https://manasbihani.netlify.app/Manas Bihani takes apart questions about AI, technology, money and products: essays, benchmarks, models, datasets and teardowns.https://manasbihani.netlify.app/work/periodic-table/https://manasbihani.netlify.app/work/periodic-table/Tue, 06 Oct 2026 00:00:00 GMTVisualizationEvery data centre is a pile of specific atoms. Fifty-six of the 118 elements have a job in it, fourteen of those are short, concentrated or behind an export licence right now, and five are newer materials trying to replace an incumbent. A map of which is which, and what each one suggests about the next bet.https://manasbihani.netlify.app/work/moat-when-the-model-isnt-yours/https://manasbihani.netlify.app/work/moat-when-the-model-isnt-yours/Sun, 27 Sep 2026 00:00:00 GMTNoteThe labs keep absorbing what used to be startup features. So what is left that a startup can actually defend?https://manasbihani.netlify.app/work/problem-selection-premium/https://manasbihani.netlify.app/work/problem-selection-premium/Sun, 27 Sep 2026 00:00:00 GMTNoteWhen building gets cheap, the scarce skill isn't building. It's deciding what deserves to be built.https://manasbihani.netlify.app/work/same-model-different-wiring/https://manasbihani.netlify.app/work/same-model-different-wiring/Sun, 27 Sep 2026 00:00:00 GMTNoteIf everyone gets roughly the same intelligence, people and companies stop differing by what they can think, and start differing by what they plug it into.https://manasbihani.netlify.app/work/selection-is-the-bottleneck/https://manasbihani.netlify.app/work/selection-is-the-bottleneck/Sun, 27 Sep 2026 00:00:00 GMTNoteA model will give you a hundred options in a minute. Knowing which one is good was always the job; now it is the whole job.https://manasbihani.netlify.app/work/the-ai-race/https://manasbihani.netlify.app/work/the-ai-race/Sun, 27 Sep 2026 00:00:00 GMTEssayIt looks like one race to build the smartest model. It is five: capability, compute, deployment, distribution and economics, with nations and safety wrapped around all of them. A plain guide to the whole machine: who is racing, what the words mean, who pays whom, and what comes after.https://manasbihani.netlify.app/work/the-convergence-tax/https://manasbihani.netlify.app/work/the-convergence-tax/Sun, 27 Sep 2026 00:00:00 GMTNoteCompanies using the same models to ask the same questions will get the same answers. Sameness is a cost, and nobody books it.https://manasbihani.netlify.app/work/the-luxury-of-reality/https://manasbihani.netlify.app/work/the-luxury-of-reality/Sun, 27 Sep 2026 00:00:00 GMTNoteAs experiences become synthetic and instant, the things that can't be made at zero marginal cost start to look like luxuries.https://manasbihani.netlify.app/work/the-translators/https://manasbihani.netlify.app/work/the-translators/Sun, 27 Sep 2026 00:00:00 GMTNoteA growing class of people aren't the best engineers, but understand the technology well enough to follow it all the way to a market. That bridge may be worth more than either end.https://manasbihani.netlify.app/work/verification-economy/https://manasbihani.netlify.app/work/verification-economy/Sun, 27 Sep 2026 00:00:00 GMTNoteWhen producing an answer stops being the hard part, checking one becomes the business.https://manasbihani.netlify.app/work/bets-against-the-wall/https://manasbihani.netlify.app/work/bets-against-the-wall/Sat, 26 Sep 2026 00:00:00 GMTVisualizationEvery startup attacking AI's bottlenecks is betting that a wall will still be standing when its product ships. Whether it's right about the wall decides what it builds. Who has to buy it tends to decide how that gets sold, and how the story goes.https://manasbihani.netlify.app/work/cant-buy-your-way-out/https://manasbihani.netlify.app/work/cant-buy-your-way-out/Sat, 26 Sep 2026 00:00:00 GMTVisualizationThe richest companies in the world can't buy their way to more AI any faster. Money buys a place in line. Only building makes the line shorter, and the lines are now measured in years while AI models arrive every few months.https://manasbihani.netlify.app/work/one-token/https://manasbihani.netlify.app/work/one-token/Sat, 26 Sep 2026 00:00:00 GMTVisualizationFollow one prompt through an open frontier model, drawn in 3D, from the electricity that moves it to the word on your screen. Why the reply comes one word at a time, why memory, not maths, is the expensive part, and what each word costs in electricity and water.https://manasbihani.netlify.app/work/the-grid-is-the-last-wall/https://manasbihani.netlify.app/work/the-grid-is-the-last-wall/Sat, 26 Sep 2026 00:00:00 GMTVisualizationInside the fence, everything a data centre needs can be bought, eventually. Outside it, the power has to come from somewhere, and every way of getting more is measured in years. An engraved walk from the data hall out along the line to the gas turbines, the reopened reactor and the geothermal wells that buyers are now paying for, then country by country: where the wall is highest.https://manasbihani.netlify.app/work/who-got-paid/https://manasbihani.netlify.app/work/who-got-paid/Sat, 26 Sep 2026 00:00:00 GMTVisualizationWhen AI hardware ran short, some companies got to charge far more. Others sold more but earned less on each sale, and a few ended up paying more themselves. Same shortage, different endings. The difference comes down to one question: could the buyer go somewhere else?https://manasbihani.netlify.app/work/why-this-paper-matters/https://manasbihani.netlify.app/work/why-this-paper-matters/Sat, 26 Sep 2026 00:00:00 GMTExplainerTwelve papers that moved where the constraint in AI sits, from the Transformer in 2017 to the models of 2025. Each one opens with the question it answered.https://manasbihani.netlify.app/work/constraint/https://manasbihani.netlify.app/work/constraint/Fri, 25 Sep 2026 00:00:00 GMTVisualizationAn AI data centre is a mill that turns electricity into words. Follow the power from the grid to the chip and back out as heat, and see why each part of the mill, in turn, became the thing everyone was short of, until the shortage reached the power grid itself.https://manasbihani.netlify.app/work/desk-dataset/https://manasbihani.netlify.app/work/desk-dataset/Thu, 24 Sep 2026 00:00:00 GMTDatasetEvery object on this site, one row each, rebuilt every time the site is. Download it and do something strange with it.https://manasbihani.netlify.app/work/renaissance-edition/https://manasbihani.netlify.app/work/renaissance-edition/Wed, 23 Sep 2026 00:00:00 GMTNoteBