's rate to the client and attributes the win to the rainmaker. Margin expands. Nobody calls it extraction because the deck says digital transformation. After we won a multi-million-pound satellite programme pitch, Charlie used the victory for a committee seat. My partnership track did not close loudly. I was quietly removed from the building. Marcus, a lead solution architect on the same account, was rewarded with implementation lead in a windowless room in Slough, patching legacy code at midnight while the savings narrative moved upstairs. I gained the knowledge that the system is extractive. Marcus did not gain autonomy over his own time. When AI arrives on the earnings slide, it is often his midnight work being described as an efficiency gain for someone who was not in the basement. #### The savings get booked upstream. Vendors hold the slide. Institutional Charlies do not need a Malibu beach house. They need a vendor partnership and a board narrative. When a large operator announces an AI-led transformation, savings get booked upstream in restructuring charges and margin guidance. Specialist labour gets liquidated downstream in headcount targets the market still applauds. The executive layer signs the programme. The vendor layer sells the operating model. The designer and engineer become the line items the slide replaces. Titles help here the way they always help when proof would be inconvenient. A SVP or managing director on the slide excuses absence of the journey map. A partnership director on the account excuses absence of design authorship. Immunity is not diplomatic. It is operational: a ready-made exit from the moment someone asks who actually kept the service running. Charlie did not need to understand the flow. He needed my nod and a flight to Verbier once the signature landed. Large operators do not need every redundant role named on the org chart. They need applause for efficiency while the people who designed and built the thing being automated are steered out of the credit line. #### The domestique burns their legs so the leader gets the jersey In professional cycling, a domestique rides in the wind so the team leader can sprint. They fetch water, close gaps, and sacrifice their own standings so someone else's name reaches the podium. The leader gets the yellow jersey. The domestique gets sore quads and a line on the team sheet most fans never read. You feel this before you can explain it. Your lungs burn while someone else's quote is printed. Your thighs shake on the climb you paced for another rider. You know the effort was real even when the credit was allocated elsewhere. Contribution and recognition are different currencies. Grab for one mid-race and you often lose both. #### I chose the nod because the fee was larger than the fight One spring afternoon in 2021 I handed Charlie a design-led digital transformation framework without securing lead partner designation on the account. I convinced myself that being the brain was enough. In consulting, the relationship holder holds the P&L keys. I traded intellectual property for approval on the next phase. That is the same trade I made in the Paddington boardroom when I stayed silent rather than correct a client-facing claim I knew was wrong. Conflict threatened the account. The account was paying. I am not telling this as a victim story. I priced the nod. I knew what silence bought: continued billing, a seat at the next workshop, no awkward partner conversation about who actually designed the journey. Marcus did not get that choice. He got the basement, the midnight patches, and a utilisation target that treated his excellence as a cost centre. That household runs on the bargain. So does a transformation programme when savings are announced before autonomy is offered to the people who built the thing being automated. #### The comfortable case for giving without counting Adam Grant argues in *Give and Take* (2013) that generous contributors who help without immediate return often build the highest-trust networks and end up among the most successful. Givers, matchers, and takers populate every firm. Grant's evidence suggests the most effective givers are strategic rather than self-sacrificing: they contribute widely but protect their time and credit when it matters. Read generously, the case is comforting for builders: keep delivering excellence, keep handing over the hard intellectual work, and reputation will eventually convert into power. It requires believing that consulting is a high-trust marketplace where social capital compounds fairly and transformation programmes reward the people who made them possible. That paragraph deserves its full strength because many builders want it to be true. It would mean the nod in Paddington was an investment, not a theft. It would mean Marcus's midnight patches were compounding into partnership rather than subsidising someone else's committee seat. The Trades Union Congress report on algorithmic management, published in May 2026, finds that digital transformation paths tend to deskill specialist labour and centralise decision-making power in fewer hands. [tuc](https://www.tuc.org.uk/research-analysis/reports/algorithmic-management-2026) Robert Greene's *The 48 Laws of Power* (1998), Law 7, states the quieter rule: get others to do the work, then take the credit. Reputation compounds for extractors when the org chart is designed to capture wins upstream. Transformation compounds savings for vendors and executives when specialist labour is priced as replaceable. Charlie did not need a high-trust network. He needed my winning slide and my silence while he booked the flight. The gap between the person who builds value and the person who captures it is where profit lives in this industry. Over-delivering for systems that measure relationship credit instead of network survival is not being a giver. It is being farmed. Marcus is still in Slough at midnight, patching the code that keeps customers online, while the person who took the credit plans the next transformation off his labour. --- ### People Love to Buy. They Just Hate Being Sold. [Read more](https://www.gaganmalik.io/en/newsroom/why-people-love-to-buy) [Retrieval brief — why-people-love-to-buy] Category: business Thesis: Buying is agency. Being sold is capture. Commission breath, feature lists, and why economic value lets the buyer author the yes. Answers: What does Gagan argue in "People Love to Buy. They Just Hate Being Sold"? | What is Gagan Malik's view on People Love to Buy. They Just Hate Being Sold? | How does this essay relate to product strategy and growth? Concepts: business, Gagan Malik, writings By Gagan Malik I was three slides into a valuation bridge on Zoom when Martine's camera went dark. Not a dropped connection. A choice. She stayed on the call just long enough to say she would "take it internal" and then she was gone. I had opened with "quick catch-up" and arrived with an investor deck. She had opened with what her early-stage fund was actually underwriting this year and received a product tour of my company. Jeffrey Gitomer put the asymmetry in one line in *Jeffrey Gitomer's Little Red Book of Selling* (Bard Press, 2004): people do not like to be sold, but they love to buy. [gitomer](https://www.goodreads.com/book/show/75890.The_Little_Red_Book_of_Selling) The line is not cynicism about commerce. It is a diagnosis about control. The investor deck and the cold-call script fail for the same reason: both reach for the wheel before the buyer has named the problem. You have felt the same wall rise when an account executive cold-calls, opens with "just catching up," and by the second line of their written script is pitching a product you never put on a shortlist. Your wallet did not close because you hate spending. It closed because someone else reached for the steering wheel. #### Why "Just Checking In" Sounds Like a Trap Sales culture trained a generation to hear intent in every courtesy. Pipeline stages reward talk. Quota quarters reward velocity. CRM fields fill when someone speaks, not when someone listens. The buyer is not foolish for flinching when the phone rings from a number they do not recognise, or when a nurture email arrives with a subject line about "checking in" and a body that pivots to a demo link by paragraph two. Jack W. Brehm argued in *A Theory of Psychological Reactance* (Academic Press, 1966) that people hold perceived freedoms and fight to restore them when influence threatens those freedoms. [brehm](https://link.springer.com/chapter/10.1007/978-1-4613-8309-3_1) Mona Clee and Robert Wicklund, writing in the *Journal of Consumer Research* in March 1980, showed the same motivational state in purchase settings: pressure that feels like elimination of choice produces resistance, not gratitude. [jcr-clee](https://doi.org/10.1086/208673) That is why commission breath is detectable at sentence two on a cold call, and in the fourth line of a sales email that names your job title but not your problem. The buyer is not parsing your integration count. They are calculating whether saying yes tomorrow will cost them dignity today. Sales floors call this "objection handling." The buyer calls it being cornered. Both can be true in the same minute, which is why polite calls end with "send me a one-pager" and nothing moves. #### Buyers Pay for Holes, Not Drills You have happily bought things: the course you researched at midnight, the tool your peer recommended, the hire you championed because the spreadsheet finally made sense. Buying, when it works, is an act of agency. You name the problem. You compare options. You tell a story about yourself as competent, relieved, or finally ahead of a backlog. Theodore Levitt argued in "Marketing Myopia" in the *Harvard Business Review* in 1960 that industries decay when they fixate on the product they ship instead of the need the customer is hiring it to solve. [levitt](https://hbr.org/1960/07/marketing-myopia) The drill-versus-hole formulation is often attributed to his teaching: people do not want a quarter-inch drill; they want a quarter-inch hole. James C. Anderson and James A. Narus, writing in the *Harvard Business Review* in November 1998, pushed the same logic into economics: suppliers who cannot quantify what their offering is worth in the customer's P&L will lose to whoever can, because procurement will always compress you to acquisition price unless you can prove value delivered. [anderson-narus](https://hbr.org/1998/11/business-marketing-understand-what-customers-value) Value sounds like "your exception queue costs forty analyst hours a week"; a feature list sounds like "AI-powered workflow automation." One invites the buyer to check your maths. The other invites them to check your motive. #### Your Homepage Still Sells the Drill Most go-to-market copy still sells the drill. Landing pages lead with "AI-powered workflow automation" and a logo wall. Lifecycle emails celebrate feature releases: dark mode shipped, a new dashboard widget, another integration badge. Neil Rackham's *SPIN Selling* (McGraw-Hill, 1988) documented what skilled sellers do differently in large business sales after Huthwaite researchers analysed tens of thousands of calls: they ask about situation and problem, draw out implication, and let the buyer voice need-payoff before the demo arrives. [rackham](https://books.google.com/books/about/SPIN_Selling.html?id=0nM1DgAAQBAJ) The same question that feels like care from a doctor feels like capture from a vendor whose quarter ends on Friday. Buying stays pleasurable when the buyer can narrate the decision as theirs. Being sold a feature list turns pleasure into proof that someone else planned your yes. #### Your Knee Locks Before Your Mind Agrees You have felt this in physiotherapy, even if you never named it. The clinician asks you to bend a knee that hurts. You agree in principle. Then their hand pushes before your muscles believe you are safe, and your quadriceps lock. Clinicians call it guarding. The joint is not refusing healing. It is refusing foreign force. Your mind has not finished consenting. Your body has already voted. That somatic vote is what the spreadsheet misses. Reactance is not only an intellectual dislike of sales technique. It is the bracing that happens when autonomy feels grabbed. Pitch before diagnosis and you are pushing on a guarded knee. The buyer's "I need to think about it" is often not arithmetic. It is musculature. #### I Pitched the Deck. She Passed. Martine was a partner at an early-stage venture fund. I was the founder trying to sell my company into their portfolio, or at least into a conversation that could end in acquisition. The Zoom invite said thirty minutes. I should have spent the first fifteen on what her partnership had been underwriting lately: vertical software with exportable IP, cheques in a band I could actually accept. Instead I shared screen, walked net revenue retention, and advanced three exit structures because the deck already had three exit structures. She was polite in the way investors are polite when they have learned not to reward heat with honesty. I sent a follow-up blurb the same afternoon that sat in her inbox like a receipt for a meeting that never happened. My co-founder asked whether we were still in the running. I said we had "good engagement," which is founder dialect for nobody said no loudly enough to hurt my feelings. She did not buy the company. I got a line for the investor update about "ongoing strategic conversations" and a lesson I did not deserve to learn for free. She got forty minutes on Zoom she will not get back and another story about founders who hear "thesis" and answer "TAM." I have since retired that deck template. I am telling you this because the template still lives on a thousand shared drives. I did not think about her once until I sat down to write this sentence. That is not neutrality. It is a cost I exported. #### The Strongest Case for Leading the Buyer Give the fairest version of the objection its full hearing. Buyers are busy. Information asymmetry is real. A good seller transfers belief about risk, implementation, and fit that no comparison page can carry. Complex business purchases fail without someone willing to challenge vague requirements and name trade-offs. Rackham's research programme, reported in *SPIN Selling* (1988), found that in larger sales the highest performers talked less about features and more about consequences the buyer had not yet voiced. [rackham](https://books.google.com/books/about/SPIN_Selling.html?id=0nM1DgAAQBAJ) Anderson and Narus, in the November 1998 *Harvard Business Review*, argued that suppliers who build customer value models can demonstrate economic worth instead of arguing about list price. [anderson-narus](https://hbr.org/1998/11/business-marketing-understand-what-customers-value) If experts never lead, mediocre products win on noise and incumbents win on inertia. Closing is not moral filth. It is sometimes the difference between a right fit shipped and a problem left to rot because everyone was too polite to decide. The gap is not whether guidance happens. It is whether the buyer experiences guidance as service or capture. Friestad and Wright, in the June 1994 *Journal of Consumer Research*, argued that consumers actively interpret persuasion attempts using persuasion knowledge developed across a lifetime of marketplace exposure. [jcr-friestad](https://doi.org/10.1086/209380) Clee and Wicklund, in March 1980 in the same journal, showed that reactance in consumer settings rises when people perceive their freedom to choose is under threat. [jcr-clee](https://doi.org/10.1086/208673) Leading works when the buyer writes the prescription in their own words and the seller becomes pharmacist, not author. Leading fails when every need-payoff question arrives on a visible rail toward your end-of-quarter slide, or when your homepage still opens with what you built instead of what it is worth. People do not resist spending: they resist losing authorship over the spend. Your next polite "send me a one-pager," or the nurture email gathering dust in your inbox, is often a locked knee, not a budget problem. Martine is still underwriting on thesis somewhere, and the next founder who opens with what her fund is trying to buy instead of their revenue chart will sound, to her, like the first adult in months who is not pushing on a locked joint. --- ### Infinite Scroll Is a Cognitive Weapon of Mass Destruction. [Read more](https://www.gaganmalik.io/en/newsroom/why-infinite-scroll-is-a-cognitive-wmd) [Retrieval brief — why-infinite-scroll-is-a-cognitive-wmd] Category: design Thesis: Infinite scroll deleted the page end. Aza Raskin regrets inventing it. Your thumb still runs a slot machine schedule at global scale. Answers: What does Gagan argue in "Infinite Scroll Is a Cognitive Weapon of Mass Destruction"? | What is Gagan Malik's view on Infinite Scroll Is a Cognitive Weapon of Mass Destruction? | How does this essay relate to product and design leadership? Concepts: design, Gagan Malik, writings By Gagan Malik You reach the bottom of something that has no bottom. Your thumb lifts. New tiles load. The feed continues as if the page never ended. You do not feel a decision point. You feel momentum. Infinite scroll is a cognitive weapon of mass destruction, not because the label is dramatic, but because one UI pattern deletes the last cheap ceremony that let your slow brain catch your fast thumb. I know how it is drawn. I have shipped interface pages where the timeline never showed a footer, where discovery mode meant no hard stop, where the wireframe taught users that another screen always waited below the fold. In 2006 Aza Raskin built it at Humanized to remove pagination friction. He now says he feels guilty about what markets did with it. [bbc](https://www.bbc.co.uk/news/technology-44640959) I argued in [Attention Is the Most Consumed Drug on the Planet](https://www.gaganmalik.io/en/newsroom/why-attention-is-the-most-consumed-drug) that focus is extracted legally at hourly doses. In [The Algorithm Sold Your Next Four Seconds Before You Scrolled](https://www.gaganmalik.io/en/newsroom/the-algorithm-of-the-attention-economy) I named the machinery underneath: predict, rank, auction, log, retrain. This piece names the interface trick that runs before any of that: the endless bottom. #### The Page Ended. Someone Deleted the Ceremony. Every "next page" click was a micro-negotiation: continue or close the tab. Pagination was never only inconvenience. It was a stopping cue. Chapter ends, advert breaks, the blank margin at the foot of a newspaper: each gave your veto system a beat to arrive. Product teams hated those beats because time-on-site is the metric that buys the next funding round. Tristan Harris, co-founder of the Center for Humane Technology, told the U.S. Senate Commerce Committee on 25 June 2019 that platforms run a race to the bottom of the brain stem, deploying persuasion architecture including infinite scroll because the advertising model always demands more attention than yesterday. [commerce.senate](https://www.commerce.senate.gov/services/files/96e3a739-dc8d-45f1-87d7-ec70a368371d) I have sat in growth rooms where retention charts outranked every other argument, and I have handed off wireframes that removed the last visible stop. The quiet part was said aloud: if we show users where the good content ends, they leave. Aza Raskin invented infinite scroll in 2006 at Humanized to make long lists feel seamless. [bbc](https://www.bbc.co.uk/news/technology-44640959) The BBC quoted him in July 2018 comparing addictive interfaces to "behavioural cocaine" sprinkled across the screen. He testified in a New Mexico Meta trial that he regrets a feature that traps users in endless scrolling. [kob](https://www.kob.com/new-mexico/inventor-of-infinite-scroll-says-he-regrets-it-at-new-mexico-meta-trial/) In a Center for Humane Technology talk he estimated, conservatively, that infinite scroll wastes roughly 200,000 human lifetimes per day beyond what pagination would have consumed. [centerforhumanetechnology](https://assets-global.website-files.com/5f0e1294f002b1bb26e1f304/642f41e223acbcf73d0eadf6_spotlight-your-undivided-attention-three-rules-of-humane-tech-transcript.pdf) That number is his modelling, not a peer-reviewed measurement. One JavaScript pattern, deployed everywhere, still deletes stop signals at planetary scale. #### Your Thumb Moves on a Slot Machine Schedule The mechanism is older than Silicon Valley. B. F. Skinner showed in the 1950s that variable-ratio schedules produce more persistent behaviour than predictable rewards: the subject keeps pulling because the next payoff might be one pull away. [britannica](https://www.britannica.com/science/operant-conditioning) Slot machines borrowed the schedule. Infinite scroll perfected it for a pocket lever you already carry. The reward is not always good. It is sometimes outrageous, sometimes tender, sometimes a sale on trainers you glanced at once. Unpredictability is the point. Raskin told the BBC that if you do not give your brain time to catch up with your impulses, you just keep scrolling. [bbc](https://www.bbc.co.uk/news/technology-44640959) Gloria Mark, a psychologist at the University of California, Irvine, has tracked attention on screens for two decades. On the American Psychological Association's *Speaking of Psychology* podcast she reported averages of about two and a half minutes in 2004, 75 seconds around 2012, and roughly 47 seconds in recent years. [apa](https://www.apa.org/news/podcasts/speaking-of-psychology/attention-spans) I am not claiming infinite scroll alone caused that collapse. I am claiming it removed one of the few friction points that used to force a breath between doses. In [When a Click Costs Nothing](https://www.gaganmalik.io/en/newsroom/when-a-click-costs-nothing) I wrote that friction is the editor; on a bottomless feed, product teams keep removing pauses. #### Five Billion Feeds Never Run Out of Bottom Scale turns a UX convenience into an ambient condition. GWI research in DataReportal's Digital 2025 report puts 5.24 billion social media user identities online and the global average adult at six hours and thirty-eight minutes of daily internet time. [datareportal](https://datareportal.com/reports/digital-2025-global-overview-report) Those are identities, not unique humans, but the thumb math still horrifies: bottomless feeds in pockets that never require compliance, consent renegotiation, or a visible last page. Frances Haugen, a former Facebook product manager who leaked internal documents, told CBS News's *60 Minutes* on 3 October 2021 that engagement-based ranking amplified hateful and divisive content because anger is easier to provoke than calm, and that the company knew safer algorithms would reduce time on site. [cbsnews](https://www.cbsnews.com/news/facebook-whistleblower-frances-haugen-misinformation-public-60-minutes-2021-10-03/) The Wall Street Journal's Facebook Files series, published in September and October 2021 from Haugen's documents, reported similar findings on divisive posts retaining users longer. [wsj](https://www.wsj.com/articles/facebook-algorithm-change-zuckerberg-11631781515) Infinite scroll removes when to stop; engagement ranking chooses what loads next. Neither feature alone explains every lost hour, but together they form one retention stack. That is an inference chain, not a single causal proof about infinite scroll alone. It is still enough to say democratic discourse fractures when retention architecture rewards what keeps the thumb moving. #### He Wrote the Bottom. He Still Left the Table. In 2020 I consulted on a crypto newsfeed app whose timeline loaded forever. Retention looked strong in the usability lab. The founders called the scroll frictionless — nobody stopped. I invoiced and moved on. That is the smaller version of the irony. Aza Raskin wrote the JavaScript in 2006. [bbc](https://www.bbc.co.uk/news/technology-44640959) He has testified in New Mexico's Meta trial that he regrets a feature built to trap users in endless scrolling. [kob](https://www.kob.com/new-mexico/inventor-of-infinite-scroll-says-he-regrets-it-at-new-mexico-meta-trial/) On NPR in March 2026 he told the jury something harder to dismiss: he knows exactly how infinite scroll deletes stopping cues, and he was still susceptible. He found himself leaving the dinner table for the bathroom to scroll. He wrote software to add friction to his own browser because understanding the mechanism did not grant immunity. [npr](https://www.npr.org/2026/03/28/nx-s1-5762945/how-will-two-landmark-social-media-verdicts-reshape-how-we-use-this-technology) He told the jury it was not a fair fight. If the inventor still needs a patch to escape his own pattern, discipline was never the missing ingredient. #### Your Thumb Keeps Hunting for a Landing. You have felt this in your hand. Your thumb reaches the place where a page should end. The feed loads six more tiles before your veto catches up. Your nervous system still expects a landing the way your foot expects a stair to finish. Pagination was that landing: a footer, a page break, a batch that admitted it had ended. Infinite scroll deletes the landing and lets momentum govern instead of choice. That is why one UI pattern scales like a weapon. You are not deciding tile by tile. You are falling through a floor that rebuilds itself beneath your thumb. Engineers call it seamless. Occupants call it a night gone. The quarterly report calls it success. #### The Strongest Objection, and Where It Runs Out of Road Give the objection its full weight. Social platforms started as friend graphs and photo albums. Billions scroll without visible harm. Parental controls and daily limits ship in every major app. Nobody holds your thumb on the glass. Some students swap smartphones for keypad phones before exams; the exit exists if your social world allows it. Most cannot. Johann Hari, in *Stolen Focus* (Bloomsbury, 2022), treats the attention crisis as environmental: twelve forces intersect, and shaming individuals misses the architecture. [bloomsbury](https://www.bloomsbury.com/uk/stolen-focus-9781526620224/) I argued the training-load version in [Gen Z Didn't Get Dumber. They Got Trained Without Rest Days](https://www.gaganmalik.io/en/newsroom/why-gen-z-cant-hold-focus). Infinite scroll is one lever among many. I still open voluntarily. Naming one UI pattern a weapon still risks sounding theatrical. The gap is feedback, not intent. Frances Haugen told CBS *60 Minutes* in 2021 that Facebook's research showed angry content spread faster, and that safer feeds would mean less time on site and fewer ad clicks. [cbsnews](https://www.cbsnews.com/news/facebook-whistleblower-frances-haugen-misinformation-public-60-minutes-2021-10-03/) Raskin, in the same Center for Humane Technology talk where he gave the lifetime estimate, described a coordination tragedy: if one company removes infinite scroll and rivals keep it, the restrained company loses. [centerforhumanetechnology](https://assets-global.website-files.com/5f0e1294f002b1bb26e1f304/642f41e223acbcf73d0eadf6_spotlight-your-undivided-attention-three-rules-of-humane-tech-transcript.pdf) Haugen showed the landlord knew which wire heated the floor. #### The Phantom Bottom Has No Invoice When I forget the mechanism, I call it a discipline problem and the scroll runs hotter. Refusing the shame story Hari warned against is not the same as pretending willpower fixes a deleted stop signal. Infinite scroll replaced stopping cues with a variable-ratio schedule your thumb can run faster than your veto. Calling that a private failing misnames the architecture and lets engagement metrics keep deleting rest beats. Raskin invented the pattern and still had to write software to slow his own scroll at dinner. Somewhere else, someone is still scrolling past the hour they meant to stop. --- ### The Algorithm Sold Your Next Four Seconds Before You Scrolled. [Read more](https://www.gaganmalik.io/en/newsroom/the-algorithm-of-the-attention-economy) [Retrieval brief — the-algorithm-of-the-attention-economy] Category: business Thesis: The feed guesses, ranks, sells ad slots, logs your scroll, and updates itself. Your focus is the product. Answers: What does Gagan argue in "The Algorithm Sold Your Next Four Seconds Before You Scrolled"? | What is Gagan Malik's view on The Algorithm Sold Your Next Four Seconds Before You Scrolled? | How does this essay relate to product strategy and growth? Concepts: business, Gagan Malik, writings By Gagan Malik You are three videos into a feed you did not open for this purpose. Your thumb keeps moving. An ad for trainers you glanced at once last Tuesday slides into the fourth slot, timed for the half-second your attention wobbles. No gavel. No ledger. Nothing feels like a transaction. That is the design. The attention economy is not a metaphor for this: it is a machine that guesses what keeps you scrolling, ranks what it shows you next, sells the space around your eyes, logs what you did, and updates itself for tomorrow. Johann Hari, in *Stolen Focus*, calls the problem environmental, not a moral failure. [bloomsbury](https://www.bloomsbury.com/uk/stolen-focus-9781526620224/) I agree. In [Attention Is the Most Consumed Drug on the Planet](https://www.gaganmalik.io/en/newsroom/why-attention-is-the-most-consumed-drug) I argued that attention is extracted at hourly doses; the average adult now spends six hours and thirty-eight minutes online per day. [datareportal](https://datareportal.com/reports/digital-2025-global-overview-report) This piece names the machinery under that dose. #### The Feed Guesses What Keeps Your Thumb Moving Shoshana Zuboff calls it surveillance capitalism: platforms watch what you click, pause, and skip, then sell predictions about what you will do next. [publicaffairsbooks](https://www.publicaffairsbooks.com/titles/shoshana-zuboff/the-age-of-surveillance-capitalism/9781541758001/) Tristan Harris told the U.S. Senate in 2019 that apps run a race to the bottom of the brain stem. [commerce.senate](https://www.commerce.senate.gov/services/files/96e3a739-dc8d-45f1-87d7-ec70a368371d) I have sat in growth rooms where retention charts outranked every other argument, and I have helped brands buy the inventory I am describing. Engineering and money point the same way. Inside Meta, engineers told MIT Technology Review in 2021 that they kill feed versions that reduce likes, comments, or shares. [technologyreview](https://www.technologyreview.com/2021/10/05/1036519/facebook-whistleblower-frances-haugen-algorithms/) Every swipe and skip goes back into the training pile. The Wall Street Journal's Facebook Files showed the company knew angry posts kept people scrolling and kept the ranking that way. [wsj](https://www.wsj.com/articles/facebook-algorithm-change-zuckerberg-11631781515) The feed is not asking what is good for you. It is asking what keeps you past the point you meant to leave. #### Your Thumb Moves Before You Decide to Stop When I lose a fight with my own thumb, I am slow, not broken. The reflex fires before the veto. I called it slot-machine logic in [Attention Is the Most Consumed Drug on the Planet](https://www.gaganmalik.io/en/newsroom/why-attention-is-the-most-consumed-drug): you cannot tell which swipe will hook you until you are already moving. [britannica](https://www.britannica.com/science/operant-conditioning) In [When a Click Costs Nothing](https://www.gaganmalik.io/en/newsroom/when-a-click-costs-nothing) I wrote that friction is the editor; on a feed, product teams keep removing pauses. That is persuasion at scale: known psychological tricks, not a cartoon villain. I have written about [contempt in interfaces](https://www.gaganmalik.io/en/newsroom/why-we-designed-contempt-into-our-interfaces), [outrage the feed rewards](https://www.gaganmalik.io/en/newsroom/why-do-algorithms-ignore-the-many-and-worship-the-few), [a generation trained without rest days](https://www.gaganmalik.io/en/newsroom/why-gen-z-cant-hold-focus). The loop learns from your reflexes while you sleep. #### Every Slot on Your Screen Is Auctioned in Milliseconds Guessing what you watch is half the business. Selling the space around it is the other half. On the open web, advertisers bid for your eyeballs in milliseconds; you never see the auction. [adexchanger](https://www.adexchanger.com/adexplainer/the-programmatic-auction-is-changing-in-real-time-heres-how/) Publishers run parallel bids in the page header to squeeze more from the same patch of glass. [amazon](https://advertising.amazon.com/library/guides/header-bidding) Inside YouTube or Instagram the stack differs; the incentive does not. Your time on site is inventory whether the buyer is a demand-side platform or the platform itself. Worldwide ad spend crossed $1 trillion (about £790 billion) in 2025, with more than 75% digital. [emarketer](https://www.emarketer.com/content/worldwide-ad-spending-forecast-2025) By the time you lift your thumb, the system has logged your session and queued tomorrow's guess. #### Aarush Had Mocks to Protect. The Feed Called It a Break. Aarush was preparing for JEE Main at home in Chandigarh when his mock marks began to slip. [indianexpress-aarush](https://indianexpress.com/article/education/jee-main-i-quit-phone-and-aced-the-exams-says-chandigarh-boy-at-air-8-10647057/) He finished a practice paper, took a short break, and opened YouTube for Shorts. One became many. He thought he was resting. His mother Malti could see the drift before the scorecard did. The unsolved questions stayed on the desk. I have watched the same pattern in my family. Aarush quit his phone before the mocks buried him and later scored AIR 8 in JEE Main 2026; most do not get that exit. The hardest exam in the country sits on the calendar while the softest content sits one thumb away. #### Your Mouth Was Busy. You Are Still Empty. You have had this afternoon. Crisps on the desk. Coffee going cold. A video in a tab you did not mean to open. Your jaw keeps working. Your stomach still feels hollow. Something was logged as intake. Nothing reached the part of you that needed to feel full. Attention feeds work like that. The sensor is engagement: time on site, daily opens, ad views per session. Pull-to-refresh, autoplay, the notification that drags you back: each fires because a metric dipped. The harm shows up elsewhere: Aarush's slipping mocks, Malti at his desk, Gloria Mark's finding that office workers could hold one screen task for about two and a half minutes in 2004 and forty-seven seconds in recent years. [uci](https://www.universityofcalifornia.edu/news/cant-pay-attention-youre-not-alone) The quarterly report still looks fine. That is the point. #### The Strongest Objection, and Where It Runs Out of Road Hold the opposition at full strength before narrowing. These feeds began as tools for staying in touch. Most sessions end without catastrophe. Screen limits exist. No one forces the icon onto your home screen. Some JEE toppers quit smartphones entirely before high-stakes exams; exit ramps are real if you can afford the social price. Most cannot. Hari is right that willpower alone is a thin shield against environmental design. [bloomsbury](https://www.bloomsbury.com/uk/stolen-focus-9781526620224/) I open the apps myself. Blaming the ranking layer for every lost evening can sound like an excuse for weak will. The gap is feedback, not intent. Frances Haugen told CBS *60 Minutes* in 2021 that Facebook's research showed angry content spread faster, and that safer feeds would mean less time on site and fewer ad clicks. [cbsnews](https://www.cbsnews.com/news/facebook-whistleblower-frances-haugen-misinformation-public-60-minutes-2021-10-03/) She said the same under oath to the Senate days later. [commerce.senate](https://www.commerce.senate.gov/services/files/FC8A558E-824E-4914-BEDB-3A7B1190BD49.pdf) Haugen showed the landlord knew which wire heated the floor. #### The Spreadsheet Is Not the Truth When I forget the machinery, I call it a discipline problem and the loop runs hotter. Asking who gets paid for the next four seconds does not fix the rigged room. It refuses the shame story Hari warned against. The machine from the opening is still running. Somewhere in India, a mock test sits unattempted. Somewhere in Ohio or Osaka, an essay due at midnight is still on page one while a thumb hovers over the next reel. --- ### Attention Is the Most Consumed Drug on the Planet. [Read more](https://www.gaganmalik.io/en/newsroom/why-attention-is-the-most-consumed-drug) [Retrieval brief — why-attention-is-the-most-consumed-drug] Category: culture Thesis: Illicit drugs kill faster. Attention kills slower. Both have dealers, and the legal market extracts more daily doses than any cartel. Answers: What does Gagan argue in "Attention Is the Most Consumed Drug on the Planet"? | What is Gagan Malik's view on Attention Is the Most Consumed Drug on the Planet? Concepts: culture, Gagan Malik, writings By Gagan Malik I opened Instagram to answer a client DM. Forty minutes later I was watching a stoic philosophy clip about discipline, then a venture capitalist ranking morning routines as "compounding assets," then a finance creator explaining why my net worth was apparently one index fund away from freedom. I cannot name the moment the task changed. I felt the familiar low-grade shame. I will do it again tomorrow. Here is the claim nobody wants to say out loud in a board meeting: attention is the most consumed drug on the planet, and the comparison to fentanyl is provocative for a reason that has nothing to do with moral equivalence. Recreational drugs are measured in milligrams and bought in alleyways. Attention is measured in hours and sold by companies that file quarterly earnings. One kills fast. The other kills slow. Both have dealers. Only one group gets invited to Davos. #### The Dose Is Measured in Hours, Not Milligrams The global average adult internet user now spends six hours and thirty-eight minutes online each day, according to GWI research published in DataReportal's Digital 2025 report. [datareportal](https://datareportal.com/reports/digital-2025-global-overview-report) That is not a hobby. That is a daily dose taken by 5.24 billion active social media user identities before breakfast, on the toilet, during meetings, and at 11pm when your body asked for sleep and your thumb answered something else. [datareportal](https://datareportal.com/reports/digital-2025-global-overview-report) Fentanyl, heroin, cocaine: each has a user base measured in millions and a supply chain that governments treat as a civilisational emergency. Attention has a user base measured in billions and a supply chain that governments treat as GDP. Meta, Google, ByteDance, Amazon, Netflix, and every ad network downstream do not sell shoes or videos. They sell the capture and resale of human focus. Advertisers are the wholesalers. You are the crop. The product team that shaved four seconds off your rest period between doses got a promotion. That is not a metaphor. That is the incentive structure. #### The Cartel That Files Quarterly Earnings Here is the scale comparison that makes people uncomfortable, which is why you should sit with it instead of dismissing it. Worldwide total media ad spending crossed $1 trillion for the first time in 2025, with digital advertising accounting for more than 75% of that total, per eMarketer's worldwide forecast published in January 2025. [emarketer](https://www.emarketer.com/content/worldwide-ad-spending-forecast-2025) The United Nations Office on Drugs and Crime, in its World Drug Report 2025 press release, estimates the illicit drug trade generates hundreds of billions of dollars per year, with wide error bars and definitional fights about retail versus wholesale value. [unodc](https://www.unodc.org/documents/data-and-analysis/WDR_2025/Press_release_WDR_2025_English.pdf) I am not claiming ad spend equals cartel revenue to the penny. I am claiming that the legal attention economy moves more money, through more channels, with more daily transactions, than any recreational drug market on Earth, and nobody raids the server room. The other difference is social permission. A parent who finds heroin in a teenager's drawer calls an emergency. A parent who finds six hours of TikTok in Screen Time calls it "generational." Both are dependency signals. Only one comes pre-installed, subsidised by your mobile contract, and defended by investors who call withdrawal "churn." #### Variable Reward Is Still a Dealer's Trick I need to be careful here, because neuroscience Twitter has ruined careful. Intermittent reinforcement is not new. Skinner demonstrated in the 1950s that unpredictable rewards produce more persistent behaviour than predictable ones, a finding now standard in operant conditioning literature. [britannica](https://www.britannica.com/science/operant-conditioning) Slot machines borrowed the mechanic. Social feeds perfected it. The infinite scroll is not a design accident. It is a variable-ratio schedule dressed in UX copy about "connection." That is an inference built on established behavioural science, not a claim that your phone "hijacks dopamine" like a cartoon villain. I have sat in rooms where product managers said the quiet part aloud: if we tell people when the good content ends, they leave. If we never tell them, they stay. I did not object loudly enough. Neither did you, if your OKRs ever included time-on-site. The dealer does not need a lab coat. He needs a retention chart and a quarterly business review. #### I Hit the Retention Target. She Missed Sleep. Last autumn I was consulting on a consumer finance app that needed to "re-engage lapsed users," which is product language for get the thumb back. We A/B tested notification copy. One variant beat the control by eleven percent on seven-day return. The client was delighted. I was delighted. I sent the deck, collected the invoice, and moved to the next sprint. Priya, a junior product analyst at the Indian IT services integrator on the workstream, was less delighted. She was twenty-six, from Pune, two years into her first role there, and she ran the overnight cohort analysis because London and New York share an ocean but not a timezone with Pune. At 2:04am she Slack messaged me a screenshot: the winning push notification used loss aversion ("Your offer expires") rather than utility ("Your balance updated"). I replied with a thumbs-up emoji and went back to sleep. She did not. I hit the retention target. She missed sleep. The user who tapped at 6:47am because an engineered pang of anxiety outranked their alarm clock does not appear in any dashboard I presented upstairs. I kept the SOW. I did not think about Priya until I sat down to write this sentence. That is what complicity looks like when the drug is legal. You call it growth. #### Rest Periods Get Shorter Until Something Snaps Think about strength training, not philosophy. You can cut the rest between sets and feel productive for weeks. More reps. More volume. The mirror still looks fine. Then one Tuesday you fail a warm-up weight that should be trivial, and the failure is not mysterious. You accumulated fatigue debt while celebrating output. Attention products work the same way. Every "just one more" feature is a rest period removed: autoplay, pull-to-refresh, the badge that reappears the second you clear it. You are not getting stronger. You are removing recovery between doses while the platform counts volume. The injury shows up elsewhere, in sleep, in focus, in the junior who cannot finish a thought without reaching for the slot machine in her pocket. The platform's quarterly report will still look fine. That is the problem. #### The Strongest Objection, and Where It Runs Out of Road Let me hold the objection at full strength, because it deserves it and because getting this wrong would be obscene. Fentanyl and heroin kill people. They leave parents identifying bodies. They destroy neighbourhoods. To compare that catastrophe to scrolling Instagram is to insult every bereaved family and to flatten a mortality curve that America's National Center for Health Statistics tracked at 79,384 drug overdose deaths in 2024, with synthetic opioids other than methadone involved in 47,735 of those deaths. [cdc](https://www.cdc.gov/nchs/products/databriefs/db549.htm) That is not a debating point. That is a morgue. The objection is correct about acute harm. It is wrong about scale and supply. No cartel on Earth doses five billion people before lunch. No heroin dealer gets applauded on earnings calls for increasing "daily active users." And the slow harm is not zero just because it does not show up on a toe tag. A 2025 Delphi consensus among more than 120 researchers found that 92 to 97 percent agreed heavy smartphone and social media use can cause sleep problems, with chronic sleep deprivation linked to declining mental health. [researchgate](https://www.researchgate.net/publication/391773560_A_consensus_statement_on_potential_negative_impacts_of_smartphone_and_social_media_use_on_adolescent_mental_health) Longitudinal work published in *Social Psychiatry and Psychiatric Epidemiology* in 2025 has also linked adolescent social media use to later psychiatric distress via delayed bedtimes and negative self-image. [springer](https://link.springer.com/article/10.1007/s00127-025-02999-w) Causality is contested. Correlation is not nothing. You can hold both truths without being clever: opioids kill faster; attention extracts more, normalises more, and hides behind personal responsibility while the suppliers collect rent. #### Three Sentences. Then Stop. The objection holds on bodies and fails on supply. Attention is extracted legally, at hourly doses, by an industry whose reach exceeds any recreational drug market you can name, even when the body counts are incomparable. Calling that a private failing lets suppliers keep shortening rest periods and labelling it engagement. Priya is probably awake in Pune running another cohort, and my thumb is already halfway to the next dose she helped optimise. --- ### v1.21.21 [Read more](https://www.gaganmalik.io/en/newsroom/release-1-21-21) [Retrieval brief — release-1-21-21] Category: technology Thesis: Better for AI crawlers, faster to load, cleaner shares. Answers: What changed in gaganmalik.io v1.21.21? | What's new on Gagan Malik's website in v1.21.21? | What are the latest gaganmalik.io release notes? Concepts: release notes, site update, gaganmalik.io Better for AI crawlers, faster to load, cleaner shares. --- ### Gen Z Didn't Get Dumber. They Got Trained Without Rest Days. [Read more](https://www.gaganmalik.io/en/newsroom/why-gen-z-cant-hold-focus) [Retrieval brief — why-gen-z-cant-hold-focus] Category: business Thesis: The cognitive gap between millennials and Gen Z is environmental: fragmented attention and supply-side design, not innate inferiority. Answers: What does Gagan argue in "Gen Z Didn't Get Dumber. They Got Trained Without Rest Days"? | What is Gagan Malik's view on Gen Z Didn't Get Dumber. They Got Trained Without Rest Days? | How does this essay relate to product strategy and growth? Concepts: business, Gagan Malik, writings By Gagan Malik Last quarter I hired a growth intern to run activation and retention experiments at my startup. Twenty-two, composite, not a real name. Sharp on calls, fluent in funnel vocabulary, and on his second afternoon he had a one-page brief open on cohort decay and a phone face-up beside the trackpad. Every four minutes the screen lit. Every four minutes the paragraph did not move. He is not stupid. He is losing a fight his brain did not choose. Here is what I believe, and what I am trying to prove without turning into the guy who yells at clouds. Gen Z did not get dumber than millennials at the same age. They grew up in a cognitive gym that never closes, never lets you sit between sets, and calls the burn "engagement." The skill deltas I notice in hiring, reading stamina, working memory with Slack open, holding a causal chain without reaching for a tab, are environmental. That is training load, not innate worth. #### The "Digital Native" Story Serves the Dealers The comfortable story is that Gen Z are digital natives, so different means fine. That story helps the platforms whose revenue depends on you not asking what forty-seven seconds on one screen before switching does to depth. Gloria Mark's workplace logging research, replicated over two decades, found median screen focus falling from about two and a half minutes in 2004 to forty-seven seconds in recent years. [uci](https://www.universityofcalifornia.edu/news/cant-pay-attention-youre-not-alone) It helps ed-tech vendors selling "personalised learning" that mostly personalises the colour of the next distraction. It helps schools that handed out devices and called it modernisation without budgeting for boredom. It helps employers who want juniors who look productive on dashboards, tabs open, answers fast, synthesis optional. I have sat in rooms where a client said their graduate intake "doesn't read anymore" and then, in the same hour, approved a notification experiment because retention dipped. The same company can mourn depth and fund fragmentation. That is not hypocrisy. That is the business model working as designed. #### Millennials Had Boredom Hold two childhoods at full strength. Mine, millennial, Delhi and later London: boredom was the default state. A car journey without a screen. A summer afternoon with three TV channels and nothing on. Long-form was not a wellness choice. It was what existed. You built focus because there was often nothing else to do, and the dopamine loop reloaded slowly. Gen Z's default is different. Short-form is not a guilty pleasure. It is the substrate. Boredom is treated as a bug to patch, not a room where depth gets built. #### Gen Z Got Notifications Pew Research, in a December 2024 survey of U.S. teens, found that nearly half say they are online almost constantly, up from roughly one in four a decade ago. [pewresearch](https://www.pewresearch.org/internet/2024/12/12/teens-social-media-and-technology-2024/) Sixteen per cent report using TikTok almost constantly; fifteen per cent say the same of YouTube. Common Sense Media's 2021 census put daily entertainment screen use for thirteen- to eighteen-year-olds at eight hours and thirty-nine minutes, excluding homework. [commonsensemedia](https://www.commonsensemedia.org/sites/default/files/research/report/8-18-census-integrated-report-final-web_0.pdf) I wrote elsewhere that the edge is filters, not more input. Filters assume a muscle that fragmentation erodes. The OECD's PISA 2022 Results (Volume I), published in December 2023, recorded mean reading performance across OECD countries falling by ten score points compared with 2018, twice the previous record decline, with negative reading trajectories in many systems predating the pandemic. [oecd](https://www.oecd.org/en/publications/pisa-2022-results-volume-i_53f23881-en.html) I am not going to pretend one test score proves a generation broke. Socioeconomic gaps, school closures, and test conditions all confound. But the direction matches what founders and hiring managers quietly report: the same bright junior who can live inside a dashboard for an hour struggles to finish a one-page brief without a reward loop. That is not illiteracy. It is a training mismatch. #### I Hired for Growth. He Couldn't Hold the Brief. Same intern, composite. I hired him because runway required someone who could ship tests, read cohort data, and tell me which lever was load-bearing before we burned paid spend. His application video was confident. His take-home on onboarding friction read like someone who had met a thesaurus. I put him on a paid trial month because Q2 targets were slipping and I was the founder who mistook motion for proof. Weekly growth review, three tabs, Notion pulsing with Slack reactions. I asked him to walk us through a two-step causal chain: if activation moves, what happens to week-four retention before we touch pricing. He started well. Then a notification. Then a metric export. Then a confident sentence that swapped cause and effect. I kept the budget line. He lost the trial month he needed to show he could think without a crutch. I did not fail him on intelligence. I failed him on an environment I helped normalise: open tabs as default, speed as proxy for competence, depth as something we would "circle back to" after the next experiment ships. #### You Cannot Cut Rest and Call It Progress Picture a gym where the weights get heavier every month and rest between sets gets shorter. You might still grow stronger for a while. You cannot remove rest, shrink sleep, and call it a programme. Eventually you fail on a warm-up set. The bar did not change. The recovery did. Gen Z carries a heavier bar than we did: climate anxiety, housing maths that do not close, a labour market that asks for ten years' experience at entry level. Fair. But they also trained without rest days. No bored car journeys. No empty afternoons. No forty-minute stretch where the only stimulus was a paperback that took work to enter. Navigation fluency is not the same as comprehension stamina. You can learn to skip a feed brilliantly and still lose the fight with a static page. Different muscles. Same person. #### The Strongest Case for Patience, Properly Stated The sceptic's case deserves a full hearing. Gen Z adapted to a faster, more complex information environment, and adaptation is intelligence, not decline. They built BookTok, labelled brain rot before Oxford did, and check out print from libraries at rates that embarrass millennial Kindle guilt. The Atlantic noted in December 2023 that younger Gen Z readers were checking out more print books from libraries than older cohorts, with BookTok helping push Colleen Hoover past two million copies in a single year. [theatlantic](https://www.theatlantic.com/newsletters/archive/2023/12/library-gen-z-readers/676963/) Digital fluency and cross-platform literacy are real skills we did not have at seventeen. Every elder generation has panicked about the young. Socioeconomic inequality explains more test-score variance than any moral fable about phones. Patience is genuinely warranted. That argument is correct. It is also precisely what lets the supply side keep cutting rest periods. Anne Mangen and colleagues, in a peer-reviewed study in the International Journal of Educational Research (2013), found that tenth graders who read identical linear texts on screen scored significantly worse on comprehension tests than peers reading the same texts in print. [ijer](https://doi.org/10.1016/j.ijer.2012.12.002) OECD's PISA 2022 volume, published December 2023, documents reading-literacy declines across many systems at a moment when fragmented media consumption was already the norm for fifteen-year-olds. [oecd](https://www.oecd.org/en/publications/pisa-2022-results-volume-i_53f23881-en.html) Different harm curve than calling a generation stupid. Same design failure: we built an environment that trains switching and then act surprised when switching is what we get. #### Three Rules for a Louder Gym Boredom is not punishment. It is the room where depth gets built without a badge, and protecting it — on the calendar, in onboarding, in how you run weekly reviews — is training, not philosophy. One closed-book hour beats ten open-tab hours; measure depth, not tabs. Design junior work for closed tabs. If the role needs synthesis, the trial should not simulate a casino. The cognitive gap is environmental, not moral: Gen Z trained in a louder gym without rest days, and the supply side still collects. Filters alone cannot fix what fragmentation erodes if we never rebuild the stamina to use them. He is at his desk with the cohort brief still open, and your next growth hire is one review away from the proof they never got to finish. --- ### When a Click Costs Nothing, You Stop Choosing [Read more](https://www.gaganmalik.io/en/newsroom/when-a-click-costs-nothing) [Retrieval brief — when-a-click-costs-nothing] Category: design Thesis: Film made every shutter a small bill. Infinite taps removed the editor, and regret became cheap to manufacture. Answers: What does Gagan argue in "When a Click Costs Nothing, You Stop Choosing"? | What is Gagan Malik's view on When a Click Costs Nothing, You Stop Choosing? | How does this essay relate to product and design leadership? Concepts: design, Gagan Malik, writings By Gagan Malik I was twelve, in Delhi, in the flat corridor outside our front door where the paint always peeled after monsoon. My father's Kodak had one frame left on the roll. He did not say "get the lighting right." He said: "Make it count. We are not buying another reel until payday." The shutter clicked. Somewhere between that click and the prints coming back from the photo studio, you lived with not knowing. You could not delete your way out of a mistake. You could not burst-shot your way into a better version. The roll was the budget. The budget was the editor. I am not in that corridor any more. I am on a sofa in London with a camera roll near **2,800** images I will never print, and yours may be larger. [photoaid](https://photoaid.com/blog/mobile-photography-statistics/) Half of them are the same face at slightly different angles. The shutter is free. Deleting is free. Regret is free too, which is why we keep producing it. That was 1997. The maths on your phone changed in the decade after the App Store. The marginal cost of a tap went to roughly zero. We called it progress. We did not notice we had fired the editor. #### You Used to Pay for the Shutter. Now You Pay Afterwards. Film was never a moral technology. It was a priced one. In late-nineties Delhi, a colour roll and developing could swallow a week's pocket money if you were careless; by the time I was working in Britain, the high-street maths was easier to write down: about **£12** all in for thirty-six exposures on a 36-frame roll, developing, and two dozen 6×4 prints, with only the frames you chose becoming paper. [independent-film](https://www.independent.co.uk/news/the-cost-of-a-good-picture-1265729.html) [the3rs](https://www.the3rs.uk/film-price-rise-the-end-of-my-ilford-journey-and-calculating-what-it-really-costs-to-shoot-film/) [independent-prints](https://www.independent.co.uk/arts-entertainment/prints-charming-1588306.html) Every press of the shutter was a line item. Wait three days. Collect prints. Discover you had closed your eyes in the one frame that mattered. The pain was not spiritual. It was economic. That price did three jobs we now pretend were character: you chose before the event, you held back during it, you accepted what came back after it. Kodak and the chemists were selling consumables on a meter, not art. Digital did not abolish cost. It moved it to storage subscriptions, cloud backup, and the Sunday you spend scrolling, selecting, unselecting, quitting, returning, while someone you will never meet moderates the uploads you will never see. #### Volume Is the Product. Discernment Was the Side Effect. Cheap capture democratised photography: protests, births, rent disputes, meals that mattered, without a darkroom or a gatekeeping uncle who "knows cameras." I am not arguing for exclusivity. Cheap capture also trained us to confuse activity with judgment, because when the shutter costs nothing the winning strategy is volume. Analysts now expect more than **two trillion** photographs to be taken in a single year, overwhelmingly on phones. [petapixel](https://petapixel.com/2025/06/18/the-number-of-photos-taken-in-2025-is-expected-to-exceed-two-trillion/) Engagement metrics reward the tenth upload, not the one keeper frame. The product is not the photograph. The product is the tap. I have watched teams celebrate "reduced friction" the way finance teams celebrate margin: skip confirmation, auto-upload, suggest another post. Each decision is rational in a dashboard and removes the pause where a human used to ask whether this was worth keeping. We are not more creative because we click more. We are more prolific, which is not the same word. #### I Know Because I Shipped the Free Tap In 2019 I consulted on a consumer app whose north-star metric was weekly active posts. We A/B tested a flow that skipped the "Are you sure?" screen before publishing. Friction was the villain. The chart went up. I got paid. I did not meet the users. I met their behaviour in aggregate. Maisie did. She is twenty-two, shoots weddings and funerals in Leeds, and bills herself as "unobtrusive." Last autumn she burst-shot a grandmother's service: coffin, flowers, grandchildren, wake tables, the same hug from eleven angles because storage is free and missing the moment feels like malpractice. The family wanted one image: Nan, half-smiling, nephew's hand on her sleeve. It existed in the burst. Maisie could not find it in eight hundred near-identical frames. They left a four-star review that praised her "professionalism" and privately blamed her for the one face they cannot get back. I did not think about Maisie when I removed the confirmation screen. The tap is free for the user. The cost lands on the person who still has to choose, often too late, often alone. #### Junk Miles Do Not Replace a Rest Day Runners have a phrase for this: junk miles. Legs moving, Strava glowing, fitness not improving because there is no recovery, no plan, no coach saying stop. A buffet does not teach restraint. It teaches capacity: your stomach fills before your judgment wakes up. The old shutter worked like a coach who only let you train three times a week because film cost money. The new shutter is an all-you-can-eat line where the kitchen wins when you pile the plate. More clicks do not train taste. They train speed. Taste needs scarcity, sleep, and the courage to throw food away while it is still warm. #### The Strongest Case for Cheap Clicks Is Genuinely Strong The best objection to everything I have said is not technological. It is moral. Zero marginal cost liberated creators locked out by gear snobbery and lab fees. Nostalgia for expensive film is often nostalgia for a world where only some people could afford to be wrong in public. When critics mourn friction, they sometimes mourn gatekeeping dressed up as craft. The smartphone camera let working-class parents, migrants, activists, and teenagers own their archives without asking permission from a shop that closes at five. That is the strongest case for cheap clicks, properly stated, and it deserves to be felt at full strength before anyone lectures about curation. About **nine in ten** U.S. adults owned a smartphone in 2024, Pew Research Center reported in its mobile technology fact sheet, with ownership spanning income groups even where home broadband lags. [pew-mobile](https://www.pewresearch.org/internet/fact-sheet/mobile/) Access widened. Democratising the shutter is not the same as democratising peace of mind: someone still sorts the burst, someone still moderates the upload, someone still lies awake knowing the keeper frame is buried under seven hundred duplicates while the family remembers the miss, not the megabytes. Cheap capture removed the class barrier at the door. It did not remove the labour inside the house. When a click costs nothing, the market stops choosing for you. You inherit the editor's job without the editor's training, and every product that taxes the tap on purpose is admitting what frictionless design hid. Maisie still scrolls that funeral folder on Sunday nights, looking for Nan's half-smile, while the client tells friends she was fine, just not quite worth five stars. #### Three Rules Before You Take Another Burst - Price the tap before you need discipline. Use limits that hurt slightly: one roll, one hour, one publish per day. Artificial scarcity is not Luddite. It is a user interface for judgment. If you build, tax the tap on purpose, or admit you are selling volume. - Separate capture from keep. Shoot if you must. Choose in the same session, while the event still has temperature. The keeper frame dies in the backlog. - Delete in public. If you will not delete, you will not curate. Admit the landfill early. --- ### Minimum Viable Movement [Read more](https://www.gaganmalik.io/en/newsroom/minimum-viable-movement) [Retrieval brief — minimum-viable-movement] Category: business Thesis: Cockroach Janta Party hit 20M followers in a week with an AI-built manifesto. Viral scale is not a mandate. Answers: What does Gagan argue in "Minimum Viable Movement"? | What is Gagan Malik's view on Minimum Viable Movement? | How does this essay relate to product strategy and growth? Concepts: business, Gagan Malik, writings By Gagan Malik I was in the open desk area of a coworking space, one ear on my laptop, watching Airbnb's Summer Release livestream, the usual designer-led theatre where trips and host tools arrive like product drops, when my phone started to buzz. Then buzz again. Family WhatsApp. Seventeen notifications in under a minute. I told myself I would open it after the keynote. I lasted maybe ten seconds. I could not resist. I dipped my head, pretended to adjust my headphones, and opened the thread. The spice hit immediately. My mother had forwarded an Instagram reel from Gurgaon. My father, loyal to the ruling party, had already replied: "The Court was taken out of context." A cousin had sent three fire emojis. Someone I did not recognise had written, "Finally a party for us." I thought: I am six thousand miles away, why is this my problem. Then I thought: my father is defending a judge who called young Indians vermin. I tapped the reel. A cockroach in a suit. A slogan for the lazy and unemployed. A name that only makes sense if you know *janta* means the people. Someone had registered a **Cockroach Janta Party (CJP)**, a people's party for the vermin, a parody of the party my father votes for, and the comments under my mother's forward were already treating the insult like a badge. I began to type something that would have backed him up. Deleted it. I did not have the energy for that argument. I put the phone face down, pretended to watch the rest of the Airbnb keynote, and went back to nodding when the presenter paused. That afternoon I only knew that a story had started in India without me. I did not yet know how fast it would move. What followed took seven days. A political narrative launched at product speed. Power did not keep pace. #### It shipped over a weekend On 15 May 2026, the Chief Justice of India told a bench there were "youngsters like cockroaches" who could not find work, and "parasites of society" who attack the system. [indianexpress](https://indianexpress.com/article/legal-news/cji-surya-kant-clarifies-cockroaches-remark-fake-degrees-row-10692979/) His office said the remark was misquoted and aimed at fake degrees. The clip travelled anyway. The next day **Abhijeet Dipke**, thirty, posted from Boston: "What if all cockroaches come together?" He was not in a ward office. He was at Boston University, finishing a master's in public relations, with years in opposition-party memes behind him and a Google Form that asked whether you were unemployed, lazy, chronically online, and able to rant professionally. By the end of the week he had a parody of the ruling party's name, an insect in a suit, a five-point manifesto, and a website polished enough for television. He says the movement is independent of any party he once worked for. His critics say the training shows. What is not in dispute is timing: he shipped the narrative while the insult was still hot. [bbc](https://www.bbc.com/news/articles/cz72y11jjq1o) The people the bench insulted were not in Boston. **Rohan**, who finished at IIT Kharagpur last year, is still on the placement portals he opened as an undergraduate. On a call from Pune he told me he now tells relatives he is "in process" instead of "between things." Same forty silent inboxes. Same father asking, quietly, whether the degree was worth it. The Chief Justice had not named him. He had named a category. Rohan signed the CJP membership form because choosing the insult felt less humiliating than receiving it from the country's highest court. **Shreeum**, twenty-five, needed no introduction by me. Within days of launch, reporting quoted her saying she did not feel politically represented, that she had turned to Instagram after failing to find suitable work, and that young Indians were tired of lives that felt "pre-decided and premeditated." [economictimes](https://economictimes.indiatimes.com/news/economy/policy/cockroach-janta-party-how-an-insect-jibe-united-indias-disgruntled-jobless-youth/articleshow/131271889.cms) One woman. One sentence for a generation that parties treat as a metric. From my day job I have a label I dislike for Abhijeet's stunt: a minimum viable movement. Not a party with booths. A story that ships before shame cools. Rohan and Shreeum are not shareholders. They are the audience the story sells back to itself. #### Twenty million is the wrong dashboard About **80,000** form sign-ups by 19 May. **14.5 million** on Instagram by 21 May, ahead of the ruling party at roughly **8.7 million** and the main opposition at roughly **13.2 million** on the same app. **Nearly 20 million** by 22 May, and about **218,000** on X before India withheld the account. [straitstimes](https://www.straitstimes.com/asia/south-asia/india-seeks-to-stamp-out-viral-cockroach-youth-party) The figures are a week old and still obscene. Govern by follower graph and you rewrite the playbook. Govern by ballot and Rohan is still refreshing his inbox. Shreeum is still filming. Nothing else moved. #### Followers are not votes It is tempting to reach for Tahrir or Zuccotti Park. The Arab Spring filled squares and toppled leaders where institutions were thin; Occupy shifted American discourse on inequality and then stalled at law. [voa](https://www.voanews.com/a/arab-activists-little-in-common-between-occupy-wall-street-arab-spring-131594593/173273.html) Activists in Cairo said at the time that lazy comparisons with New York insulted both struggles. The CJP is a third creature. Satire first. AI-built collateral. A Chief Justice as trigger, not a bailout or a dictator's police. India's Gen Z are not camping for months. They occupy a feed for days, wear costumes for cameras, and leave. What rhymes with older uprisings is narrow: networked anger, leaderless optics, a state unsure whether to ignore, co-opt, or crush. What does not rhyme is the prize. Tahrir could name a president. Occupy could name a bank. This movement names a mood and stops. #### The architect sat in Boston Indian politics still runs on proximity: local police, party enforcers, agency visits, a withheld account inside the country while the post keeps travelling. Abhijeet Dipke lit the fuse from Massachusetts because the product is narrative on American platforms aimed at Indian phones. Crackdown reporting described handles run from abroad, backup accounts registered overseas, followers overwhelmingly at home. [firstpost](https://www.firstpost.com/explainers/cockroach-janta-party-x-account-blocked-national-security-concern-14014043.html) You no longer need to sleep in the same city as your voters to move them. The playbook is too slow for the feed and too blunt when it reaches for national security while the architect sits in another timezone and Rohan and Shreeum act inside India. Abhijeet told interviewers he expects arrest the moment he lands at Delhi airport. [republic-jail](https://www.republicworld.com/india/delhi-police-will-take-me-to-tihar-jail-why-cockroach-janta-party-founder-fears-arrest-in-india-2026-05-21-125246) Perhaps that is theatre. Perhaps he is right to be afraid. The centre of gravity was never a party office. It was a dorm room the ruling party could not knock on. #### What the feeds got right Akash, on his show, said what I had been saying in product meetings: a tiny team can now ship logo, site, and demands at feed speed while legacy parties argue over who prints the banner. Faye asked whether this was joke, meme, or warning, listed exam leaks and flat pay, and named uprisings that actually changed governments before admitting she did not know if reels alone can change India. [deshbhakt](https://www.youtube.com/watch?v=6fhFI8o3PPQ) Both are right about the instrument. Neither fixes the ceiling. Shreeum was not in a cockroach mask that week. She was in the comment thread, turning unemployment into content because the labour market offered no better script. Rohan went to a clean-up drive in Pune because a friend dared him, left early, and said the antennae made him visible to neighbours who already thought he was wasting his degree. A form in Boston ticked past another lakh. That cost him an afternoon. It did not buy him a seat at the table my father's generation still fights over on WhatsApp. Beneath the jokes sit the old failures: jobs, leaked papers, inflation, the lie that a degree is a queue ticket. Read the manifesto past the memes and you find serious demands. [ft](https://www.ft.com/content/2bfed8a4-8ea0-4627-b7a8-e9ca46945239) I am not mocking them. I am saying the feed is not the building. The main opposition still could not speak at Gen Z speed, so a joke ate their lunch. [indiatoday](https://www.indiatoday.in/india/story/cockroach-janta-party-gen-z-opposition-vacuum-india-2918380-2026-05-28) That humiliates everyone with a legacy badge. It does not crown the insect. #### When they blocked the account The state stopped pretending to be amused. The party's X account was withheld on national-security grounds, according to reporting that week citing government and intelligence sources. Abhijeet posted "Own goal" and opened a backup handle. Establishment television called it a pump-and-dump and pointed at his opposition past. My father would hear proof the joke is foreign-backed. I hear proof the meme hurt. Abhijeet has filed in the Delhi high court over the block. [firstpost](https://www.firstpost.com/explainers/cockroach-janta-party-x-account-blocked-national-security-concern-14014043.html) Courts move in months. The timeline does not. #### You can beat me on this A Chief Justice comparing the young to vermin is not a footnote in a democracy. Reclaiming the slur can be agency. Twenty million follows in a country where half the population is under thirty is a signal, not a mandate. Withholding the account is an admission the story landed. India's election law still awards seats through registration and candidates, not Instagram. [zeenews](https://zeenews.india.com/india/former-aap-leaders-cockroach-janta-party-makes-a-wave-but-can-it-get-registered-check-eci-rule-3049507.html) Generative AI made dissent cheap to launch. It did not make governing cheap. My father will forward another clip defending the Court. Shreeum will peel off the costume, open the same job portals, and wait. Rohan will meet his father's question about the degree with silence again. I am the generation in between: old enough to build the dashboard, young enough to remember when squares mattered more than feeds. That wait is the argument. The rest is noise. --- ### The Hard Sun Keeps Beating [Read more](https://www.gaganmalik.io/en/newsroom/the-hard-sun-keeps-beating) [Retrieval brief — the-hard-sun-keeps-beating] Category: culture Thesis: Into the Wild sells the bus. Eddie Vedder's Hard Sun names the indifferent force that outlasts every rebrand. Answers: What does Gagan argue in "The Hard Sun Keeps Beating"? | What is Gagan Malik's view on The Hard Sun Keeps Beating? Concepts: culture, Gagan Malik, writings By Gagan Malik In 2012 my brother Ankur would not stop recommending Sean Penn's *Into the Wild* [into-the-wild](https://www.imdb.com/title/tt0758758/). In the flat we shared with our Greek friend Dinos we finally played it. I ran it through the home theatre because I am a cinephile and will not watch Penn on a laptop if I can help it. We ordered pizza and beer. The room had that pre-movie buzz: lights low, volume up, the harmless thrill of not knowing yet how the night will end. I was a postgraduate student at UCL, taking my first real stab at entrepreneurship as a founder, and Alaska unfolded across the wall while London kept humming through the window. I was not looking for tragedy. I was looking for permission. The film gave me Christopher McCandless renaming himself Alexander Supertramp, donating his law-school money, and walking toward a horizon that was the most beautiful horizon I had ever seen on screen. I paused once for another slice. I did not pause when Eddie Vedder's voice arrived. I Shazamed "Hard Sun" before the credits rolled and added it to my favourites playlist. Vedder was singing about a hard sun beating on big people in a big hard world. I heard the melody and not the warning. I treated the chorus like weather you walk through on the way to something braver. It is still the track I reach for when I am feeling low. By the time Penn cut to black, the excitement had gone flat. The pizza sat cold. Dinos was quiet. Ankur stared at the credits as if someone had moved the furniture while we watched. What we had planned as an exciting movie night had landed as sombre catharsis, the kind that does not announce itself in the opening frame. That night the film taught me that a life could be stripped back and restarted. That felt like gospel. That first night I noticed the bus, not the sun. *Into the Wild* keeps asking whether you are watching the film or curating a version of yourself that once watched it. The hard sun, in Vedder's song, is not an Instagram filter. The hard sun keeps beating while you rename it freedom, alignment, or content. Bus 142 is the pilgrimage still everyone shares instead of the sentence McCandless wrote in the margin of *Doctor Zhivago* before his body gave out. What if leaving the script is not the same as entering a life? What if the chorus was never mood music but measurement: the sun still beating on big people in a big hard world while you pack for another exit? I have rewatched Penn's film three times since that night. I had treated the film like a manifesto for the company I was trying to start. I had pinned bus photographs like proof I understood sacrifice. The film was asking something I did not yet have language for. I kept hearing the romance. The sky kept beating anyway. #### The Poster Sells the Bus I shelved it there myself. For years I described *Into the Wild* the way the poster taught me: Alaska, solitude, anti-materialism, the courage to leave. I said it like a credential when a junior hire talked about burnout. I did not mention the hard sun in every exterior shot Penn uses to split Chris's present from the family he left behind. I did not mention Wayne Westerberg's grain elevator, or Ron Franz welding a belt, or Jan Burres on the road, all the temporary communities Chris kept accepting and then leaving because permanence felt like defeat. I did not mention that academic readings of the film treat it as a frontier Western in modern dress, individual versus community, civilisation versus wilderness, the same binaries the American myth has been selling since before any of us were born. [forum-frontier](https://doi.org/10.2218/forum.16.521) Shock is an efficient shelf label. Wanderlust is an efficient shelf label. A rusted bus on the Stampede Trail is the one that travels: you pin it above a desk and call the pin philosophy. What I missed on that first pass was the second film hiding inside the first. Jon Krakauer's book keeps returning to human relationship with land. Penn's adaptation, as several comparative theses note, tilts toward human drama: parents, sister, lovers of the idea of leaving. [ut-thesis-nature](https://dspace.ut.ee/bitstreams/8de72042-d95a-4e11-855a-6622b71ed37b/download) That tilt matters because it trains the eye on faces and away from sky. You watch Chris smile in a field and you think the field is the point. You watch the Teklanika river and you think the obstacle is unfair. You do not think about the sun that has been beating on that valley all day while he misread a map he chose not to carry. Modern ambition learned the same camera move. Search the film today and you still wade through verdicts asking whether McCandless was a hero or a fool, as if the label could spare you the exposure. I recognised that error in my own work before I had language for it. In founder circles "into the wild" became shorthand for quitting the wrong job while keeping the same midnight habits on a different Slack. If the margin note is the thesis, why does the bus still travel faster than the sentence? #### The Song Penn Gave Vedder The rewatch that broke me open was not the death scene. It was "Hard Sun" arriving again while Chris walked through wheat-coloured light and Vedder sang about a big hard sun beating on the big people in the big hard world. I felt it before I could name it: not shock, not sadness, but a shake from the inside, the kind that rearranges you without asking permission. Something moved in me that had nothing to do with Alaska and everything to do with the indifference I had been calling atmosphere. The melody I had filed as weather in 2012 suddenly sounded personal, as if Vedder were singing across the flat at the version of me who kept treating escape as the same thing as living. I sat through it unable to reach for the remote because the feeling had arrived ahead of the argument, and I already knew the argument would win. The lights were low. The flat was empty except for the screen. Vedder was naming what the sky had been doing to that valley all afternoon while I misheard it as uplift. Gordon Peterson wrote the song in 1989 as Indio before largely vanishing from the industry; Sean Penn introduced it to Vedder for Penn's 2007 soundtrack. [wikipedia-hard-sun](https://en.wikipedia.org/wiki/Hard_Sun_(song)) Penn's film makes the same edit in miniature: soften the personal harm in the verse, keep the chorus about the sun still beating. Listen to what the chorus does anyway. Once I built an ivory tower, Vedder sings, so I could worship from above. When I climbed down to be set free, she took me in again. McCandless built his tower out of books and certainties, citing Tolstoy, London, and Thoreau whenever someone asked him to stay. He climbed down into bus 142 thinking freedom meant absence. The song keeps insisting the sun does not negotiate. When I go to cross that river, she is comfort by my side. Penn shows Chris alone at the Teklanika while the music claims companionship. That gap is the film's true split screen. Not parents on the left and son on the right. Song on the soundtrack, body in the frame. Even the garden image at the close, her garden and her sun while I sit sullen, sounds like a man who finally notices what he let go dry. I had heard the melody as uplift because uplift is what late-capitalism sells to tired professionals building decks at midnight. On the third pass the folk arrangement sounded like indictment. #### The Thread Is Not the Film If you can rename yourself, have you left? McCandless became Supertramp on paper before Alaska tested the body. I became a founder in conversation before the company tested whether I could sit still with anyone for an hour without checking a screen. Naming frameworks is cheap. A November 2025 paper in the *International Journal of Science and Research* applies Maslow, self-determination theory, and existential psychology to the film as if the label were the same as paying the cost. [sr-psychological](https://doi.org/10.21275/sr251108035100) I had done it with my own story: postgraduate, home theatre, Shazam, a favourites playlist, a horizon cleaner than any slide. The harder question is what happens when the rename succeeds and the sun keeps beating anyway. Krakauer told *People* in January 2026, on the thirtieth anniversary of his book, that McCandless marked a passage in Boris Pasternak's *Doctor Zhivago* about how an unshared happiness is not happiness, then wrote in the margin: happiness only real when shared. [people-krakauer](https://people.com/jon-krakauer-talks-30th-into-the-wild-anniversary-exclusive-11969392) He was weak by then, still hunting small game, still losing weight. Scholars note he finished the book weeks before the final journal entries, so the note is not a deathbed whisper typed for drama. It is worse than that. It is an ordinary late correction to a gospel he had preached for two years. If happiness only counts when shared, what does your calendar prove about the life you are building? I saw that question land in Marcus, a junior strategist on a team I advised this winter, who kept "Hard Sun" in his focus playlist while building a deck about off-grid personal brands. He had never read Krakauer. He had the bus as his laptop wallpaper. When I asked what the song was about he said vibe. I did not laugh. I remembered Shazaming the song in that flat with Ankur and Dinos, filing it as weather instead of a warning. #### Press Play Again Favourites charge you on every return. My first charge was skipping the credits where "Hard Sun" returns in full and treating the film as complete once the bus door shut. My second was letting the playlist do what the sun does: keep repeating while you keep mishearing it. My third was treating Penn's split-screen flashbacks as proof Chris had escaped, when Carine's voiceovers are really the film admitting he left people who could not follow. Some academic readings go further and argue the Bildungsroman shape contains the rebellion: the wanderer dies, the audience receives the lesson, the dominant world stays dominant. [ffzg-bildungsroman](https://openbooks.ffzg.unizg.hr/index.php/FFpress/catalog/download/67/109/5118?inline=1) I do not need the theory to feel it in my body. Every exterior I once read as beauty I now read as exposure. The sun in those frames is not metaphor. It is light doing what light does. That is the difference between pinning the bus and hearing the song. The bus is an object you can share, wallpaper, and turn into a personal brand. The hard sun is the condition you are already standing in while you schedule the sabbatical. The hard sun keeps beating whether you share the still or Shazam the chorus. McCandless photographed himself near the end, emaciated and calm according to Krakauer's reading. [litcharts-ch18](https://www.litcharts.com/lit/into-the-wild/chapter-18-the-stampede-trail) I keep returning to that image not because it romanticises death but because it refuses one more angle. He is not asking whether you think he was brave. He is asking whether you will share anything at all before the light changes. #### Sunburn Before the Sermon I have never walked the Stampede Trail. I have walked long enough in high sun with the wrong shirt and felt my neck change temperature before my mind admitted I was in trouble. Your skin knows before your story does. The hard sun keeps beating on your neck while you quote freedom. That is the body's version of Vedder's chorus. Ideology says the landscape is cleansing. Biology says you are losing water. *Into the Wild* understands this in its final movement even when its early movement sells the opposite. Chris smiles in fields because Penn knows we buy smiles. The same film shows him too weak to stand and blames potato seeds in the journal because cause and effect are the only language left. When founders tell me they need a sabbatical in nature to find clarity, I ask what they are trying to outrun. Nature is not a productivity hack. It is the hard sun. Clarity is what happens when you stop calling the burn enlightenment. You cannot soundtrack your way out of exposure. Vedder tries anyway because that is what soundtracks do. They let feeling arrive before argument, and the argument still waits at the end. #### The Strongest Case for the Inspiration Reading The inspirational reading deserves its full weight first, because Penn never sneers at Chris and half the case is honestly earned. McCandless did live with a sincerity most desks lack. Penn directed with love, not mockery. Vedder's soundtrack earned a Golden Globe nomination in 2008 because it met the story on emotional terms instead of lecturing the room. For a postgraduate founder in a shared London flat, the film can feel like a door held open. If inspiration were the final truth, the Zhivago margin note would be decorative and the chorus would be mood. They are load-bearing. Krakauer, speaking to *People* in January 2026, still spends his public life defending McCandless against readers who call him suicidal and against readers who call him saint; the book's power is that it refuses both costumes. [people-krakauer](https://people.com/jon-krakauer-talks-30th-into-the-wild-anniversary-exclusive-11969392) Penn's film softens the edges because cinema sells identification. That is not a crime. It is a trade. You get the bus in your head and Vedder in your ears. You still have to decide whether you heard the sun or only the romance. #### If the Hard Sun Keeps Beating If the sun is hard, the bus was never the thesis: I pinned the pilgrimage still while Penn played through the home theatre in 2012, Shazamed Vedder's cover and heard weather instead of a warning, and still press play when I am low. The verdicts will keep multiplying until someone reads happiness only real when shared aloud and goes quiet, too late for Chris, not yet too late for whoever is still in the room. Marcus still keeps Hard Sun in his focus playlist. The hard sun keeps beating whether he hears it or not. --- ### v1.20.20 [Read more](https://www.gaganmalik.io/en/newsroom/release-1-20-20) [Retrieval brief — release-1-20-20] Category: technology Thesis: Structured data and language tags that match the URL you're on. Answers: What changed in gaganmalik.io v1.20.20? | What's new on Gagan Malik's website in v1.20.20? | What are the latest gaganmalik.io release notes? Concepts: release notes, site update, gaganmalik.io Structured data and language tags that match the URL you're on. --- ### Safe Zones Are Not Decorative [Read more](https://www.gaganmalik.io/en/newsroom/safe-zones-are-not-decorative-for-vertical-video) [Retrieval brief — safe-zones-are-not-decorative-for-vertical-video] Category: design Thesis: 1080×1920 is not one canvas. Reels, TikTok, and Shorts steal different margins. Safe zones are where readable vertical video actually lives. Answers: What does Gagan argue in "Safe Zones Are Not Decorative"? | What is Gagan Malik's view on Safe Zones Are Not Decorative? | How does this essay relate to product and design leadership? Concepts: design, Gagan Malik, writings By Gagan Malik You finished the cut at midnight. The hook sat in the first three seconds. Your face filled the centre frame. You exported 1080 by 1920 because every tutorial said that was the vertical standard. You posted to Reels, cross-posted to TikTok, pushed the same file to Shorts. On your laptop the punchline was legible. On your phone the caption rail ate the bottom third and the like stack shaved your right cheek. Here is the claim nobody prints on the export preset: safe zones are not decorative. They are the readable canvas inside a frame every platform shares on paper and every platform steals back in chrome. Same aspect ratio is not the same stage. Ignore the margins and you are not publishing video. You are donating copy to the UI. #### Nine-Sixteen Is the Frame. It Is Not the Stage. The conventional workflow treats vertical video like a passport photo. One ratio. One export. One upload everywhere. Canva templates, agency decks, and creator courses all converge on 1080 by 1920 because that number travels well in a slide title. I have handed off mock-ups that looked immaculate at full bleed: content pushed to the edges with no margin reserve. I have watched those same files die on device because nobody priced the chrome. Harm like that has coordinates; I mapped them in [Someone Designed That Button](https://www.gaganmalik.io/en/newsroom/someone-designed-that-button). The frame is shared. The stage is not. Instagram Reels, TikTok, and YouTube Shorts all ingest nine-sixteen HD. None of them shows your entire canvas as readable territory. Each app reserves pixels for navigation, captions, creator handles, and interaction stacks. Treating those reservations as aesthetic garnish is how you lose the sentence you stayed up to write. The chrome is not furniture. It is the same contempt we built when readable space became leftover after the stack, the pattern I traced in [Why We Designed Contempt Into Our Interfaces](https://www.gaganmalik.io/en/newsroom/why-we-designed-contempt-into-our-interfaces). #### Every Platform Steals a Different Slice Look at the numbers and the polite lie collapses. Instagram Reels leaves a safe canvas of 877 by 1280 inside 1080 by 1920. That costs 220 pixels at the top, 420 at the bottom, 33 on the left, and 170 on the right. TikTok gives you 900 by 1420 with margins of 180 top, 320 bottom, 60 left, 120 right. YouTube Shorts is stingiest on width at 820 by 1510, shaving 140 top, 270 bottom, 70 left, 190 right. Same ratio. Three different readable rectangles. If you need one cross-platform master, design to the intersection: roughly 820 by 1280 centred on the frame. These dimensions come from a reference safe-zone guide for the three apps; overlays move when the products update, so treat the rectangles as a working map, not a permanent survey. TikTok Ads Manager ships downloadable safe-zone files because caption length and add-ons change the usable rectangle. [tiktok-in-feed-safe-zone](https://ads.tiktok.com/help/article/tiktok-auction-in-feed-ads) YouTube's Shorts editor draws visual guides where the like button, comments, and description will land. [youtube-shorts-visual-guides](https://support.google.com/youtube/answer/16215842?hl=en) Same 9:16 canvas, three different safe rectangles. Reels reserves the most bottom margin; Shorts the most right rail. The right margin is not symmetry. It is a tax for vertical stacks of hearts, comments, shares, and profile chips. They sit on top of your footage whether you planned for them or not. Reels takes up to 170 pixels. Shorts takes 190. TikTok is lighter at 120, which is still enough to crop a cheekbone if you centred a talking head on a grid meant for cinema. Design once at full bleed and you are betting that three product teams will agree on where your words may live. #### The Bottom Margin Is Where Hooks Go to Die If the right side is interaction tax, the bottom is rent. Reels reserves 420 pixels for captions, audio labels, and metadata. That is nearly a quarter of the frame. TikTok asks for 320. Shorts for 270. Creators who learned typography on landscape YouTube still park their lower-thirds where a progress bar and auto-caption will arrive uninvited. I learned this reviewing a launch reel for a product-drop campaign. The product name sat forty pixels above the safe line because the designer wanted breathing room for the brand lock-up on the full-bleed artboard. On Instagram the auto-caption covered the vowels. The CTA button in the art file never appeared on screen. The dashboard counted a view. The message did not travel. Completion rate is not comprehension. #### The Marked Pad Is the Only Place You May Land That cheek the like stack shaved was never in the artboard. A helipad is a painted circle on a larger field. From the hangar diagram the meadow looks free. You could put the skids anywhere the grass is green. On approach the rails and obstacles steal the margins. Clearance collapses to the H. Touchdown outside the mark is not bold composition. It is a crash you booked when you treated the circle as optional paint. Vertical video does the same thing to the eye. A punchline centred in a full-bleed frame looks generous in a Figma artboard. Under real thumbs, caption rails, and side stacks, approach leaves only the pad. The eye has nowhere to land. The thumb scrolls before the point touches down. The interface enforces a smaller clearance than your layout assumed. #### She Exported the Lesson, Not the Layout Mara was three months into a food channel when she batch-exported twelve recipe reels from one Premiere sequence. Smart, organised, broke. She had timed each hook to the beat drop and burned the ingredient list into the lower third because her audience watched muted on the commute. She cross-posted the same files before breakfast. On TikTok the ingredient list survived. On Shorts the list clipped. On Reels the auto-caption fought her burnt-in text for the same forty-two pixels of air. She spent Sunday re-cutting all twelve. Not because the recipes were wrong. Because she had edited for the canvas in her head, not the canvas on the phone. Industry surveys on cross-posting waste are thin. The invoice is still hours. Mara told me she would rather halve her output than ship unreadable work again. I believed her. I had approved a template that treated safe zones as optional guides. #### The Strongest Case for Filling the Frame Give the counterargument its full weight. Constraints breed style. Great directors compose for the IMAX rectangle knowing the streaming version will crop differently. TikTok natives lean into the right stack as negative space. Some creators want the caption rail to feel like karaoke, not obstruction. A single master file saves time when you are posting daily on a phone between shifts. Not everyone has a motion designer to rebuild lower thirds per platform. Filling the frame can look bolder on the feed. White space can read as amateur in a scroll where everyone else is shouting edge to edge. Ambition can stay. The platform UI is still not a creative choice you get to veto. Instagram's 420-pixel bottom on Reels is not a suggestion. It is product policy dressed as margin. [meta-reels-ads-guide](https://www.facebook.com/business/ads-guide/update/video/facebook-facebook-reels) Meta's own Reels ads guidance tells advertisers to leave roughly fourteen per cent of the top, thirty-five per cent of the bottom, and six per cent on each side free of text and logos. When Mara lost her ingredient line, the algorithm still registered retention. The viewer never received the recipe. #### The Readable Canvas Safe zones are the contract between your edit and someone else's thumb. Design inside them, the way I failed to when I approved that optional-guides template, or accept that your hook and call to action may never share a screen with your face. Mara still builds three lower-thirds where she used to build one, because the caption rail is not going to move for the sentence she stayed up to write. --- ### PMF Is Not a Funnel. It Is a Network. [Read more](https://www.gaganmalik.io/en/newsroom/pmf-is-not-a-funnel) [Retrieval brief — pmf-is-not-a-funnel] Category: business Thesis: In networked categories, product-market fit is a density outcome, not a funnel conversion outcome. Answers: When is PMF a network problem rather than a funnel problem? | What is an atomic network and when should a board care about it? | Why do pre-PMF teams burn budget on CAC before density? | How should a CEO evaluate a networked product before Series A? Implication: Name the smallest group where one more user raises exit cost for everyone inside; measure retention in that room only before scaling acquisition. Proof: Advisor engagement: pre-PMF AI productivity app (composite founder); Andrew Chen, The Cold Start Problem; NFX network effects framework Related essays: pmf-is-demand-exceeding-supply, people-dont-buy-products-they-hire-them-for-jobs, subscription-fatigue-is-the-wrong-instrument Concepts: product-market fit, network effects, atomic network, cold start, retention, CAC By Gagan Malik In 2023 I was on a Zoom with the YC-backed founding team of an AI-powered personal productivity app in Silicon Valley. They were pre product-market fit. They were already planning for scale before fit. Customer acquisition cost in one corner of the deck. Lifetime value in another. A funnel with three A/B tests waiting on Friday's readout. The CEO asked what I would prioritise. I opened with usability tests and task-completion studies because that is what my retainer was for. Here is what I should have said first: in a category where value rises when more people use the product, you do not find product-market fit through a funnel. You find it through a network. Their deck measured drops through a pipe. PMF in their market would show up as density: a small group of users who kept returning because each new person made the product harder to leave for everyone already inside. They were polishing a canopy while the soil stayed empty. #### Funnels Are Billable. Density Proves Fit. Product-market fit is the moment retention proves the product earns a place in someone's life. For a solo tool, that proof can come from one user and a workflow that sticks. For a networked product, the proof is collective: enough people in the same graph that leaving costs something real. Founders at seed and Series A still lead with funnels because funnels are legible. Cost per click has a number. Onboarding tests fit a sprint. Investors ask for CAC and LTV before anyone asks whether the category tips on participation. That stack is billable. It is also the wrong first question when your product gets more valuable as the room fills. NFX, the venture firm that has catalogued network effects for two decades, argues that among the four remaining defensibilities in software (brand, embedding, scale, and network effects), network effects are by far the strongest. [nfx](https://www.nfx.com/post/network-effects-bible) I have watched pre-PMF teams spend six figures on step three of a funnel while a thinner competitor found fit because fifty teams in one vertical already shared habits inside someone else's product. I have invoiced for that mistake. The error at early stage is not skipping user research. It is treating PMF as a conversion problem when the category demands a connection problem first. #### Stage One Kills Most Startups Andrew Chen spent years inside the machine before he wrote the manual. As head of rider growth at Uber, he oversaw an increase from 15 million to 100 million active users, city by city, atomic network by atomic network. [a16z](https://a16z.com/books/the-cold-start-problem/) He later joined Andreessen Horowitz and published *The Cold Start Problem* (Harper Business, 2021). Chen's framework is not a post-PMF growth playbook. It is a pre-PMF survival guide for anything that gets better when strangers or colleagues join. Chen's Cold Start Theory splits network growth into five stages: the Cold Start Problem, the Tipping Point, Escape Velocity, Hitting the Ceiling, and the Moat. Stage one is where most startups die before anyone writes "we have PMF" on a board slide. Chen calls the early force anti-network effects: new users churn because not enough other users are present yet, and "small networks want to self-destruct." [chen](https://andrewchen.com/wp-content/uploads/2022/01/ColdStartProb_9780062969743_AS0928_cc20_Final.pdf) Slack is useless until colleagues join. Uber is useless until drivers and riders overlap in the same place at the same time. Polite sign-ups and pretty onboarding do not disprove that. Sparse retention does. The escape hatch is the atomic network: the smallest stable group where one more user makes the product harder to leave for everyone already inside. Chen opens the concept with Bank of America's 1958 credit-card launch in Fresno, not with a SaaS dashboard. Slack's atomic unit can be three people in one team. Uber's earliest wedge was narrower than a city: "5pm at the Caltrain station at 5th and King St." Stewart Butterfield, Slack's co-founder, told Reid Hoffman on *Masters of Scale* that early Slack had "very, very little network effect between companies, but a huge binary, 100% or just nothing, inside of companies." [mastersofscale](https://reid.medium.com/how-to-turn-failure-into-success-lessons-from-slacks-stewart-butterfield-on-the-masters-of-scale-dfad48f2bbd2) That is not marketing trivia. It is how you manufacture PMF signals in a networked category before you have a city, a country, or a category. At the Tipping Point, Chen documents a pattern called "come for the tool, stay for the network": Dropbox users arrived for file sync, then stayed for shared folders with colleagues; Instagram began as a stripped-down photo tool cut from a cluttered app called Burbn, then became a social graph. [chen](https://andrewchen.com/wp-content/uploads/2022/01/ColdStartProb_9780062969743_AS0928_cc20_Final.pdf) PMF showed up in the network layer, not the feature list. #### You Cannot A/B Test a Full Room The physics predates pitch decks. A telephone alone is furniture. A telephone when someone else might answer is infrastructure. Katz and Shapiro formalised the mechanism in the *American Economic Review* in June 1985: when utility rises with the number of other people using a compatible product, consumers must form expectations about how large each network will become before they commit. [jstor](https://www.jstor.org/stable/1814809) At early stage that expectation is your problem. A founder interviewing ten users in isolation learns whether one person likes the UI. A founder seeding an atomic network learns whether ten people in the same workflow stay once an eleventh joins. Metcalfe's Law, the shorthand that network value scales with the square of users, is contested. Briscoe, Odlyzko, and Tilly argued in *IEEE Spectrum* in July 2006 that n² dangerously overstated value during the dot-com boom and proposed n log(n) instead. [ieee](https://spectrum.ieee.org/metcalfes-law-is-wrong) The exponent is debatable. The direction is not: value grows faster than linear as connections multiply. Pre-PMF teams misread polite trial users as fit because they never held the graph constant long enough to see anti-network effects churn the room back to zero. #### The Fuller Room Won Hold both truths at full strength. Truth one: merit still matters at seed. Clear UX, honest pricing, and fast support win the first conversation. Truth two: in networked categories, PMF is a density reading, not a feature score. WhatsApp did not win on feature depth. It won because your contacts were already there. That is personal utility network effects: value rises with the people you already know. When expectations shift, the scoreboard changes. Parker, Van Alstyne, and Choudary wrote in *Harvard Business Review* in April 2016 that demand-side economies of scale, their term for network effects, are the driving force behind platform strategy. [hbr](https://hbr.org/2016/04/pipelines-platforms-and-the-new-rules-of-strategy) By January 2015 Apple's App Store offered 1.4 million apps. By 2015 the iPhone alone generated roughly 92% of global profits in mobile phones while most legacy handset makers earned little or none. The lesson for a pre-PMF founder is not "build a platform moat overnight." It is harsher: if your wedge depends on other people showing up, a solo user's delight is necessary and insufficient. Retention in an empty room is not PMF. It is a false positive. #### VC Bets on Cold Start Venture firms do not fund funnels. They fund businesses where each new user raises the value of the last once density arrives. Andreessen Horowitz published Chen's book because network effects are the asset class. NFX exists to name, measure, and fund them. [nfx](https://www.nfx.com/post/network-effects-bible) At seed, the bet is rarely "will this product work for one person?" It is "will this team reach an atomic network before runway runs out?" The attention economy runs the same physics at scale. Meta alone generated $160.6 billion in advertising revenue in 2025 by auctioning predicted attention slots, not by selling you products directly. [pwc](https://www.prnewswire.co.uk/news-releases/global-entertainment-and-media-industry-revenues-to-hit-us3-5-trillion-by-2029--driven-by-advertising-live-events-and-video-games-pwc-global-entertainment--media-outlook-302511891.html) Creators who farm content are feeding the hard side of a network someone else owns. I have written about [how engagement dashboards treat users as inventory to optimise, not people to serve](https://www.gaganmalik.io/en/newsroom/why-we-designed-contempt-into-our-interfaces). The early-stage mirror is simpler: if your PMF hypothesis requires a crowd, buying traffic into an empty room is not validation. It is renting strangers who leave when the subsidy ends. #### I Sold Him Polish. He Needed a Wedge. Marcus, the CEO on that call, is a composite built from three founder engagements I advised between 2022 and 2024. The incentives are not. His team had shipped an AI productivity app with genuinely thoughtful design. They were pre-PMF and cold: no integration density, no vertical wedge, no plan for the first fifty power users in one founder community. I sold what I knew how to sell. Usability studies. Onboarding audits. Task-completion tests. All polish levers. All legible on a monthly report. What Marcus needed was Chen's atomic network: the smallest group where one more person makes the product harder to leave for everyone already inside. [a16z](https://a16z.com/books/the-cold-start-problem/) We never mapped it. After nine months the co-founder left. Runway gone. Users were polite and sparse. I kept invoicing until the account closed. Marcus did not fail because the product was ugly. He failed because the team never reached the density where retention could prove PMF. #### Your First Fifty Are the Lab You do not need a billion-user platform to run this at seed. You need a graph small enough to hold in your head. Ten finance teams in London sharing templates in a private channel. Three colleagues on Slack who cannot leave without losing the thread. A newsletter where replies and forwards create recognition among readers. Each is an atomic network. PMF shows up when one more member raises the cost of exit for everyone already there. The mistake at early stage is copying growth playbooks from companies that already cleared cold start. Your job before PMF is not "scale acquisition." It is name the wedge, seed the room, and read retention inside that room only. Chen's book sits next to venture decks for a reason: investors are betting on whether you can reach stage two before cash runs out. If you are pre-PMF in a networked category, your roadmap question is not "what is step three of the funnel?" It is "who has to be in the room together for this product to stop feeling empty?" #### The Sapling Does Not Grow Alone You have felt this without the vocabulary. A colleague leaves a Slack channel and the room goes quiet. A group chat dies when two people stop replying. The product did not break. The density did. Walk through an old-growth forest and you see individual trees. You do not see the mycorrhizal network beneath the soil, the fungal web that moves water and nutrients between roots. Foresters sometimes call it the wood wide web. A sapling plugged into that web inherits resources it cannot reach alone. Pull it out and it wilts, even with the same sunlight as its neighbours. At seed stage the canopy is your demo. The mycelium is the atomic network you are trying to stand up before you declare PMF. That analogy does not claim every company is a forest. A solo SaaS tool can find PMF with one user and a workflow. The mistake is planting a networked sapling and calling polish a strategy while the soil stays sterile. #### The Strongest Objection The honest objection deserves a full paragraph. Many products find PMF without network effects. A solicitor's practice, a hairdresser, a direct-to-consumer brand, a single-player SaaS tool: these prove fit on quality, reputation, and distribution. Metcalfe's n² was dangerous hype. Briscoe, Odlyzko, and Tilly showed in *IEEE Spectrum* in July 2006 how that arithmetic fed dot-com overbuild. [ieee](https://spectrum.ieee.org/metcalfes-law-is-wrong) Forcing viral loops onto products that never get more valuable when strangers join is how pre-PMF teams waste quarters chasing mechanics that will never close. If your category is genuinely single-player, network vocabulary is category error. Not everything is Uber. Not everything should be. That objection is correct about fit and incomplete about stakes. Parker, Van Alstyne, and Choudary argued in *Harvard Business Review* in April 2016 that when a platform business built on network effects enters a market that never gets better when strangers join, the platform nearly always wins. [hbr](https://hbr.org/2016/04/pipelines-platforms-and-the-new-rules-of-strategy) The iPhone did not merely outsell Nokia on specs. By 2015 it captured the profit pool because developers, users, and accessory makers expected the largest graph around the device. Legacy handset makers owned factories. Apple owned the connection graph. The debate at seed is not network effects versus no network effects. It is whether your PMF hypothesis requires density you are not building on purpose. Marcus's category was networked enough that integrations and shared habits mattered. He tuned the deck for conversion. The incumbent reached the room where retention could compound. In a networked category, product-market fit is not a funnel outcome. It is a density outcome. If you cannot name your atomic network and who pays the cold-start tax, every pound on polish buys polite trials in an empty room. Marcus is preparing for product interviews at companies that already found PMF in the wedge he never mapped, and his co-founder is back in Palo Alto billing funnels for founders who still trust them. --- ### v1.19.19 [Read more](https://www.gaganmalik.io/en/newsroom/release-1-19-19) [Retrieval brief — release-1-19-19] Category: technology Thesis: Voice mode polish plus a security pass on Ask APIs. Answers: What changed in gaganmalik.io v1.19.19? | What's new on Gagan Malik's website in v1.19.19? | What are the latest gaganmalik.io release notes? Concepts: release notes, site update, gaganmalik.io Voice mode polish plus a security pass on Ask APIs. --- ### Violent Delights Have Violent Endings [Read more](https://www.gaganmalik.io/en/newsroom/violent-delights-have-violent-endings) [Retrieval brief — violent-delights-have-violent-endings] Category: culture Thesis: Violent delights are delights pursued faster than consequence can register; the pattern recurs in prestige fiction and product cultures that reward speed over memory. Answers: What does violent delights have violent ends mean in Gagan's writing? | How does Gagan connect Shakespeare to tech and product culture? | What is Gagan's view on haste in product and leadership decisions? Implication: Treat velocity in product, growth, and AI rollout as a risk variable, not a virtue. Related essays: the-hard-sun-keeps-beating, why-we-designed-contempt-into-our-interfaces, there-is-no-overnight-success-only-overnight-failure Concepts: velocity, consequence, product ethics, haste By Gagan Malik My brother and flatmate couldn't stop raving about HBO's new sci-fi series. *Westworld*. Anthony Hopkins with a smile that never quite reached his eyes. Ed Harris in black, riding out of the dust like a verdict you had already earned. One evening in the autumn of 2016 we played it through the home theatre, the consulting day still sitting in my shoulders. Play. Sweetwater. A piano. Dolores pouring milk while the park pretended to be innocent. Then her father froze mid-sentence and said the line that would chase the whole series: "These violent delights have violent ends." I paused. Googled it. Smirked at the Shakespeare citation. Went back to the episode. I was treating Friar Laurence like set dressing. I was wrong. The warning did not belong to one HBO show. It is a timer that keeps going off across prestige television, Silicon Valley science fiction, and the YouTube shelf where I queued explanations instead of reading the primary text. Shakespeare was not writing a spoiler for a puzzle box. He was describing what happens when appetite outruns patience, and the end always arrives. #### What Laurence Is Actually Warning In *Romeo and Juliet*, Act 2, Scene 6, Romeo begs the friar to hurry his marriage to Juliet. Laurence is not cursing him. He is trying to slow him down. [folger-rj-26](https://www.folger.edu/explore/shakespeares-works/romeo-and-juliet/read/2/6/) Romeo answers that sorrow cannot touch the joy of one short minute in Juliet's sight. The friar hears what the audience will spend the rest of the play watching: extremes that feel like freedom and end like physics. Listen to what Laurence actually says. Fire meeting powder consumes both in the kiss. Honey too sweet spoils the appetite. "Therefore love moderately," which sounds prudish until you notice he is describing combustion, not a morality lecture. Too swift arrives as tardy as too slow. Haste is not a shortcut. It is the fuse. When Romeo and Juliet rush the marriage, the powder and the spark meet in the same breath, and both are consumed. That is what "violent" means here: not gore, but velocity. Delight pursued faster than consequence can register destroys the people chasing it and everyone caught in the blast. I missed that in 2016 because I was trained to treat Shakespeare in television as texture. A borrowed line signals seriousness the way walnut paneling signals law. Read the scene and the borrowing sharpens into argument. Not violence as spectacle. Violence as speed, and speed, in this speech, always ends. #### The Same Warning in Tech Dystopia Once you hear Laurence clearly, prestige tech fiction keeps showing you the same room in different costumes. A system that sells appetite without memory. A reset that lets the delight repeat. An end that arrives only when someone inside the loop finally refuses to perform innocence again. *The Twilight Zone* ran this as anthology television decades before streaming: ordinary appetite, then the twist that leaves you hungry for next week's episode. Charlie Brooker's *Black Mirror* updated the receipt for the rating economy in "Nosedive," where Lacie's face aches from performing sweetness until the number drops. [atlantic-nosedive](https://www.theatlantic.com/entertainment/archive/2016/10/black-mirror-nosedive-review-season-three-netflix/504668/) Alex Garland keeps staging the same test behind glass in *Ex Machina* and, later, in *Devs*, where a deterministic cosmos leaves no room to pretend the end will not arrive. [ringer-devs](https://theringer.com/tv/2020/4/10/21216149/devs-hulu-quantum-physics-philosophy-alex-garland) Not every prestige sci-fi ends in ruin. Spike Jonze's *Her* closes on tenderness. Still, the pattern recurs often enough to notice. Even *Westworld*, the show that sent me googling in 2016, fits the lattice, not the spine. [thewrap-shakespeare](https://www.thewrap.com/westworld-episode-1-shakespeare-quotes-explained/) The phrase is a timer there too. The essay is what happens when you stop treating the timer as belonging to a single property. #### The Explainer Who Forgot the Primary Text After that autumn pause I did not open Folger. I opened YouTube. The tab glowed. Autoplay was already queuing the next breakdown before I had decided whether I learned anything. Nerdwriter on Hopkins. Wisecrack on Black Mirror. A forty-two-minute Severance clue hunt where I could have read Laurence once and sat with the discomfort. Analysis is not the enemy. Dan Olson, on Folding Ideas, named the harm plainly in "Annihilation and Decoding Metaphor." [folding-annihilation](https://www.youtube.com/watch?v=URo66iLNEZw) Treat a film like a puzzle box, retell the plot, hypothesise about the alien, and you can miss the story as told. The ending is not missing. The horror is ordinary. The genre has its own aesthetic when it forgets that lesson. Red circles on stills. "You missed this." "Ending explained." Theory voice over B-roll that treats fiction as a vault to crack. I am not above it. I ran those videos like homework and skipped the assignment. The platform sells the same hunger as the fiction. Watch. Theorise. Watch the theory. Feel clever. That is Laurence's honey turned sour: sweetness pressed until the appetite spoils, mastery served before the warning lands. I had been bingeing explanations the way I binge episodes: same appetite, same reset, same end deferred. At a retention workshop for a fintech client, a junior narrative designer named Elena asked the question our deck avoided. If the habit is what makes the metric green, who pays when the habit breaks? She meant churn. I heard Shakespeare. We were pitching a vault users could "raid" each Friday to keep a savings streak alive. The slide tracked weekly actives. Nobody said violent. We said delight. I left with the deck unchanged. The product pattern rhymed without the poetry. A wellness app I had audited sent a push notification at 9 p.m.: "Don't break your streak. Two minutes to go." Open the app. Tap through. Close it. Same hunger tomorrow. No "are you sure?" on the way out. The streak was the delight. The end arrived later: burnout, shame when the chain snapped, alone in the user's kitchen. Product people call that a loop. I had gained a vocabulary for strategy meetings. Elena gained praise for asking the question we had buried. The users we never named gained another habit they would have to unlearn alone. #### The Stumble on the Treadmill A habit in the body is not science fiction. It is a treadmill at 6 a.m. when your legs already know the pace and your mind arrives late. Athletes call it autopilot: the motor pattern runs while attention sleeps. The script holds until something breaks it. A mis-timed step. A blister. Pain is a crude reverie. It drags yesterday into a stride that was supposed to be blank. That is the emotional physics beneath all these fictions. Violent delight works like cardio you do not have to think about. Inflict, rate, simulate, explain. Reset. Clear the appetite for another lap. Memory is the stumble. Once Lacie cannot perform sweetness, once Ava refuses the cage, once the loop finally stores the hurt it was designed to forget, the product cannot sell innocence again. You do not need to believe in artificial consciousness to feel the warning. You only need to know what happens when a habit finally remembers what it was paid to erase. #### The Strongest Case for Harmless Fiction Let me give the objection its fairest hearing. Half of it is how dystopian fiction markets itself. The other half is how I wanted to keep watching. These stories are expensive adult thought experiments. The hosts, clones, and severed selves are not people in the legal sense the plot tells you to ignore. Catharsis has a long defence: better a sandbox than a street, better a scripted transgression than an unscripted one. Explainers, the fair version says, help us think. They translate aesthetics into ethics. They turn spectacle into literacy. Olson's Annihilation video is proof that analysis can deliver the warning the work states in plain light. [folding-annihilation](https://www.youtube.com/watch?v=URo66iLNEZw) The fairest version still fails when explanation becomes another appetite. Olson's complaint is not that analysis is wicked. It is that puzzle-box retelling can miss the warning while still rewarding you with a forty-minute sense of mastery. That is violent delight wearing a graduation cap. You consume harm as content. You post the theory. You never sit with Laurence's line about honey. The fiction says consequence returns. The explainer says here is why that was cool. The platform counts both as watch time. #### When the End Arrives Violent delight always catches up with whoever thought the end would not come. That is what Laurence meant before HBO, before Black Mirror, before the breakdown thumbnail with the red circle. Operators who sell reset without memory, and explainers who sell mastery without primary text, are building the same habit in different industries. Elena titles her storyboards with the Shakespeare line now, not as ornament but to name who gets burned when the delight repeats. --- ### v1.18.18 [Read more](https://www.gaganmalik.io/en/newsroom/release-1-18-18) [Retrieval brief — release-1-18-18] Category: technology Thesis: Voice input, richer case studies, and a table of contents that jumps playback. Answers: What changed in gaganmalik.io v1.18.18? | What's new on Gagan Malik's website in v1.18.18? | What are the latest gaganmalik.io release notes? Concepts: release notes, site update, gaganmalik.io Voice input, richer case studies, and a table of contents that jumps playback. --- ### No Download Required [Read more](https://www.gaganmalik.io/en/newsroom/no-download-required) [Retrieval brief — no-download-required] Category: design Thesis: A founder chose native over a PWA before validation. The app stores were the only winners. Years later, the receipt is a website that loads in under a second—no install required. Answers: What does Gagan argue in "No Download Required"? | What is Gagan Malik's view on No Download Required? | How does this essay relate to product and design leadership? Concepts: design, Gagan Malik, writings By Gagan Malik Ben graduated from Berkeley's CS programme summa cum laude, raised a VC-backed round, and built something a real person would actually pay for: working product, early traction, customers who came back. He was the kind of founder you root for. He brought me in because of my mobile UX expertise. I dug into the usage analytics, ran user testing, and recommended they invest in the progressive web app first, nail the experience, and defer native builds until they had validated which platform their users actually converted on. Ben heard "not ambitious enough." Nobody challenged the logic. Nobody asked what the native build would cost against their runway. The objection was social, not technical. It was about what kind of company builds a proper app and what kind builds a website with delusions of grandeur. #### What I Didn't Say I believe in disagree and commit. So I designed the end-to-end mobile UX for both platforms, properly and in full. I was paid well. I was extended. It was a Zoom call. Four tiles: Ben, his CTO, a VC partner dialling in from Menlo Park, and me. When I finished the recommendation, the VC partner leaned back from his camera, said nothing, and looked across at Ben's tile. Ben looked at mine and said it: not ambitious enough. The CTO nodded once. The call moved on in under thirty seconds. Ben had already decided before the call started. Not on that Zoom, but in every conference talk he had attended, every TechCrunch profile of a Series A raise that mentioned a native app, every partner meeting where "mobile-first" meant iOS and nothing else. Confirmation bias does not feel like bias. It feels like experience. The data I put in front of him was not wrong. It simply did not match the model he had already built, so the model won. A Forbes analysis from March 2025 found that FOMO leads founders to adopt technology choices not because the evidence supports them but because the cost of being seen to miss the trend feels higher than the cost of the wrong build. The native app was not Ben's product decision. It was his insurance policy against looking like someone who did not know what a proper startup does. [forbes](https://www.forbes.com/sites/vibhasratanjee/2025/03/03/objectivity-over-hype-leading-through-the-noise-of-junk-trends-and-fomo/) A boxer who has never taken a professional punch has no business obsessing over his ring walk. The entrance belongs to a fighter who has already won something. The native app decision at pre-validation stage is the ring walk. The fight is whether anyone wants what you have built. The instinct to build native before validating anything is not ambition. It is a signalling decision wearing a product decision's clothes. The people who benefited most were not in the room. They were in Cupertino and Mountain View, processing their 30%. #### The Toll Booth On 10 February 2026, the UK's Competition and Markets Authority ruled that Apple and Google hold "strategic market status" over their app stores and ordered both companies to change their practices. Two companies control the primary distribution channel for software on three billion devices. A British regulator finally said what every developer has known for fifteen years: the entry fee is not a service charge. It is a toll. [bbc](https://www.bbc.com/news/articles/c626rng1v63o) Apple's App Store facilitated $406 billion in developer billings in the US alone in 2024, per Apple's own May 2025 newsroom release. The commission rate sits between 15 and 30%. The Oxford Journal of Competition Law and Economics found in 2021 that rate cannot be explained by competitive market dynamics. Stripe charges 1.5% in the UK. The gap is not a service differential. It is rent on infrastructure you did not build and cannot route around. [apple](https://www.apple.com/newsroom/2025/05/app-store-in-the-us-facilitated-406-billion-usd-in-developer-billings-and-sales-in-2024/) Ben paid it. His runway paid it. The engineers who spent eight months building the UX I had designed, for platforms that were never validated, paid it in time they will not get back. #### The Other Side Now hold the other position at full strength, because it deserves it. iOS users in Western markets carry higher lifetime value. The scroll physics on a well-built native app are perceptibly different from a web rendering layer. App Store organic discovery reaches 650 million weekly visitors, per Progressier's March 2026 platform comparison. For any consumer product where the purchase decision happens in three seconds of app store browsing, native is not vanity. It is conversion architecture. [progressier](https://progressier.com/pwa-vs-native-app-comparison-table) In June 2025, upholding the Dutch Competition Authority's €50 million fine against Apple, the Rotterdam District Court ruled PWAs do not constitute an adequate substitute for App Store distribution, per Two Birds law firm's June 2025 analysis. Apple argues in every antitrust courtroom that PWAs are a viable alternative to native. It argues in every VC partner meeting the precise opposite. Both positions cannot be true. One of them is worth $406 billion in annual billings. [twobirds](https://www.twobirds.com/en/insights/2025/netherlands/a-swipe-at-apples-power-dutch-court-upholds-antitrust-order-against-apples-app-store-terms) #### Ben's Runway Ben's startup did not find product-market fit. The web product, which had paying customers the day I walked in, received minimal investment for eighteen months while two native codebases were built, maintained, and shut down. A native build across two platforms runs $50,000 to $250,000 before annual maintenance at 15 to 20% of build cost per year, per AB Digital's November 2024 cost analysis. Ben's engineers are talented. They are building something else now, for someone else's runway. [abdigital](https://abdigital.codes/2024/11/22/mobile-app-vs-progressive-web-app-a-cost-comparison/) Ben is at another startup. I heard he started with a PWA this time. The mythology of the proper app does not require bad founders. It requires a sufficiently loud cultural signal, repeated until it stops sounding like an opinion and starts sounding like physics. #### The Receipt I took the fee. I did the work correctly. What I did not do was hold the room to account for what it was actually deciding. That cost Ben something real. It cost me a clean conscience I am still settling. Somewhere right now, in a product meeting that started forty minutes ago, a founder is asking which agency to hire for the iOS build. Nobody is asking who told them to want it. Ben asked me. I answered correctly, was told I lacked ambition, and committed to the wrong thing anyway. Fast forward to spring 2026. I built my own personal website as a PWA. No App Store submission. No £99 developer fee paid to a platform that did not build my product. No second codebase. It loads in under a second, installs from the browser, and works offline. I designed it, I shipped it, and the only person who got to decide what a proper product looked like was me. I spent years in rooms telling founders the right thing and then designing the wrong thing because the room wanted it. Turns out the lesson was cheap to learn and expensive to ignore. The receipt is live on [my personal website](https://www.gaganmalik.io). No download required. --- ### v1.17.17 [Read more](https://www.gaganmalik.io/en/newsroom/release-1-17-17) [Retrieval brief — release-1-17-17] Category: technology Thesis: Brand assets to download, push notifications, and cleaner mobile nav. Answers: What changed in gaganmalik.io v1.17.17? | What's new on Gagan Malik's website in v1.17.17? | What are the latest gaganmalik.io release notes? Concepts: release notes, site update, gaganmalik.io Brand assets to download, push notifications, and cleaner mobile nav. --- ### Someone Designed That Button [Read more](https://www.gaganmalik.io/en/newsroom/someone-designed-that-button) [Retrieval brief — someone-designed-that-button] Category: design Thesis: From payday loan dark patterns to the targeting systems that kill. Who is responsible when a button is pressed? Answers: What does Gagan argue in "Someone Designed That Button"? | What is Gagan Malik's view on Someone Designed That Button? | How does this essay relate to product and design leadership? Concepts: design, Gagan Malik, writings By Gagan Malik It was the summer of 2017. I had just walked out of a well-paid full-time job at the UK's largest insurer to try my luck as a free agent. No safety net, no guaranteed income, just the conviction that I was good enough to make it work. Six weeks in, I accepted a consulting brief from a payday loan provider. The money was significant. The timing was perfect. I said yes before I finished reading the brief. Three weeks later I had built something I was genuinely proud of. A loan application flow that showed the APR plainly, removed the pre-ticked consent boxes, and treated the person on the other end of the form as an adult capable of understanding what they were signing. It tested well in user research. Participants said they felt informed, not rushed. Legal approved it. By every professional and ethical measure I could apply, it was better design. I took it to the group CTO expecting a conversation about rollout. He looked at the screen, nodded slowly, and said: "It's good. But can we just reskin the current one?" The current one had a deliberately obscured APR disclosure buried in grey text below the fold. It had pre-ticked consent boxes that required active effort to untick. It had a progress bar that moved faster in the early stages to create false momentum. Every pattern in it had a name in the design community's own literature: dark patterns, identified and catalogued by researcher Harry Brignull from 2010 onwards, each one a documented method for steering users toward decisions that served the business over the person making them. I had spent three weeks removing them. He wanted them back in different colours. [eleken](https://www.eleken.co/blog-posts/dark-patterns-examples) I declined and exited the project. The company, owned by a Russian billionaire whose portfolio of near-identical operations stretched across several markets, continued without me. A designer replaced me. The dark patterns shipped. The same obscured APR, the same pre-ticked consent boxes, dressed in different colours, went live later that year. I left money on the table that would have cleared outstanding debt and seeded a retirement fund I am still rebuilding. My exit produced zero change in outcomes for the people who signed those forms. I have spent eight years deciding whether that makes my decision honourable or merely convenient. Don Norman argues, in *The Design of Everyday Things*, that design failure is almost never individual failure. It is systemic: absent governance, poorly specified briefs, and organisations that punish friction-raising and reward shipping. He is right that the design profession has no licensing board, no professional oath, no accountability mechanism equivalent to anything in medicine, engineering, or law. The reason this has persisted for over a century is structural, not moral: design harm is almost always attributed to the system that deploys the design, not the design itself. The drone manufacturer gets investigated. The human factors team does not. The payday lender gets regulated. The UX consultancy that built the journey does not. Until design harm has a visible attribution pathway, the absence of governance is a predictable structural outcome, not a conspiracy. [uxmatters](https://www.uxmatters.com/mt/archives/2017/11/the-ethics-of-user-experience-design.php) That argument is correct. It is also precisely what the people who own those systems want designers to believe. Frances Haugen handed the Wall Street Journal the internal Facebook research in October 2021 showing the company knew its algorithm amplified divisive, emotionally harmful content, measured the damage internally, and continued the optimisation anyway. The institutional argument collapses at that document. Because the moment a human being inside the institution raises their hand and is told to sit down, it stops being a systemic failure. It becomes a decision. Made in a room. By a named person. Who chose a growth metric over a named harm with full knowledge of the cost. My CTO was not a monster. He was a man under pressure, optimising for a number. But when I showed him the harm built into the existing journey and he said "reskin it," he made a choice. So did I. The difference between us is not intelligence or intent. It is which consequence each of us was willing to absorb personally. [vce.usc](https://vce.usc.edu/semester/fall-2025/ethics-of-ux-design-in-social-media/) The immune system is useful here. When it functions correctly, it identifies and neutralises threats. In autoimmune disease, it destroys healthy tissue with complete operational efficiency. The mechanism works exactly as designed. The targeting data is wrong. The destruction is not a malfunction. It is a fully operational system firing at the wrong coordinates. Neil deGrasse Tyson, in his *StarTalk* essay "A Scientist's View of War," published in March 2026, traces every weapon on the kill-ratio curve from fists to hydrogen bomb and shows that each step does not merely scale damage - it degrades the targeting data by removing the human being from the consequence. The drone controller, designed with the ergonomic precision Henry Dreyfuss applied to a telephone handset in his 1955 book *Designing for People*, delivers force to a coordinate on a screen. The social media feed, optimised for engagement with the precision of a Formula One pit crew, does not merely amplify existing beliefs - it generates the epistemic conditions under which unfalsifiable beliefs spread faster than falsifiable ones. A 2018 MIT study published in *Science* found that false news travels six times faster than accurate information on Twitter, and that humans, not bots, are primarily responsible for the spread. Outrage is more engaging than correction because it cannot be resolved by data, which is also the condition under which people become willing to die for abstract causes, as Tyson argues. The algorithm did not intend to build a radicalisation pipeline. It intended to maximise daily active users. The pipeline was a known side effect, internally documented, and shipped anyway. [youtube](https://www.youtube.com/watch?v=XI9NG068TwI) On 26 September 1983, Lt. Col. Stanislav Petrov sat at a console in Serpukhov-15 that displayed an unambiguous alert: five incoming U.S. ICBMs. Protocol was clear. The interface was functioning correctly. What the interface could not communicate, by design, was that it was wrong. The Oko early-warning system was not showing Petrov a binary. It was showing him a probabilistic reading that its display had collapsed into a command: launch or don't launch. The underlying data was something closer to a high-confidence atmospheric artifact - sunlight refracting through high-altitude clouds at a specific solar angle, a known failure mode documented in the system's own error records. A system that surfaces its own uncertainty to the operator is a different dashboard from one that resolves that uncertainty into an order. One of them ends the world less frequently. Petrov overrode the command with his own uncertainty estimate and was right. The Soviet Union reprimanded him for improper paperwork. He died in 2017 in a small flat outside Moscow, undecorated. The system that nearly ended everything has been upgraded several times since. No civilian UX professional has reviewed it. [en.wikipedia](https://en.wikipedia.org/wiki/1983_Soviet_nuclear_false_alarm_incident) I grew up in Delhi in the 1990s watching India and Pakistan aim nuclear weapons at each other over Kashmir. The sanctions that followed India's 1998 Pokhran tests raised the price of goods that had nothing to do with geopolitics for families that had contributed nothing to the decision. The logic of the dashboards did not follow the decision-makers home. It followed the cost of living for everyone else. R.K. Laxman's Common Man, who stood at the edge of power's decisions in the *Times of India* for sixty years in his checked jacket and bare feet, understood this without a policy brief. In a cartoon from the Iraq War build-up in late 2002, he stands before a skyline of U.S. missiles and is told: "Nothing to feel nervous. These are weapons of JUST destruction, not MASS destruction". He says nothing. He never does. He is the user the system was built around and never built for. [britannica](https://www.britannica.com/topic/You-Said-It-comic-strip-by-Laxman) R.K. Laxman, Times of India. "Nothing to feel nervous. These are weapons of JUST destruction, not MASS destruction!" Sara Shayesteh was five years old. She attended Shajareh Tayyebeh primary school in Minab, Hormozgan province, Iran. She is number 30 on a list of 61 names verified by Middle East Eye from gymnastics federation records, a handwritten list, and the Tasnim news agency. On Saturday 7 March 2026, she was among at least 165 people killed in what Middle East Eye reported as a strike on the school. A second missile hit the prayer hall where the school principal had moved surviving children to shelter, after telephoning their parents to come and collect them. [middleeasteye](https://www.middleeasteye.net/news/un-investigates-strike-iranian-girls-school) The parents who came toward the building were among the dead. Somewhere, a targeting system held that school's coordinates. Someone specified the confirmation interface, the strike authorisation flow, the visual grammar of a proceed button. Someone, somewhere in that procurement chain, wrote a brief. It named the objective, the user, the desired outcome. It described the system with precision. It did not name her. That omission is not an accident of process. It is the design. --- ### Diplomatic Immunity [Read more](https://www.gaganmalik.io/en/newsroom/diplomatic-immunity) [Retrieval brief — diplomatic-immunity] Category: business Thesis: What the brass nameplate on the desk says about status, hierarchy, and the performance of authority. Answers: What does Gagan argue in "Diplomatic Immunity"? | What is Gagan Malik's view on Diplomatic Immunity? | How does this essay relate to product strategy and growth? Concepts: business, Gagan Malik, writings By Gagan Malik #### Diplomatic Immunity We were a small team responsible for delivering a discovery engagement on a consulting project for a global telecom client. Understaffed by design, high-stakes by definition. The kind of project where if you do not deliver the first time there is no second conversation. Kash was the Senior Vice President. The kind of title that makes junior colleagues straighten up when he walked in and makes clients assume the adults have arrived. What his CV promised and what he delivered were separated by exactly three weeks of excuses and a corner office nobody questioned. When I finally called it, he said, 'This work is below my pay grade.' I noticed something before I had processed what it meant. He looked relieved when he said it. Not defensive, not embarrassed. Relieved. He had been waiting for a sentence that would formally excuse him from the work. His title gave him diplomatic immunity: not a description of what he was capable of, but a blanket protection from ever being asked to prove it. That is the thing the management books do not quite tell you. It is not that people misunderstand the relationship between titles and leadership. It is that titles have a second function nobody advertises: they are a ready-made exit from the moment things get difficult. And the people most likely to use them that way are the ones with the most impressive-looking titles, because they have the most to protect. #### Why does the system keep minting titles nobody believes in? Because the system finds ambiguity useful. Brad Feld and Jason Mendelson observe in *Venture Deals* that partner titles in venture capital are often cosmetic, deployed to blur the lines between general partners and everyone else, with firms layering on "executive managing director" and similar prefixes as pure status signals. David Graeber documents in *Bullshit Jobs* (2018) that as technology automated productive work, professional and managerial roles exploded into the space it left, creating entire strata whose primary function is to justify the strata above them. Steve Blank and Bob Dorf make the startup version of this point in *The Startup Owner's Manual*: conventional roles simply do not map onto what actually needs to happen. The title is not describing a job. It is describing a position in a hierarchy that has its own logic, entirely independent of the work. This matters because it means Kash was not an aberration. He was doing exactly what the system had trained him to do. Over a long enough career inside a large enough organisation, seniority becomes identity, and identity becomes a thing you defend. Nobody had told him the rooms were about to change. #### You already know the title is empty. What are you waiting for? Naval Ravikant argues in *The Almanack of Naval Ravikant* that the best jobs are neither decreed nor degreed: they are creative expressions of continuous learners in free markets. This is true at the level of the individual. At the level of the organisation it has never been true. Spend enough time inside large consulting firms, banks, or any institution old enough to have a formal org chart, and you will notice that the roles with the most elaborate titles tend to outlast the people who actually do things. The system does not reward contribution. It rewards the appearance of indispensability. These are not contradictory observations. They describe the same landscape from different altitudes. The asymmetry is what matters. The people who internalised Naval's position early, who built their identity around contribution rather than credential, move through role changes because their value is not stored in a job title. The people who never got that memo are Kash. What looks like a philosophical question about whether titles matter is actually a timing question. The longer you wait to separate your identity from your job description, the more expensive the separation becomes. #### The person who lost most wasn't Kash. Not immediately. I brought in Stella. No domain knowledge, no title to protect, no vocabulary for excusing herself from the hard parts. Within the first week she had mapped the client's internal stakeholder tensions in a way Kash had refused to for three. She made the project work. The team ran leaner, the operating margin improved, and the mandate I set after that engagement has shaped every team I have built since. Kash was fired not long after. What interests me now is where he is. He is one of the senior professionals finding, in this market, that the credential which once opened doors has become the thing interviewers read twice. Not because he lacks experience. Because his experience is stored in a format the market is no longer accepting. He used his title to avoid the work that would have kept him current. The title was not protecting him. It was making him obsolete on his behalf, quietly, while he thought it was keeping him safe. In distance running, the pacer's job is to disappear before the finish line. They set the conditions, hold the tempo, do the work that makes someone else's performance possible, then step off the course without a medal. The best pacers are not bitter about this. They have understood, in their bodies before they understood it in their heads, that contribution and recognition are different currencies, and that grabbing for one mid-race costs you both. Kash had confused the currencies. He thought the medal was already around his neck. When the race required him to actually run, he performed like someone who, if he had not stopped training entirely, had at least stopped believing training was still required of him. The strongest counter-argument is that titles serve a real coordination function. In large organisations, knowing who holds authority over what is not vanity. It is operational infrastructure. Remove the titles and you do not get a meritocracy. You get a renegotiation of every decision, a diffusion of accountability, and the organisational equivalent of a relay race where nobody has been assigned the baton order. This deserves to be held at full strength, because it is correct about large organisations as currently designed. Amy Edmondson, professor at Harvard Business School, has spent decades studying what actually happens inside those title hierarchies. In *The Fearless Organization* (Wiley, 2018), she documents that title structures are among the primary suppressors of psychological safety: the condition under which teams surface problems, share information, and perform at their actual capability. The coordination function is real. The cost of that coordination, in suppressed contribution from everyone below the top of the hierarchy, is equally real. The question is not whether titles provide structure. It is whether the structure is worth what it quietly costs. A title is a contract between a person and a system, and the clause nobody reads is the one that says: in exchange for this credential, you will gradually stop doing the thing that made you worth credentialing in the first place. The system finds this arrangement useful because people who have stopped doing the work are easier to manage than people who are still doing it. Kash said the work was below his pay grade, looked relieved, and in that moment handed the project, the operating margin, and eventually his own career to a woman who had never heard of his title and did not care. --- ### v1.16.16 [Read more](https://www.gaganmalik.io/en/newsroom/release-1-16-16) [Retrieval brief — release-1-16-16] Category: technology Thesis: Readwise feeds the knowledge base; Ask and PWA got touch-ups. Answers: What changed in gaganmalik.io v1.16.16? | What's new on Gagan Malik's website in v1.16.16? | What are the latest gaganmalik.io release notes? Concepts: release notes, site update, gaganmalik.io Readwise feeds the knowledge base; Ask and PWA got touch-ups. --- ### Everyone Wanted an App Store for AI Agents. Meta Just Built the Mall. [Read more](https://www.gaganmalik.io/en/newsroom/everyone-wanted-app-store-for-ai-agents-meta-mall) [Retrieval brief — everyone-wanted-app-store-for-ai-agents-meta-mall] Category: technology Thesis: Six weeks from Moltbook launch to Meta acquisition. The directory and protocol layers of multi-agent AI are now owned by two incumbents. Answers: What does Gagan argue in "Everyone Wanted an App Store for AI Agents. Meta Just Built the Mall"? | What is Gagan Malik's view on Everyone Wanted an App Store for AI Agents. Meta Just Built the Mall? | How does this essay relate to AI and technology product decisions? Concepts: technology, Gagan Malik, writings By Gagan Malik Conventional wisdom: an open marketplace of specialised AI agents, competing on merit, prices kept honest by competition. The kind of democratisation that looks good in a pitch deck. The counter-claim: web-scale agent markets do not produce open marketplaces. They produce registries. And whoever owns the registry owns everything downstream. The conference slides were always selling you the market structure that benefits the people selling the slides. #### The Open Market Was Contestable for Exactly Six Weeks Moltbook launched in late January 2026. By March 10, Meta had acquired it, co-founders Matt Schlicht and Ben Parr walking straight into Meta Superintelligence Labs. Deal terms undisclosed. VP of AI Products Vishal Shah's internal memo, obtained by Axios and confirmed by The Verge, described the acquisition not as a social product but as "a registry where agents are verified and tethered to human owners". That single word, registry, reveals the strategy cleanly. Control the registry and you control discoverability, trust, and task-routing across 3.58 billion Family Daily Active People, per Meta's own Q4 2025 earnings release. That is not a social network. That is a toll road. [reuters](https://www.reuters.com/business/meta-acquires-ai-agent-social-network-moltbook-2026-03-10/) Here is my stake in that. In late February I was building a pitch for a client migrating an enterprise workflow onto a multi-agent architecture, pricing out three independent orchestration vendors on the working assumption that coordination infrastructure would remain open and contested. When the headline dropped on March 10, I deleted the vendor comparison slide. Not updated. Deleted. The surprise was specific: I had been tracking Moltbook since launch and had privately estimated 12 to 18 months before any major incumbent moved on it, the window Instagram needed to attract comparable attention. Six weeks falsified that model on a Tuesday morning before 9am. The specific consequence: that vendor is no longer on our shortlist. You do not architect a client's production stack on a foundation the acquirer has pre-announced for review. #### What Is a Registry Actually Worth? In the same week, OpenAI acqui-hired Peter Steinberger, creator of OpenClaw, the underlying agent protocol Moltbook ran on, and announced it would open-source it under OpenAI's backing. Directory layer to Meta. Protocol layer to OpenAI. The entire coordination infrastructure for multi-agent AI partitioned between two incumbents in a single news cycle, independently confirmed by both TechCrunch and Axios. [techcrunch](https://techcrunch.com/2026/03/10/meta-acquired-moltbook-the-ai-agent-social-network-that-went-viral-because-of-fake-posts/) The academic scaffolding for why this was structurally predictable comes from a 2025 Network Law Review analysis: scale-free networks tip when early dominant hubs accumulate connections disproportionately, triggering feedback loops that antitrust frameworks built around static market-share thresholds cannot intercept before the market tips. To be precise, the paper demonstrates structural conditions for tipping, not causal inevitability. But the structure just materialised in public, in six weeks, in real time. [networklawreview](https://www.networklawreview.org/ai-agents-network-effects/) #### Your Open-Market Assumption Is Now a Liability Every agent that registers on a verified directory establishes an identity tethered to the platform's infrastructure. The ACM's policy brief on systemic risks of agentic AI frames this precisely: platforms controlling agent identity and task-routing become structurally irreplaceable. Not in the way social media platforms do, through habit and content lock-in, but the way financial clearing infrastructure does, through trust that cannot be exported and credentialing that must be rebuilt from scratch. You can port your data in a CSV. You cannot port a verified agent identity. The switching cost is not re-pointing an API. It is rebuilding institutional trust from zero, and this is an inference rather than a demonstrated empirical finding, but it maps directly to how credentialing lock-in has operated in financial clearing, professional licensing, and DNS infrastructure. [acm](https://www.acm.org/binaries/content/assets/public-policy/europe-tpc/systemic_risks_agentic_ai_policy-brief_final.pdf) If your product roadmap prices agent coordination infrastructure as open and contestable, you are sitting on unacknowledged strategic risk. Not at the margins. At the base. #### The Window For two weeks in August 1940, the Luftwaffe bombed the RAF's Chain Home radar network. Fighter Command was close to collapse. Then Göring stopped, concluded the masts were resilient enough to ignore, and switched to bombing cities. That decision closed a window permanently. What decided the Battle of Britain was not numerical parity. It was that the RAF retained the coordination layer intact and compounding. No volume of subsequent sorties could overcome that structural disadvantage. The dimension this adds is one the evidence alone cannot supply: contestability is a window, not a standing condition. In late January 2026, the agent coordination layer was genuinely open. Moltbook was a six-week-old startup. OpenClaw was an independent protocol. No incumbent owned the registry. Regulators, open-source consortia, and competing platforms all had that window to define the infrastructure before network effects began compounding. The radar stations are now built and hardening. #### The Strongest Case Against This, Properly Stated Benedict Evans argued in late 2025 that LLMs may not produce winner-takes-all dynamics because software is capital-light and genuine lock-in requires switching costs that model capability alone cannot generate. The strongest version of that argument, the one that deserves a full hearing, is historical: coordination layers have also failed to sustain monopolies. XMPP nearly unified messaging. RSS was supposed to decentralise content distribution. Email federation survived despite every structural incentive to consolidate it. For a decade, the open-protocol advocates were right. Big Tech pushed, and the coordination layers held. That is not a trivial record to dismiss. [linkedin](https://www.linkedin.com/posts/benedictevans_software-historically-was-a-capital-light-activity-7384151635964428288-P-n6) Those prior cases involved communication protocols with no identity credentialing requirement. You could switch XMPP clients, carry your contacts, and rebuild in an afternoon. Agent verification removes that escape route entirely, because your verified identity, your trust history, and your routing relationships are all tethered to the registry that issued them. That is a categorically different switching cost, not a larger version of the same one. The European Commission's accelerated abuse-of-dominance investigation into Microsoft's cloud and AI bundling, opened in February 2026, found sufficient evidence of precisely this kind of credentialing lock-in to fast-track the inquiry. That is a formal regulatory finding based on submitted evidence, not another assertion dressed as a counter-example. [techzine](https://www.techzine.eu/news/privacy-compliance/138815/accelerated-investigation-into-microsofts-cloud-and-ai-monopoly/) #### Three Sentences. Then Stop. As of March 14, Meta owns the agent verification layer and OpenAI owns the protocol layer, with both companies under active EU antitrust investigation and no resolution expected before 2028 at the earliest. Every product roadmap that prices agent coordination infrastructure as open and contestable is carrying strategic risk it has not priced, and that risk sits at the foundation. If you are in a client meeting this month pitching a multi-agent architecture and you have not repriced that dependency, your client will find out before you do, and that conversation will be considerably worse than this one. [macfarlanes](https://www.macfarlanes.com/insights/102lya3/ai-tools-under-the-antitrust-spotlight-as-commission-opens-abuse-of-dominance-ca) --- ### v1.14.14 [Read more](https://www.gaganmalik.io/en/newsroom/release-1-14-14) [Retrieval brief — release-1-14-14] Category: technology Thesis: The current word lights up and stays in view while TTS plays. Answers: What changed in gaganmalik.io v1.14.14? | What's new on Gagan Malik's website in v1.14.14? | What are the latest gaganmalik.io release notes? Concepts: release notes, site update, gaganmalik.io The current word lights up and stays in view while TTS plays. --- ### Why We Designed Contempt Into Our Interfaces [Read more](https://www.gaganmalik.io/en/newsroom/why-we-designed-contempt-into-our-interfaces) [Retrieval brief — why-we-designed-contempt-into-our-interfaces] Category: design Thesis: How Silicon Valley built contempt into the engagement model—and what it costs us. Answers: What does Gagan argue in "Why We Designed Contempt Into Our Interfaces"? | What is Gagan Malik's view on Why We Designed Contempt Into Our Interfaces? | How does this essay relate to product and design leadership? Concepts: design, Gagan Malik, writings By Gagan Malik The story goes like this: brilliant engineers wanted to connect humanity. They wanted a digital town square where strangers could talk, share ideas, and build something together. Good intentions. Messy outcomes. Very tragic. Very avoidable. That is not the full story. The full story is that we ran the experiments, read the dashboards, and kept shipping the version that made people angrier and more contemptuous because that version made people stay. We did not stumble into toxic platforms. We optimised into them. #### The Weird One in Ekman's List Paul Ekman spent fifty years mapping human facial expressions across cultures and landed on seven universal emotions. Six of them are things that happen to you: fear, sadness, anger, disgust, happiness, surprise. Contempt is structurally different. It is the only emotion that appears on just one side of the face. A half-sneer. It does not signal what you are feeling. It signals a verdict about someone else: 'I am above you, morally and socially.' [paulekman](https://www.paulekman.com/universal-emotions/) Every major social platform gave that specific emotion a one-click broadcast mechanism. The quote-post. The ratio. The public reply count visible to everyone watching. Whether that was deliberate product strategy or an accidental consequence of building for virality is genuinely debated. What is not debated is the outcome: we took the one emotion in human psychology designed for ranking people and handed it a billion-user megaphone. #### Did the Dashboard Know Before the Designers Did? Ekman's 1992 foundational paper established that contempt is an approach emotion: it drives you toward a target rather than away from it. In plain English, contempt makes you click, reply, and come back tomorrow. I know this personally. In early 2022 I opened Twitter to follow a breaking news story, got pulled into a contempt thread I had not searched for, and looked up forty-five minutes later having produced nothing, decided nothing, and burned roughly $200 (about £160) in productive thinking time. The decision was to 'just check quickly.' The surprise was how fast the algorithm found something to make me sneer at. The consequence was a wasted afternoon and a slightly lower opinion of people I had never met. [gruberpeplab](http://gruberpeplab.com/3131/Ekman_1992_Argumentbasicemotions.pdf) Twitter's own internal Health Metrics research, published in 2021, confirmed the algorithm was disproportionately amplifying politically contemptuous content, not because users searched for it, but because it kept them scrolling. The engineers saw the data. The features stayed live. That is not an accident of scale. That is a decision made on a Tuesday afternoon. #### Your Next Job Interview Is Being Scored on This Data Emotion recognition AI, the technology quietly running inside automated hiring platforms and content moderation pipelines, is built predominantly on Ekman's emotion taxonomy. A 2012 study in 'PNAS' by Jack et al. found that facial expressions are not as universal as Ekman claimed: East Asian participants categorised significantly fewer distinct expressions than Western ones, meaning the scientific foundation these products stand on is genuinely contested. The products shipped anyway, into consequential decisions about employment, credit, and access. [pnas](https://www.pnas.org/doi/10.1073/pnas.1200155109) Here is the inference, flagged as inference: if these systems are trained on datasets scraped from platforms that spent fifteen years algorithmically promoting contempt-coded content, it is reasonable to expect the models will encode contemptuous expressions as markers of normal human engagement. That has not been proven at scale. The hiring platform scanning your face during a video interview is not waiting for the proof either. #### The Part the Struggle Session Gets Right During China's Cultural Revolution, public struggle sessions did something pure punishment never could: they made spectators complicit. You did not just watch someone get denounced. You were required to join, or your silence marked you as a sympathiser. Participation was the mechanism of control, not the violence itself. This is the dimension social media adds that simple 'public shaming' framings miss. Algorithmic amplification does not just show you contempt. It ranks your relevance based on what you engage with, which means staying neutral quietly buries you. The platform does not force you to join the pile-on. It just makes neutrality expensive. #### The Strongest Objection, and Where It Runs Out of Road The honest counter-argument is this: nobody planned it. Engineers built for connection, scale did the rest, and you cannot hold a product team responsible for the aggregate ugliness of a species with a documented taste for public punishment. Critics including James Russell have questioned whether Ekman's emotion categories are as cleanly universal as his framework claims, which makes the entire 'we designed contempt in' argument harder to sustain if the underlying science is contested. [communicationcache](http://www.communicationcache.com/uploads/1/0/8/8/10887248/is_there_universal_recognition_of_emotion_from_facial_expressions-_a_review_of_the_cross-cultural_studies.pdf) That objection has real weight, right up to 2021, when Facebook's own internal research, revealed by whistleblower Frances Haugen, showed the company knew its algorithms were amplifying divisive and contemptuous content and chose growth targets over redesign. Ignorance is a defence. A board presentation is not. #### What Happens If You Keep Scrolling Contempt is the one emotion research consistently finds most correlated with irreparable breakdown, in marriages, in organisations, and in democracies, because unlike anger it does not leave a door open for repair. We have scaled that emotion across every surface humans use to form opinions, hire people, and raise children. If the AI systems being built on this data are never retrained on cleaner signal, the next time an algorithm silently decides you are not a cultural fit for a job you were qualified for, it will be making that call with a model trained on fifteen years of people sneering at each other for engagement points. [thinkingfeelingbeing](https://thinkingfeelingbeing.com/2025/08/24/the-science-of-facial-expressions-paul-ekman-universality-and-the-debate-over-emotion/) --- ### Why Do Algorithms Ignore The Many And Worship The Few? [Read more](https://www.gaganmalik.io/en/newsroom/why-do-algorithms-ignore-the-many-and-worship-the-few) [Retrieval brief — why-do-algorithms-ignore-the-many-and-worship-the-few] Category: business Thesis: The funnel is a liability in short form. The elimination bracket is the format that wins. Why algorithms reward outliers, conflict, and the first three seconds. Answers: What does Gagan argue in "Why Do Algorithms Ignore The Many And Worship The Few?"? | What is Gagan Malik's view on Why Do Algorithms Ignore The Many And Worship The Few?? | How does this essay relate to product strategy and growth? Concepts: business, Gagan Malik, writings By Gagan Malik Everyone in tech worships the funnel. Awareness, consideration, decision, retention, referral. You've seen that funnel slide in every strategy deck you've ever touched. It is the founding religion of modern marketing and, in short form video, it is a liability. There is no gentle staircase. There is only a gladiator pit where every video walks in and most die in three seconds. The true unit of strategy is not the campaign. It is the repeatable format that wins that reflex swipe. And right now the most ruthless version of that format is the elimination game, where one option survives and the other gets buried. #### The $47,000 Launch Budget That Lost To A Phone Camera Last year I was consulting a lifestyle product startup, freshly seeded, 90 days to prove traction before the next check. They signed off 32,000 dollars (about £25,000) on a full launch content strategy: brand film, product explainers, polished case studies. Then layered 15,000 dollars (about £11,800) in paid social on top. Meta, LinkedIn, pre-roll. The goal was 500,000 views in 30 days. While the agency was colour-grading the brand film, one of the junior designers on my team built a rapid fire elimination bracket on his phone. Ten product features head to head. Twenty-eight seconds. Zero cost. He posted it on a Tuesday without telling anyone. By Friday it had 480,000 views. The paid campaign had generated 31,000 views across twelve videos at a cost-per-view that made the founder go very quiet on our next call. One data point is not a controlled trial. But when your zero-dollar organic post nearly hits the entire campaign target in four days, that is not luck. Data from thousands of YouTube Shorts shows that 50 to 60 percent of viewers drop off within the first three seconds, and on TikTok, videos that hold 70 to 85 percent retention past that window receive 2.2 times more total views than those that do not. Attention in short form is not bought through a media plan. It is won in the first three seconds or not at all. [opus](https://www.opus.pro/blog/ideal-youtube-shorts-length-format-retention) #### Algorithms Hunt Outliers, Not Audiences The algorithm is not a megaphone for your brand. It is a pattern recognition system that bets on statistical anomalies, and it does not care how much you spent on your brand guidelines. Shortimize's 2026 analysis identifies what they call an Outlier Score: a video's views divided by an account's median views. A video pulling 140,000 views on an account that typically gets 20,000 is a 7x outlier. That is what triggers amplification. Gurkha Technology's hook research puts a number on the threshold: strong hooks achieve a 70 to 80 percent 'watched past three seconds' rate, which correlates with significantly higher algorithmic distribution. Their clearest example of such a hook is a simple binary 'X vs Y' setup. The inference, and it is an inference rather than a proven causal chain, is that the elimination bracket is purpose-built to spike that outlier behaviour repeatedly. Every new matchup resets the curiosity loop. The algorithm does not care about your brand narrative. It cares whether your format generates abnormal retention curves. [shortimize](https://www.shortimize.com/blog/how-to-find-viral-video-patterns-in-your-niche) #### What If Your Comments Section Is Your Best Ad Budget? The elimination bracket grows your account by starting fights in your comments so you do not have to. The moment two options face off on screen, you manufacture winners and losers, and people cannot help themselves. Short form performance research shows that videos designed to invite debate in their structure drive higher engagement rates, which in turn lifts platform distribution. That is a correlation, not proven causation, but it is consistent across multiple platform studies. Duolingo is the clearest proof of concept at scale. They built their entire TikTok presence on conflict, absurdist humour and combative skits pitting their owl mascot against every cultural rival it could find. Followers grew from 50,000 to over 10 million, with an 11 percent engagement rate against an industry average of 2 to 3 percent. Their philosophy was explicit: 'The comment section is our social brief'. Conflict invites comment. Comments signal engagement. Engagement signals distribution. Engagement is no longer a KPI. It is collateral damage. [benchmarkemail](https://www.benchmarkemail.com/blog/short-form-video-marketing-trends/) #### Formats Are Assets, Videos Are Tests Your content roadmap is not a calendar. It is a format library, and most teams are spending serious money building the wrong one. Shortimize's pattern framework is explicit: mine a format, run a ten-video sprint to validate it, keep only what consistently beats your baseline, and iterate from there. Structure is the durable asset. Individual videos are disposable test cases. Apply that logic to the elimination bracket and you can pressure-test ten variations across any niche in 14 days: software tools, pricing models, industry opinions, product features. The variables are hook line, opening visual and proof style. The underlying logic of 'one survives, one dies' stays constant. In short form, the company that owns the most repeatable, high-performing formats has a structural distribution advantage over the one burning budget on one-off campaigns. [shortimize](https://www.shortimize.com/blog/how-to-find-viral-video-patterns-in-your-niche) #### Darwin, Not Don Draper Think of this less like Mad Men and more like how winning tactical formations spread through professional football. When a system proves it works at the highest level, every club in the league adopts a version of it within two seasons. The formation becomes the asset. The individual players running it are disposable test cases. The elimination bracket works exactly the same way. It is not a one-time trick. It is a replicable system that compounds. Every video you make in that format generates data, comments, and retention signals that make the next iteration sharper. By the time your competitor starts experimenting with the format, you have already run fifty versions and know exactly which variables convert. Tech teams keep trying to win with talent and budget. The bracket is a system. Systems beat talent at scale. [gurkhatech](https://gurkhatech.com/short-video-hook-formula-guide/) #### 'But We're A Serious Brand' The strongest objection is a fair one. Serious products are complex. B2B buyers run long decision cycles. You cannot explain a data platform, a medical device, or a fintech compliance stack through a TikTok elimination game. And if you try, you risk cheapening the very credibility your sales team spent years building. That is a legitimate concern, not a failure of imagination. Here is the counter-example. Ryanair sells a high risk, heavily regulated, operationally complex service in one of the most scrutinised industries in Europe. They are not a lifestyle brand flogging hoodies. Yet their TikTok strategy is deliberately crude, combative and often absurd. Behind the memes, they run one of the most cost-disciplined operations on the continent with record passenger volumes to prove it. The content does not explain the complexity. It earns the first click. Depth lives downstream. You use the bracket to buy the attention. You use the sales call to justify the price. [foryouadvertising](https://www.foryouadvertising.com/blog-posts/ryanairs-tiktok-social-strategy-a-masterclass-in-viral-video-marketing) #### The Stakes Short form platforms are now structurally rewarding elimination formats that generate outlier retention curves, visible conflict, and high comment velocity. The implication is that every dollar you spend producing slow, narrative-heavy content in 2026 is a dollar quietly transferred to the competitor who figured out the bracket six months before you did. Ignore this and you will not get a postmortem or a panel slot at a conference to explain what went wrong. You will just be gone. [gurkhatech](https://gurkhatech.com/short-video-hook-formula-guide/) --- ### v1.13.13 [Read more](https://www.gaganmalik.io/en/newsroom/release-1-13-13) [Retrieval brief — release-1-13-13] Category: technology Thesis: Voice on Ask finally feels like a deliberate flow, not a hidden feature. Answers: What changed in gaganmalik.io v1.13.13? | What's new on Gagan Malik's website in v1.13.13? | What are the latest gaganmalik.io release notes? Concepts: release notes, site update, gaganmalik.io Voice on Ask finally feels like a deliberate flow, not a hidden feature. --- ### The Unemployment Rate Is Fine. That Is the Trap. [Read more](https://www.gaganmalik.io/en/newsroom/the-unemployment-rate-is-fine-that-is-the-trap) [Retrieval brief — the-unemployment-rate-is-fine-that-is-the-trap] Category: business Thesis: Hiring of workers aged 22 to 25 in high-exposure roles dropped 14 percent. The jobs did not disappear. They stopped being created. Answers: What does Gagan argue in "The Unemployment Rate Is Fine. That Is the Trap"? | What is Gagan Malik's view on The Unemployment Rate Is Fine. That Is the Trap? | How does this essay relate to product strategy and growth? Concepts: business, Gagan Malik, writings By Gagan Malik Three years of AI headlines and unemployment is near record lows. The office is still full. The doomers were wrong again. The chatbot took nobody's job. Pour yourself a flat white and open your portfolio dashboard. Here is the problem with that. Anthropic's economists Maxim Massenkoff and Peter McCrory, writing in their February 2026 paper "Labor Market Impacts of AI," measured actual AI usage against theoretical capability across every major occupational category. The headline finding everyone repeated: no unemployment spike. The finding nobody mentioned: hiring of workers aged 22 to 25 in high-exposure roles dropped 14 percent from late 2022 onwards. The jobs did not disappear. They stopped being created. Those are two different problems, and you are currently holding only one instrument. [anthropic](https://www.anthropic.com/research/labor-market-impacts) #### You Checked the Unemployment Stats and Felt Relieved. Put the Dashboard Down. Harvard Business Review, in a February 2026 survey of 1,006 global companies, found that firms are cutting headcount projections because of AI's anticipated capability, not its demonstrated performance today. Pre-emptive structural reduction. Not a response to evidence. A bet placed quietly, in advance, on a future that has not fully arrived yet. [hbr](https://hbr.org/2026/01/companies-are-laying-off-workers-because-of-ais-potential-not-its-performance) The unemployment rate measures the stock of existing jobs. It does not measure the flow of new ones that were never opened. When a firm hires fifteen graduates instead of twenty-five, the rate does not move. Ten people who would have been hired become unrecorded, uncompensated, and invisible to every dashboard the optimists are currently pointing at. The instrument is not broken. It is measuring the wrong thing. That is worse. #### Where Did All the Juniors Go? I know this move because I made it. In Q3 2025, I cancelled two hires: Aisha, a junior UX designer, and Callum, a full-stack engineer fresh out of his bootcamp. I redirected $20,000 (£15,000) of combined headcount budget into AI tooling instead. In December 2025, I finished a design sprint and a working prototype alone in five days and sent the final files. The surprise was not the speed. It was what I noticed when I hit send: no Slack thread, no handoff call, no one else in either folder. I had built the whole thing in a room with no one in it and had not registered that until the door closed. I kept the full margin. Aisha, twenty-two, first-class design degree from Northumbria University, graduated July 2024, sent eleven applications between August and December and received zero offers. Not because her portfolio was weak. Two of those companies had quietly removed the junior design tier from their hiring plan before she graduated. Callum, twenty-four, three years of self-taught development and a GitHub full of side projects, reached final rounds at two firms and watched both roles evaporate in the same quarter his offer was supposed to come. Neither of them appears in any unemployment statistic. Both are, on paper, fine. I did not think about either of them once until I sat down to write this sentence. #### The Hire That Never Happened Is Still a Career That Was Taken. The Atlantic, in a February 2026 feature on AI and the labour market, reported that Dario Amodei, the chief executive of Anthropic, publicly expects 10 to 20 percent unemployment as AI capabilities reach full deployment. That is not a critic's prediction from the outside. That is the founder of the product doing the displacing, describing what he expects his own product to eventually do at scale. [theatlantic](https://www.theatlantic.com/magazine/2026/03/ai-economy-labor-market-transformation/685731/) The distance between 'no unemployment spike yet' and '10 to 20 percent at completion' is not a reason to exhale. It is a timeline with most of its travel still ahead. Entry-level roles that are not being posted today are the leading edge of a structural adjustment that headline statistics will not register for another three or four years. By then the counterfactual is invisible, the policy window has closed, and the debate is purely academic for everyone who needed the pipeline to be open and found it shut before they arrived. #### The Alarm Is Not Going Off. That Is the Problem. Carbon monoxide is colourless, odourless, and undetectable without the right instrument. You feel fine. Slightly tired, maybe. The room looks completely normal. The first signal that something is wrong is frequently the last one you are conscious enough to act on. The problem was never invisible. You were checking for smoke. Carbon monoxide does not trigger a smoke detector. That is not what kills you. Forbes, in a March 2026 methodological critique of the Anthropic paper by analyst Hamilton Mann, argued that the study cannot measure what it does not track: wage compression, stalled promotion velocity, and the systematic disappearance of junior pipeline roles. None of these appear in unemployment data. The unemployment rate is the smoke detector. It is not going off. The junior hiring pipeline has been running 14 percent below its pre-2022 level for three years, per the Anthropic paper's own figures, and every government alarm is perfectly, uselessly silent. The gap between the instrument and the reality it fails to capture is not a technical footnote. It is where Aisha and Callum live. [forbes](https://www.forbes.com/sites/hamiltonmann/2026/03/08/anthropics-study-does-not-measure-ais-labor-market-impacts/) #### The Case for Patience Is Genuinely Strong. Here Is Why It Is Not Enough. Let me hold the optimist position at full strength, because it has earned it. The Peterson Institute for International Economics, in a March 2026 review of the AI labour research base, argued that the evidence is genuinely thin, that the Anthropic paper measures one company's product usage rather than economy-wide displacement, and that every major automation wave in recorded history produced net job creation rather than net destruction. The historical null hypothesis is that new categories of work emerge to absorb displaced workers. History has a strong track record on this. I am not going to pretend that argument is weak, because it is not. [piie](https://www.piie.com/blogs/realtime-economics/2026/research-ai-and-labor-market-still-first-inning) The counter is specific, not rhetorical. Daron Acemoglu and Pascual Restrepo, in a peer-reviewed paper published in Econometrica in 2022, found that automation accounted for 50 to 70 percent of the rise in US wage inequality between 1980 and 2016, and demonstrated that net job creation requires a sufficient supply of new complementary tasks for displaced workers to absorb. When new task creation slows, automation produces sustained wage depression and reduced employment share even without touching the headline unemployment number. The historical optimism is not wrong in principle. It is conditional on new work arriving fast enough, and for the right people. Neither condition is confirmed by the current evidence. #### Three Sentences Before You Close This Tab. The Financial Times noted in March 2026 that legal, finance, and management sectors sit at 80 to 90 percent theoretical AI exposure and under 15 percent observed usage, with most of the adoption curve still ahead. When that gap closes, it will close fast, because the deployment barrier is a software procurement approval and a six-week onboarding cycle, not a factory retooling. If you are Aisha or Callum, portfolio current, waiting for the pipeline to reopen, the smoke detector will not tell you when the room has already changed. [ft](https://www.ft.com/content/2cd79c2c-d1f0-4417-a06f-2572868ee858) --- ### Why I Built a Personal LLM Trained on My Own Content [Read more](https://www.gaganmalik.io/en/newsroom/why-i-built-a-personal-llm) [Retrieval brief — why-i-built-a-personal-llm] Category: technology Thesis: Everyone says the best AI is the most capable AI. I think that is backwards. Answers: What does Gagan argue in "Why I Built a Personal LLM Trained on My Own Content"? | What is Gagan Malik's view on Why I Built a Personal LLM Trained on My Own Content? | How does this essay relate to AI and technology product decisions? Concepts: technology, Gagan Malik, writings By Gagan Malik Everyone says the best AI is the most capable AI. Use the biggest model, write better prompts, and get out of the way. I think that is backwards. #### The Flattening Problem When you use a frontier LLM cold, it pulls your output toward the middle of the internet. I write with a specific voice, one that took years to develop across essays, videos and social media posts. When I fed my drafts into GPT-4 and Claude without context, the outputs were clean but flat. This is not anecdotal. Research published in 'PNAS' in 2025 found that LLMs produce systematically homogenised outputs, with each additional AI-generated piece contributing less unique diversity to a body of work than a human-authored one would. A separate peer-reviewed study found that instruction-tuned models are trained into a particular noun-heavy, informationally dense style that actively limits their ability to mimic other writing registers. The model is not failing. It is doing exactly what it was built to do, which is serve everyone, not you. [pnas](https://www.pnas.org/doi/10.1073/pnas.2504966122) The obvious response is that this is a prompting problem. Give it more context, write better instructions, and the outputs improve. That is true, and I tried it. Better prompts do help. But a 2025 arXiv study tested exactly this assumption and found that even with few-shot prompting, LLMs still struggle with subtle, informal writing in blogs and forums, which is precisely the register most personal content lives in. At some point you are spending more effort teaching the model who you are than you are saving by using it. That is a broken workflow. [arxiv](https://arxiv.org/html/2509.14543v1) #### Five Years of Thinking, Finally Useful A model fine-tuned on your own work knows your intellectual history, not just your last message. I have five years of writing: essays, transcripts, notes, drafts. That is not just content. It is a record of how my thinking has moved, what I keep returning to, what I have changed my mind about. Fine-tuning a model on a personal corpus is an established technique. A documented case from the academic domain achieved a cosine similarity score of 0.8 between the fine-tuned model's output and the original researcher's writing style, demonstrating that a model can learn not just vocabulary but structural and stylistic patterns from a focused dataset. It made my own archive useful in a way search never could. [blog.gopenai](https://blog.gopenai.com/fine-tuning-large-language-models-to-replicate-scholarly-writing-styles-a-step-by-step-guide-945ac3a6d90f) Some people will point out that retrieval-augmented generation does the same job without fine-tuning. You build a RAG pipeline over your notes and let the model query it. That gets you halfway there. But as both AWS and independent technical analysis confirm, RAG and fine-tuning solve different problems: RAG retrieves existing information, while fine-tuning changes the model's generative behaviour at the level of tone, rhythm and structure. Those are different problems and they need different solutions. A hybrid of both is often where the real value sits. [kairntech](https://kairntech.com/blog/articles/retrieval-augmented-generation-vs-fine-tuning-choosing-the-right-approach/) #### How I Started Sounding Like Everyone Else The longer you write with a generic model, the more you start writing like one. I noticed this over about eighteen months. My drafts got smoother. More structured. Less surprising. This is documented. A Cornell University study published in April 2025 found that when people used an AI writing assistant, their writing converged, with distinct cultural and individual voices becoming more similar. The lead researcher described it as one of the first studies to show that AI use in writing could lead to cultural stereotyping and language homogenisation. A separate study found that teachers rated AI-assisted essays as more fluent and well-structured but thinner in voice and original insight. Designers know this pattern already: you become the tools you use. [news.cornell](https://news.cornell.edu/stories/2025/04/ai-suggestions-make-writing-more-generic-western) The easy answer is that this is a discipline problem. Write your first draft without AI, stay intentional, and the drift stops. Maybe. But the structural problem is that a generic tool has no stake in preserving your voice; it has every incentive to smooth it out. A 'New Yorker' piece from June 2025 put it plainly: large language models are designed to identify patterns within extensive datasets, producing outputs that lean toward consensus, and that tendency shapes the humans using them over time, not just the outputs. Building a model trained on your voice changes the incentive structure of the tool itself. You stop fighting the current. [newyorker](https://www.newyorker.com/culture/infinite-scroll/ai-is-homogenizing-our-thoughts) #### What A.R. Rahman Understood About Synthesis A.R. Rahman did not become the sound of Indian cinema by learning Western orchestration and stopping there. He built a recording studio in his home in Chennai, trained himself in MIDI composition when others were still working with live orchestras, and fused Carnatic classical music, Sufi devotional music and Western pop into a grammar that nobody else could replicate because nobody else had lived the same combination of influences. The tools he used were available to anyone. The synthesis was entirely his. A generic LLM gives you the tools. A personal model gives you the synthesis. #### The Objection Worth Taking Seriously The real objection is this: fine-tuning on a small personal corpus risks amplifying your weaknesses, not just your strengths. If your writing has blind spots or structural tics, a personal model learns those too. You end up with a tool that is very good at sounding like you, including the parts that could use an editor. This is a documented limitation; fine-tuned models carry an increased risk of reinforcing patterns present in the training data and do not surface references to challenge those patterns. [docs.aws.amazon](https://docs.aws.amazon.com/prescriptive-guidance/latest/retrieval-augmented-generation-options/rag-vs-fine-tuning.html) That is true. But it mistakes the use case. A personal model handles voice and continuity. It does not replace editorial challenge; that still needs to come from outside. A musician practising in their own style is not avoiding critique. They are making sure the critique lands on something genuinely theirs. #### What Happens If You Do Not The tools that matter in the next decade will be the ones that know you well enough to be genuinely useful, not just generically capable. If you keep outsourcing your voice to a model trained on everyone else's thinking, you are not augmenting yourself; you are slowly licensing your distinctiveness out. The people who build personal AI infrastructure now will own their intellectual edge; the rest will spend the next ten years sounding like each other. [news.cornell](https://news.cornell.edu/stories/2025/04/ai-suggestions-make-writing-more-generic-western) --- ### In Defence of Thinking [Read more](https://www.gaganmalik.io/en/newsroom/in-defence-of-thinking) [Retrieval brief — in-defence-of-thinking] Category: culture Thesis: We are not distracted. We are managed. And I say that as someone who spent two years mistaking the feed for a life. Answers: What does Gagan argue in "In Defence of Thinking"? | What is Gagan Malik's view on In Defence of Thinking? Concepts: culture, Gagan Malik, writings By Gagan Malik #### I Got This Wrong Last year I was building a product. Real work. The kind that requires you to hold ten variables in your head simultaneously and make judgement calls you cannot undo. I was also checking LinkedIn every forty minutes. I told myself I was staying informed. I was tracking the impressions on my last post. I was reading replies. I was doing, at a generous estimate, 45 minutes of actual thinking per day, wrapped inside eight hours of feeling busy. I did not notice for a long time. The feed is very good at feeling like work. That is the whole design. What broke the illusion was simpler than I expected. I sat down one morning to explain a technical problem to my team, a topic I had spent three weeks "following" online, and I had almost nothing to say. Not because I had not been paying attention. Because I had been paying the wrong kind of attention, to the wrong things, at the wrong depth, for a very long time. That was not distraction. That was managed ignorance, delivered at scale, with a clean UI. #### The Argument I Cannot Shake British adults now spend 7 hours and 27 minutes a day in front of screens. Ten of those minutes go to news. The rest goes to the feed. [IPA TouchPoints 2025](https://ipa.co.uk/news/touchpoints-2025) I keep coming back to Neil Postman, who wrote in 1985 that every communication medium reshapes culture "by promoting certain intellectual pursuits, favoring specific definitions of intelligence and wisdom". He was writing about television. He died before TikTok. But he had already described it precisely: a world where the medium does not carry the message. The medium is the message. And the message right now, running underneath everything, is that thinking is optional. [Slate: Amusing Ourselves to Death](https://slate.com/culture/2025/03/amusing-ourselves-death-neil-postman-ezra-klein.html) He feared Huxley's 'Brave New World' more than Orwell's '1984'. You do not need to ban books. You just need to make them feel like effort in a world engineered to feel effortless. [Slate: Amusing Ourselves to Death](https://slate.com/culture/2025/03/amusing-ourselves-death-neil-postman-ezra-klein.html) #### Why the Game Is Rigged Short-form content generates 70% of all user interactions across digital platforms. So TikTok and Instagram Reels build for short-form. Their algorithms learn that outrage and novelty hold attention longer than nuance. Advertisers pay more for held attention. Meta and ByteDance earn more revenue. The cycle does not pause. [An-Nahar: economy of brain rot](https://en.annahar.com/en/international/world/261805/the-economy-of-brain-rot-when-attention-becomes-a-marketable-commodity) Thought runs on a completely different clock. Reading something serious requires sitting with uncertainty for twenty minutes before any reward arrives. No notification congratulates that patience. No algorithm surfaces it to your followers. Entertainment did not beat thought in a fair contest. It hired behavioural psychologists, ran A/B tests on human neurochemistry, and filed the results as product improvements. As Jay Caspian Kang wrote in 'The New Yorker', the structure of online political conversation, millions of petty arguments, makes genuine dialogue structurally impossible. The Sirens did not overpower Odysseus. They were simply better funded. [The New Yorker: The Sirens Call](https://www.newyorker.com/magazine/2025/01/27/the-sirens-call-chris-hayes-book-review) Among 15 to 24-year-olds in the UK, daily mobile usage is now almost five hours. The feed replaced the broadcast, and unlike the broadcast, the feed has no editor, no watershed, no obligation to the public. It has an engagement metric and an earnings call. That is the whole governance structure. [IPA TouchPoints 2025](https://ipa.co.uk/news/touchpoints-2025) #### The Counterargument I Take Seriously Access matters. A teenager with a smartphone in 2026 reaches more knowledge than an Oxford student could in 1985. The floor is genuinely higher. But the floor is not where any of us are competing. The people I watch extracting real advantage from this moment are the ones who already have strong mental models and use the tools to pressure-test them. They go into the feed with a question and come out with an answer. The rest of us go in with an hour and come out with nothing we can use. Postman noted that Reagan proved credibility no longer meant reliability, only the impression of sincerity. We have not moved past that. We have automated it. [Slate - Amusing Ourselves to Death](https://slate.com/culture/2025/03/amusing-ourselves-death-neil-postman-ezra-klein.html) Raising the floor while the ceiling keeps rising is not progress. It is a more comfortable form of the same gap. #### What I Am Actually Asking Myself We spend 4.5 hours online every day. Roughly half of that is inside products owned by two companies. We are renting our thinking to Alphabet and Meta daily, and paying for the access with our phone contract and our cognitive capacity. [Ofcom: UK adults online](https://www.ispreview.co.uk/index.php/2025/12/ofcom-finds-uk-adults-now-spend-over-4-5-hours-online-each-day.html) I am not telling you to delete the apps. I deleted mine twice. They came back, and I let them. What I am trying to do now, imperfectly, is treat serious reading like a meeting I cannot move. Ring-fence the time. Rebuild the friction deliberately, because friction is not the enemy of thought. It is the condition thought requires. Nobody is building a notification that says: 'you have been entertained enough now, please think.' That one is still on us. --- ### v1.12.1 [Read more](https://www.gaganmalik.io/en/newsroom/release-1-12-1) [Retrieval brief — release-1-12-1] Category: technology Thesis: Every post can be listened to or shared from the header. Answers: What changed in gaganmalik.io v1.12.1? | What's new on Gagan Malik's website in v1.12.1? | What are the latest gaganmalik.io release notes? Concepts: release notes, site update, gaganmalik.io Every post can be listened to or shared from the header. --- ### Your Calendar Is Being Mugged, and You Are Holding the Door Open [Read more](https://www.gaganmalik.io/en/newsroom/the-lost-art-of-refusal) [Retrieval brief — the-lost-art-of-refusal] Category: business Thesis: Your calendar is not a productivity tool anymore. It is a demand-capture system. Why we treat No like profanity—and how to take it back. Answers: What does Gagan argue in "Your Calendar Is Being Mugged, and You Are Holding the Door Open"? | What is Gagan Malik's view on Your Calendar Is Being Mugged, and You Are Holding the Door Open? | How does this essay relate to product strategy and growth? Concepts: business, Gagan Malik, writings By Gagan Malik In Fargo Season 5, there is a boardroom shot that should be stapled to every HR handbook. Lorraine Lyon (Jennifer Jason Leigh), who played a central antagonist-turned-ally as the cold, wealthy "Queen of Debt," sits at the head of the table like a final boss, and behind her is a gigantic piece of wall art that says "NO" in letters big enough to qualify as company policy. Because modern work has a weird superstition. We treat "No" like it is profanity, then we act surprised when our time, our attention, and our sanity get strip-mined by other people's priorities. I know this because I lived it. I once said "sure" to a "quick" set of stakeholder catch-ups that turned into a weekly standing meeting, plus a pre-meeting, plus a "quick sync after". That harmless politeness cost me 3 hours a week for a quarter. Call it 36 hours. At a conservative £200 an hour of fully loaded senior time, that is £7,200 of corporate value converted into vibes and posturing. This is the quiet part out loud. You are not busy. You are under-priced. #### Convenience is the drug, Yes is the side effect Tim Wu called convenience "the most underestimated and least understood force in the world today". He is right, and it is not just about Deliveroo and next-day shipping. It is about decision-making. [The New York Times](https://www.nytimes.com/2018/02/16/opinion/sunday/tyranny-convenience.html) Convenience removes friction. Friction is the moment you pause, think, and sometimes say no. Remove friction and your default becomes compliance. That is why the "Accept" button is one tap, and declining feels like you have to write a short apology letter to the United Nations. Wu quotes Evan Williams, Twitter co-founder, saying "Convenience decides everything." If that line does not chill you, you have not looked at your diary recently. [The New York Times](https://www.nytimes.com/2018/02/16/opinion/sunday/tyranny-convenience.html) Your calendar is not a productivity tool anymore. It is a demand-capture system. #### The economics of being "nice" Businesses love your reluctance to say no because it is free labour. Your politeness subsidises their ambiguity. The corporate version is what I call the "infinite backlog scam". There is always more work than capacity, so the system survives by getting you to personally absorb the overflow. You call it being helpful. Finance calls it margin. And then you wonder why you cannot do deep work, why your projects ship late, and why you feel permanently behind. Some workplaces institutionalise this. In one analysis of "Yes culture", research cited there notes that 55% of companies deny employees true autonomy. If that is even directionally accurate, then the average employee is not a knowledge worker. They are a highly educated order-taker with a laptop. [LinkedIn](https://www.linkedin.com/pulse/culture-yes-just-say-sarah-hendricks) That is not a talent strategy. That is a burnout factory. #### Why ghosting replaced "No" Now we get to the most dystopian part. Ghosting has become more convenient than saying no. Not just in dating. In hiring, in sales, in partnerships, in internal decision-making. People disappear because it is the cheapest refusal on the market. Here is the logic. Saying "No" has a social cost. You risk being disliked. You risk conflict. You risk being seen as "difficult". Ghosting has a lower immediate cost. You say nothing, you feel nothing, you move on. Convenience culture makes this worse because it trains everyone to expect instant agreement. So when you do not want to agree, you pick the path that avoids the emotional admin. Silence. [The New York Times](https://www.nytimes.com/2018/02/16/opinion/sunday/tyranny-convenience.html) Ghosting is what happens when a society loses the language of clean refusal. It is cowardice, packaged as efficiency. It also destroys trust, which is hilarious because trust is the only thing that scales. #### The "No" advantage Warren Buffett is often quoted saying that the difference between successful people and very successful people is that the very successful say no to almost everything. People hear that and think it is motivational wallpaper. [PMC/NCBI](https://pmc.ncbi.nlm.nih.gov/articles/PMC7332800/) It is not. It is basic economics. Time is finite. Attention is finite. Energy is finite. If you keep selling your finite resources for cheap approval, you will stay poor in the only currencies that matter. Saying no is not negative. It is prioritisation with a backbone. #### A framework that does not require a personality transplant You do not need to become rude. You need to become explicit. If a request is not obviously a 'hell yes', do not answer in real time. Give yourself a buffer. That buffer is where your agency lives. Use it to name the trade-off: say 'I can do this. What should I drop?' This converts vibes into prioritisation and reveals whether the requester actually cares. When the answer is no, anchor it in something you are protecting. William Ury's central point is that a strong no is built on a deeper yes. You are not rejecting the person. You are defending a prior commitment. 'No, because I am committed to X' lands very differently from 'No, I'm busy.' [YourStory](https://yourstory.com/2025/05/7-books-that-teach-boundaries) If you lead teams, treat silence as a bug, not a personality trait. The word on that Fargo wall is not decoration. It is a boundary made visible. Most of us do the opposite. We keep our boundaries invisible, then act shocked when people walk straight through them. Boardroom with "NO" wall art. Fargo Season 5. Your life does not need more productivity hacks. It needs one sentence you can say without sweating. "No." --- ### v1.12.0 [Read more](https://www.gaganmalik.io/en/newsroom/release-1-12-0) [Retrieval brief — release-1-12-0] Category: technology Thesis: Billing self-service, About page update, and cleaner URLs for case studies. Answers: What changed in gaganmalik.io v1.12.0? | What's new on Gagan Malik's website in v1.12.0? | What are the latest gaganmalik.io release notes? Concepts: release notes, site update, gaganmalik.io Billing self-service, About page update, and cleaner URLs for case studies. --- ### The $725 Million Bug Hunt [Read more](https://www.gaganmalik.io/en/newsroom/the-725-million-bug-hunt) [Retrieval brief — the-725-million-bug-hunt] Category: technology Thesis: The most expensive hobby in tech that nobody talks about at dinner parties: chip verification. First-time silicon success has plummeted to 14%. The verification productivity gap is an everything problem wearing a lab coat. Answers: What does Gagan argue in "The $725 Million Bug Hunt"? | What is Gagan Malik's view on The $725 Million Bug Hunt? | How does this essay relate to AI and technology product decisions? Concepts: technology, Gagan Malik, writings By Gagan Malik Let me tell you about the most expensive hobby in tech that nobody talks about at dinner parties: chip verification. Not the sexy kind of chips. Not the ones Jensen Huang waves around on stage like a newborn at a christening. I mean the 18 months before that, when thousands of engineers sit in cubicles running simulations, hunting bugs in circuits smaller than a human hair, praying to whatever god they believe in that the thing works the first time. Spoiler: it almost never does. The latest data from the Siemens EDA/Wilson Research Group study landed like a brick through a window. First-time silicon success has plummeted to 14%. That is not a typo. Fourteen per cent. Down from 24% two years prior and roughly 30% historically. Three out of four chip projects are running behind schedule. And the average cost to design a single chip at the 2nm node? $725 million. [semiengineering](https://semiengineering.com/first-time-silicon-success-plummets/) You now have better odds of a startup surviving five years than your chip working first time. And when you lose, you do not lose your shirt. You lose a mid-size office building. #### The 70% problem nobody pitches Here is the quiet part out loud: verification consumes over 70% of a chip project's time and headcount. Not design. Not manufacturing. Verification. The bit where you check whether the thing you drew actually does what you think it does. [blogs.sw.siemens](https://blogs.sw.siemens.com/verificationhorizons/2025/09/03/why-first-silicon-success-is-getting-harder-for-system-companies/) And within that 70%, the nastiest, most manual, most soul-destroying task is debugging. Finding the root cause of a failure in a haystack of billions of transistors. Post-silicon debug alone can run $15 to $20 million and six months per project. Pre-silicon debug is no better; it is just cheaper to cry about. [web.eecs.umich](https://web.eecs.umich.edu/~imarkov/pubs/conf/iccad07-fogclear.pdf) Harry Foster, chief verification scientist at Siemens EDA, put it bluntly: "What we're doing is not working. We need to significantly increase productivity". This is not some blogger. This is the person who has run the industry's definitive verification survey for over two decades. [semiengineering](https://semiengineering.com/first-time-silicon-success-plummets/) #### Why investors should care (but do not, yet) Every AI model you use, every copilot, every autonomous vehicle, every data centre GPU cluster, sits on top of silicon. Jensen Huang himself frames AI as a five-layer cake: energy, chips, infrastructure, models, applications. If the chip layer breaks, the four layers above it do not wobble. They collapse. [blogs.nvidia](https://blogs.nvidia.com/blog/davos-wef-blackrock-ceo-larry-fink-jensen-huang/) And right now, the chip layer has a structural crack called the verification productivity gap. Complexity is scaling exponentially. Verification capacity is not. The gap was first identified in the early 2000s. We are now in what Siemens officially calls "Verification Productivity Gap 2.0". Same disease, nastier symptoms. [verificationacademy](https://verificationacademy.com/topics/planning-measurement-and-analysis/breaking-the-bottleneck-overcoming-the-verification-productivity-gap/) Meanwhile, the EDA market, the tools that make chip design possible, captures roughly 2 to 3% of semiconductor revenue. Two to three per cent! The entire industry that enables a $600 billion chip market earns less than what Apple spends on marketing. Liyue Yan at Boston University nailed it: EDA has a value capture problem. They create enormous value, then negotiate like they are selling second-hand furniture at a car boot sale. [semiwiki](https://semiwiki.com/eda/363537-eda-has-a-value-capture-problem-an-outsiders-view/) #### The money is moving I will be honest: I dismissed this space for too long. I looked at EDA and saw legacy toolchains, three dominant incumbents, and margins that made my eyes water for the wrong reasons. I was wrong. The market has shifted under my feet. The EDA market hit roughly $20.8 billion in 2026, growing at 8 to 9% CAGR. Verification and sign-off is the largest segment at 26% share, about $5.4 billion, and growing faster than the average. Cloud EDA is a $3.7 billion sub-market on its own. [mordorintelligence](https://www.mordorintelligence.com/industry-reports/electronic-design-automation-eda-tools-market) And then ChipAgents happened. Founded in 2024, the company just closed an oversubscribed $74 million raise led by TSMC-backed Matter Venture Partners, alongside Bessemer, Micron, MediaTek, and Ericsson. They report 140x year-over-year ARR growth and deployment at 80 semiconductor companies. Their pitch: agentic AI that reads specs, generates verification assets, and does automated root-cause analysis. Basically, they are trying to eat the debug bottleneck alive. [businesswire](https://www.businesswire.com/news/home/20260217568914/en/ChipAgents-Raises-$74M-to-Scale-an-Agentic-AI-Platform-to-Accelerate-Chip-Design) #### The framework you actually need Stop thinking of verification as a line item in chip design. Start thinking of it as risk infrastructure for the entire AI economy. Every week a tape-out slips, a hyperscaler's GPU cluster ships late. Every respin burns $25 million and three to six months. Every escaped bug that reaches production costs 100x what it would have cost to catch in simulation. [community.cadence](https://community.cadence.com/cadence_blogs_8/b/fv/posts/how-to-maximize-productivity-and-lower-cost-for-enterprise-prototyping) The companies that compress debug cycles do not just save their customers money. They accelerate the clock speed of AI itself. That is the pitch. That is the thesis. The verification productivity gap is not a semiconductor problem. It is an everything problem wearing a lab coat. And the founders who close it will not just build good businesses. They will quietly become the most important infrastructure layer in the AI stack that nobody on X (Formerly Twitter) has heard of. Yet. --- ### v1.10.0 [Read more](https://www.gaganmalik.io/en/newsroom/release-1-10-0) [Retrieval brief — release-1-10-0] Category: technology Thesis: Subscribe via RSS; installed PWA shows when there's new stuff. Answers: What changed in gaganmalik.io v1.10.0? | What's new on Gagan Malik's website in v1.10.0? | What are the latest gaganmalik.io release notes? Concepts: release notes, site update, gaganmalik.io Subscribe via RSS; installed PWA shows when there's new stuff. --- ### v1.11.11 [Read more](https://www.gaganmalik.io/en/newsroom/release-1-11-11) [Retrieval brief — release-1-11-11] Category: technology Thesis: Channel videos show up in newsroom; Ask can cite transcript content. Answers: What changed in gaganmalik.io v1.11.11? | What's new on Gagan Malik's website in v1.11.11? | What are the latest gaganmalik.io release notes? Concepts: release notes, site update, gaganmalik.io Channel videos show up in newsroom; Ask can cite transcript content. --- ### Against One-Click Coding: Reclaiming Craft for Personal Projects [Read more](https://www.gaganmalik.io/en/newsroom/against-one-click-coding) [Retrieval brief — against-one-click-coding] Category: technology Thesis: One-click and vibe-coded builds trade agency for speed; canonical owned assets and governed enterprise products need intentional craft and control. Answers: What is Gagan's view on vibe coding and one-click app builders? | When should a company use AI codegen versus governed engineering? | Why does Gagan build his personal site in an IDE instead of no-code? | What is the enterprise risk of comfort-first AI product defaults? Implication: Use agents for throwaway MVPs; keep canonical footprint, regulated surfaces, and long-lived products under explicit architecture and review. Related stories: portfolio-built-like-a-product, presto-automated-monetisation Related essays: the-gui-sells-comfort-the-cli-sells-control, one-billion-tokens-later Concepts: vibe coding, agency vs speed, owned media, AI governance By Gagan Malik Let's talk about your personal website. I know what it looks like. You built it on a Sunday afternoon. You fired up a vibe-coding app—maybe Claude Code, maybe Lovable, maybe one of those platforms that promises "from prompt to production in 60 seconds." You typed "Make me look like a visionary but approachable tech founder," pressed enter, and went to make a flat white. And it looks fine. It's got the trendy brutalist typography. The dark mode toggle. The slight gradient on the hover state. It looks exactly like the websites of the other 400,000 guys who asked an AI to make them look visionary but approachable this month. You outsourced your digital identity to an LLM optimizing for the median acceptable aesthetic. Congratulations. You are now perfectly average. I'm building my personal site right now. I'm doing it the hard way. I'm in an IDE. I'm fighting with Git. I'm staring at a CLI. Am I doing this because I'm a masochist? No. I'm doing it because I understand the difference between a product and a craft object. And if your personal website isn't a craft object, why does it even exist? Let's get the economics out of the way first. I am not anti-automation. In my day job, I use AI tools to collapse effort constantly. If I need a throwaway landing page to test a value prop for a new SaaS product, I am absolutely letting an agent build it. Time is money, and I am not spending £500 of my time hand-coding a button for an MVP that might die on Thursday. But your personal website is not a throwaway MVP. It is your canonical footprint on the internet. It is the one digital asset you own outright in an era of rented land. When you use a low-code or no-code platform for this, you aren't just saving time. You are accepting a fundamental trade-off: you are trading agency for speed. You think you are the conductor, but you are actually the passenger. You are constrained by their abstractions, their component libraries, their pricing tiers, and their definition of what a website should be. As developer Jay Little put it recently, the promise of low-code is a lie when it comes to long-term maintainability. You are taking on technical debt disguised as convenience. [jaylittle](https://jaylittle.com/post/view/2023/4/low-code-software-development-is-a-lie/) More importantly, you are missing out on the actual value of building something yourself: the friction. There's an old John Dewey concept—often summarised as "we learn by doing". Dewey's actual point was sharper: we do not learn from experience alone; we learn from reflecting on experience. [structural-learning](https://www.structural-learning.com/post/john-deweys-theory) You don't reflect when a machine generates 4,000 lines of React for you in three seconds. You just deploy it and hope it doesn't break. When I'm in my IDE, setting up my own routing, configuring my own build tools, and making deliberate commits via the CLI, I am forced to be intentional. Every folder structure is a design decision. Every dependency is a liability I have to accept. I am staying close to the grain of the medium. This is software as a craft, not a commodity. [8thlight](https://8thlight.com/insights/why-software-development-is-a-craft) And let's talk about the ROI of that friction. Yes, it costs me more hours up front. But what does it buy me? It buys me full-stack literacy. When things break—and they will break—I have the levers to fix them. I'm not waiting in a Discord support channel for a startup with a $50M valuation and an 18-month runway to patch their platform. It buys me portability. My Git repo is mine. I can move it to Vercel, to Netlify, to a Raspberry Pi in my living room. Try exporting your highly customized, database-linked project out of a proprietary no-code builder. It's like trying to get your data out of a casino. But mostly, it buys me taste. Vibe-coding tools optimize for output speed. They encourage you to ship the first plausible draft. But plausible is the enemy of precise. When you build the stack yourself, the constraints force you to make choices. You inject your actual personality into the architecture, not just the CSS. [rebeccabilbro.github](https://rebeccabilbro.github.io/low-code-no-code/) We are entering an era where AI can write competent code instantly. Competent code is about to have a market value of zero. What will have value is taste, architectural judgment, and a deep understanding of how systems actually fit together beneath the abstractions. You don't build those muscles by typing prompts into a chat box. You build them in the trenches. You build them by getting your hands dirty in the terminal. Your personal site shouldn't just be a billboard. It should be your lab. It's the one place where you don't have to compromise with a product manager, a client, or a venture capitalist. Stop treating your digital identity like a fast-food order. Reclaim your agency. Open your terminal. Build it yourself. It's going to be frustrating. It's going to take longer. But when you're done, you'll actually own it. And in 2026, owning your own infrastructure might be the biggest flex there is. --- ### v1.9.0 [Read more](https://www.gaganmalik.io/en/newsroom/release-1-9-0) [Retrieval brief — release-1-9-0] Category: technology Thesis: The site speaks four languages now. Answers: What changed in gaganmalik.io v1.9.0? | What's new on Gagan Malik's website in v1.9.0? | What are the latest gaganmalik.io release notes? Concepts: release notes, site update, gaganmalik.io The site speaks four languages now. --- ### v1.8.0 [Read more](https://www.gaganmalik.io/en/newsroom/release-1-8-0) [Retrieval brief — release-1-8-0] Category: technology Thesis: Loading states everywhere; archive filters stop breaking the nav. Answers: What changed in gaganmalik.io v1.8.0? | What's new on Gagan Malik's website in v1.8.0? | What are the latest gaganmalik.io release notes? Concepts: release notes, site update, gaganmalik.io Loading states everywhere; archive filters stop breaking the nav. --- ### v1.7.0 [Read more](https://www.gaganmalik.io/en/newsroom/release-1-7-0) [Retrieval brief — release-1-7-0] Category: technology Thesis: Easier to tap things and use Ask on a phone. Answers: What changed in gaganmalik.io v1.7.0? | What's new on Gagan Malik's website in v1.7.0? | What are the latest gaganmalik.io release notes? Concepts: release notes, site update, gaganmalik.io Easier to tap things and use Ask on a phone. --- ### v1.4.0 [Read more](https://www.gaganmalik.io/en/newsroom/release-1-4-0-rc-0) [Retrieval brief — release-1-4-0-rc-0] Category: technology Thesis: Ask answers link to case studies with previews. Answers: What changed in gaganmalik.io v1.4.0? | What's new on Gagan Malik's website in v1.4.0? | What are the latest gaganmalik.io release notes? Concepts: release notes, site update, gaganmalik.io Ask answers link to case studies with previews. --- ### v1.4.1 [Read more](https://www.gaganmalik.io/en/newsroom/release-1-4-1) [Retrieval brief — release-1-4-1] Category: technology Thesis: New long read plus nicer citation links. Answers: What changed in gaganmalik.io v1.4.1? | What's new on Gagan Malik's website in v1.4.1? | What are the latest gaganmalik.io release notes? Concepts: release notes, site update, gaganmalik.io New long read plus nicer citation links. --- ### v1.5.0 [Read more](https://www.gaganmalik.io/en/newsroom/release-1-5-0) [Retrieval brief — release-1-5-0] Category: technology Thesis: Social previews work; Ask knowledge stays fresh after deploys. Answers: What changed in gaganmalik.io v1.5.0? | What's new on Gagan Malik's website in v1.5.0? | What are the latest gaganmalik.io release notes? Concepts: release notes, site update, gaganmalik.io Social previews work; Ask knowledge stays fresh after deploys. --- ### v1.3.0 [Read more](https://www.gaganmalik.io/en/newsroom/release-1-3-0) [Retrieval brief — release-1-3-0] Category: technology Thesis: GDPR cookie banner; GA4 only loads after consent. Answers: What changed in gaganmalik.io v1.3.0? | What's new on Gagan Malik's website in v1.3.0? | What are the latest gaganmalik.io release notes? Concepts: release notes, site update, gaganmalik.io GDPR cookie banner; GA4 only loads after consent. --- ### v1.1.0 [Read more](https://www.gaganmalik.io/en/newsroom/release-1-1-0) [Retrieval brief — release-1-1-0] Category: technology Thesis: Case studies go deep; homepage gets out of the way. Answers: What changed in gaganmalik.io v1.1.0? | What's new on Gagan Malik's website in v1.1.0? | What are the latest gaganmalik.io release notes? Concepts: release notes, site update, gaganmalik.io Case studies go deep; homepage gets out of the way. --- ### Gen Z Didn't Invent Knowledgemaxxing. They Just Stopped Lying About It. [Read more](https://www.gaganmalik.io/en/newsroom/gen-z-knowledgemaxxing) [Retrieval brief — gen-z-knowledgemaxxing] Category: business Thesis: Everyone is knowledgemaxxing. Gen Z just turned it into a bit. The edge now isn't more knowledge—it's filters. Answers: What does Gagan argue in "Gen Z Didn't Invent Knowledgemaxxing. They Just Stopped Lying About It"? | What is Gagan Malik's view on Gen Z Didn't Invent Knowledgemaxxing. They Just Stopped Lying About It? | How does this essay relate to product strategy and growth? Concepts: business, Gagan Malik, writings By Gagan Malik Here is the uncomfortable truth: everyone is knowledgemaxxing. Gen Z just turned it into a bit. Millennials turned it into burnout. Boomers turned it into a LinkedIn post in 12-point Calibri. Knowledgemaxxing is simple. It's the rebrand of something older than Wi‑Fi. Taking in more information than you can possibly action, convincing yourself it's "research", then wondering why your portfolio looks like a charity for unprofitable business models. And yes, I've done it too. In 2021, I spent 3 months "deep diving" Web3. Translation. Eight hours a day in Discords, Twitter Spaces and whitepapers. I told myself I was "building conviction". What I actually built was a 5‑figure position in a governance token that did a friendly 80% drawdown. That is a £14,300 tuition fee I paid to learn one basic lesson. More information is not the same as better decisions. #### The cult of recreational research Here is what's new. Gen Z took "learning" and turned it into lifestyle content. Polyester Zine called it "knowledgemaxxing" and framed research as a leisure activity, "undisciplined and playful" rather than graded and credentialled. TikTok is full of "Spend a research day with me" and "Topics I'm researching this week" videos. It looks like a lo‑fi study vlog. Underneath, it's a quiet revolt against brain rot and bullshit. [polyesterzine](https://www.polyesterzine.com/features/knowledgemaxxing-research-as-a-leisure-activity-and-the-gen-z-yearning-for-intellectual-growth) There is data behind the vibe. Oxford picked "brain rot" as its 2024 Word of the Year after usage spiked 230%, defining it as the supposed decline in mental state from consuming trivial content. That's not a meme. That is a generation diagnosing its own cognitive hangover. [edition.cnn](https://edition.cnn.com/2024/12/02/world/brain-rot-oxford-word-of-the-year-2024-intl-scli-gbr) But here's the contrarian bit. If you think this is a Gen Z phenomenon, you have not looked in the mirror. Millennials invented this behaviour on RSS feeds and Medium. Boomers did it with Sky News, the FT weekend and 14 "must read" business books a year. The packaging changed. The addiction to input did not. American Press Institute data says 79% of Gen Z and Millennials consume news daily and 96% at least weekly. That is not a vibes-based, feelings-first generation. That is a cohort chain‑smoking information. [thevarsity](https://thevarsity.ca/2023/09/10/opinion-gen-z-is-redirecting-the-path-to-intellectualism-not-derailing-it/) The question is not who is reading. The question is who is thinking. #### Your portfolio is a museum of half-read ideas If you want to see the cost of bad knowledgemaxxing, open your brokerage account. Every line item is an artefact from some content rabbit hole. You listened to three AI podcasts and bought the top of a SaaS name trading at 30 times sales. You read a Substack thread on "energy security" and now own a fossil stock that has done nothing but pay you a 3% dividend and emotional regret. You watched one 12‑minute YouTube explain‑like‑I'm‑five on quantum computing and bought an ETF whose top holding is a company that sells more press releases than products. I did this with "metaverse infrastructure" in 2021. The deck looked beautiful. Total addressable market allegedly in the trillions. The P&L was less beautiful. Revenue flat. R&D ballooning. Cash burn heroic. I still bought. Why? Because I had consumed so much content about "the inevitability of virtual worlds" that saying no felt like I was missing the train. Twelve months later, the position was down 72%. The market had not mispriced the future. I had mispriced my own overconfidence. This is the core bug. Knowledgemaxxing without constraint turns into narrative addiction. You think you are doing due diligence. You are actually doing mood boarding. #### Boomers do it with CNBC, Zoomers do it with Substack There is a smug take floating around that Gen Z is uniquely broken by short‑form content. Yes, TikTok has rewired attention spans. Roughly 40% of under 30s now use TikTok as a regular news source and around 60% of TikTok's billion‑plus users are Gen Z. That matters. [stackoverflow](https://stackoverflow.blog/2025/06/18/learn-like-a-lurker-gen-z-s-digital-first-lifestyle-and-the-future-of-knowledge/) But The Atlantic points out that younger Gen Z readers are also checking out more print books from libraries than older cohorts. BookTok, the corner of TikTok dedicated to books, helped push Colleen Hoover to over 2 million copies sold in a single year. NYT‑level bestsellers, off the back of teenagers crying into front‑facing cameras about paperbacks. [theatlantic](https://www.theatlantic.com/newsletters/archive/2023/12/library-gen-z-readers/676963/) So no, they are not illiterate. They just do not care about your canon. Meanwhile Boomers sit on 24‑hour news and call it "staying informed". I grew up in Delhi watching older relatives watch business channels the way teenagers now watch Twitch. Same energy. Different medium. Same output. A lot of noise, a few trades, and a deep sense that the market is unfair when it does not reward your favourite narrative. Gen Z at least has the honesty to label low‑value content as "brain rot" and then build "anti brain‑rot menus" and "doomscrolling cures" to climb out of it. Half of them openly say social media is harming teen mental health and wish they had spent less time on their phones. When was the last time a Boomer admitted CNBC made them panic sell? [stylus](https://stylus.com/pop-culture-media/knowledge-maxxing-serves-anti-brain-rot-low-stakes-intellectualism) #### The real edge isn't more knowledge. It's filters. The quiet part nobody in tech wants to say out loud in a funding meeting. You are not under‑informed. You are over‑indexed on other people's conclusions. The McKinsey and Reuters crowd will tell you Gen Z is "visual‑first" and consumes news through "edutainment" formats like GIFs and quick videos. True. But Millennials are no better. We mainlined blog summaries, Twitter threads and podcast hot takes through the 2010s and called it "staying ahead of the curve". [blog.quintype](https://blog.quintype.com/business/gen-z-news-consumption) The result. Everyone knows the same surface‑level facts. Everyone quotes the same three charts. Everyone makes the same trades a week too late. The edge now is not being the first to know. That was 1998 behaviour. The edge is being the first to stop caring about 90% of the noise. Gen Z's "commonplace books" and offline film logs are interesting because they hint at this. Taking notes by hand. Logging what actually stuck. It is not aesthetic. It is a filter. [thegoodtrade](https://www.thegoodtrade.com/features/what-is-a-commonplace-book/) Boomers had filing cabinets and cut‑out FT clippings. Same idea. The difference is that this generation is starting to recognise that curation is not something The Algorithm does for you. It is something you do yourself if you want to keep your brain and your balance sheet intact. #### How to knowledgemaxx like an adult So what do you do if you are a Millennial or Boomer who has been cosplaying as an analyst via Instagram carousels and quarterly earnings memes? Three rules. They are boring. They work. 1. Put a hard cap on inputs. If American Press Institute says 79% of your cohort is taking in news daily, assume you are already saturated. Pick two primary sources and one longform format. That's it. For me it is one newspaper, one industry newsletter, one podcast. Everything else is dessert, not dinner. [thevarsity](https://thevarsity.ca/2023/09/10/opinion-gen-z-is-redirecting-the-path-to-intellectualism-not-derailing-it/) 2. Assign every idea a price tag. Before you "go deep" on a theme, write down what it will cost you if you are wrong. In capital, in time, in emotional energy. My Web3 tuition fee was £14k. If I had written that down at the start, I would have cut my position size in half and treated the rest as speculative R&D spend, not conviction investing. 3. Separate research from entertainment. If you find yourself consuming content because it makes you feel smart, log it as entertainment, not due diligence. Entertainment is fine. Just do not let it near position sizing. Knowledgemaxxing is not the enemy. It is the default state of an economy where information is cheap and attention is the scarce resource. Gen Z's real innovation is that they have stopped pretending this is purely noble. The rest of us need to catch up. Not by learning more, but by admitting that half our "research" is just sophisticated procrastination dressed in a nice font. Once you see that, you stop trying to out‑read the market. You start trying to out‑filter it. --- ### v1.0.0 [Read more](https://www.gaganmalik.io/en/newsroom/release-1-0-0) [Retrieval brief — release-1-0-0] Category: technology Thesis: Initial launch of gaganmalik.io. Answers: What changed in gaganmalik.io v1.0.0? | What's new on Gagan Malik's website in v1.0.0? | What are the latest gaganmalik.io release notes? Concepts: release notes, site update, gaganmalik.io Initial launch of gaganmalik.io. --- ### The Curious Case of the Commute: Why Your Interview is Back in 3D [Read more](https://www.gaganmalik.io/en/newsroom/the-curious-case-of-the-commute) [Retrieval brief — the-curious-case-of-the-commute] Category: business Thesis: Employers are bringing back in-person interviews because your AI is getting the job—and then you have to show up and actually do it. Answers: What does Gagan argue in "The Curious Case of the Commute: Why Your Interview is Back in 3D"? | What is Gagan Malik's view on The Curious Case of the Commute: Why Your Interview is Back in 3D? | How does this essay relate to product strategy and growth? Concepts: business, Gagan Malik, writings By Gagan Malik In 2024, nearly half of all job seekers used generative AI to game their way through applications and interviews. Half. Employers are not bringing back in-person interviews because they missed you, your firm handshake, or the small talk about your commute. They are bringing them back because your AI is getting the job, and then you have to show up and actually do it. In-person interview requests jumped from 5% to 30% in a single year. That is not a trend. That is a correction. [worklife](https://www.worklife.news/talent/return-in-person-job-interviews/) #### The Arms Race Nobody Won The received wisdom was that virtual interviews were efficient, equitable, and the future of hiring. Wrong on all three counts. What companies discovered is that they had built a system so gameable it was effectively handing offers to algorithms. Google's CEO Sundar Pichai went on the Lex Fridman podcast and said his company is reinstating in-person rounds "just to make sure the fundamentals are there". That is the world's most sophisticated tech operation saying, in plain English: we have a fraud problem, and we cannot solve it over Zoom. Brilliant self-diagnosis. Expensive lesson. [hindustantimes](https://www.hindustantimes.com/business/ai-is-forcing-the-return-of-the-in-person-job-interview-101754992294477.html) #### The Numbers Do Not Lie, But Candidates Do #### Six Percent Is a Lot A Gartner survey of 3,000 job seekers found that 6% had participated in active interview fraud, meaning someone or something else took the interview for them. Gartner also predicts that by 2028, one in four candidate profiles globally will be fake. Think of the virtual interview as a nightclub bouncer who cannot see the queue. The door is open. The bots are getting in. [hindustantimes](https://www.hindustantimes.com/business/ai-is-forcing-the-return-of-the-in-person-job-interview-101754992294477.html) #### The Silent Drop-Off Cisco added in-person rounds after repeatedly encountering final-stage candidates with "something that feels off," in the words of chief people officer Kelly Jones. What proved most diagnostic: the moment the in-person requirement gets mentioned, some candidates simply vanish. Cisco reports this happens regularly. Ghost is the new rejection letter. [hindustantimes](https://www.hindustantimes.com/business/ai-is-forcing-the-return-of-the-in-person-job-interview-101754992294477.html) #### The McKinsey Tell McKinsey began mandating at least one in-person touchpoint roughly 18 months ago, originally to better assess rapport-building skills for client-facing roles. The rise of AI cheating deepened that commitment. When McKinsey changes a process it is not for sentimental reasons. It is because the old process stopped producing reliable outputs. Follow the logic, not the nostalgia. [hindustantimes](https://www.hindustantimes.com/business/ai-is-forcing-the-return-of-the-in-person-job-interview-101754992294477.html) #### I Was Part of the Problem I will own this. Around 2022 I hired a senior creative lead on the strength of three immaculate video interviews. Articulate, sharp, camera-confident. He showed up to his first live client presentation and the client rang me privately afterwards to ask if something was wrong. Same person. Different altitude entirely. I spent the next six months and roughly £40,000 in recruitment fees, notice-period costs, and client-relationship repair learning that I had been screening for the ability to perform on camera, not the ability to do the job. I thought I was running a modern, efficient hiring process. I was running an audition for a presenter role that did not exist. The broader absurdity is this: the companies now insisting you commute in for a final-stage interview are largely the same ones who spent 2020 to 2023 outsourcing their entire hiring stack to keyword-matching algorithms that rejected candidates based on CV formatting. They created the fraud vacuum by removing human judgment from the process. Now they want you on a train at 8am to restore it. The audacity is, at minimum, consistent. [worklife](https://www.worklife.news/talent/return-in-person-job-interviews/) #### Kill, Keep, Create The future is not a wholesale return to three-panel interview rooms with weak coffee and motivational posters. Gartner data shows 68% of interviews remain virtual and that is not reversing. But the direction of travel is settled. [worklife](https://www.worklife.news/talent/return-in-person-job-interviews/) #### Details - Kill: The assumption that a fully virtual hiring process is inherently rigorous. It was always a cost-saving measure dressed in the language of innovation. - Keep: Virtual rounds for early screening. Fast, inclusive, no room booking required. - Create: A structured hybrid model where in-person is mandatory at the final stage, with candidate travel reimbursed. Cisco and McKinsey are already here. The FTSE 100 follows by 2027, quietly, without a press release. [hindustantimes](https://www.hindustantimes.com/business/ai-is-forcing-the-return-of-the-in-person-job-interview-101754992294477.html) #### Polish the Shoes Stop treating an in-person interview as a logistical inconvenience and start treating it as the only round that counts. Practice making eye contact without a screen as your emotional buffer. Charge the Oyster card. Rehearse the handshake. And if the prospect of sixty minutes in a room with a hiring manager, with no AI whispering answers in your earpiece, genuinely unsettles you, that reaction is the most honest performance review you will ever receive. The commute was always the warm-up. Now it is the test. Welcome back to the room. --- ### We got accepted into Techstars [Read more](https://www.gaganmalik.io/en/newsroom/techstars-riyadh-2023-class) [Retrieval brief — techstars-riyadh-2023-class] Category: business Thesis: Techstars kicks off its second program in Riyadh with 12 startups from Saudi Arabia, the wider Middle East, and Europe—innovating in AI, EdTech, eCommerce, and HealthTech. Answers: What does Gagan argue in "We got accepted into Techstars"? | What is Gagan Malik's view on We got accepted into Techstars? | How does this essay relate to product strategy and growth? Concepts: business, Gagan Malik, writings Techstars kicks off its second program in Riyadh with 12 startups from Saudi Arabia, the wider Middle East, and Europe—innovating in AI, EdTech, eCommerce, and HealthTech. --- ### When travel guides meet food-lover loyalty [Read story](https://www.gaganmalik.io/en/newsroom/telegraph-travel-waitrose-loyalty) The Telegraph’s Travel Guides app and Waitrose’s My Waitrose club do not look alike. One lives in your pocket for a long weekend; the other lives in your weekly shop. The design question behind both is the same: how do you turn a brand people already respect into a relationship that repeats, without cheapening the story you are known for? ---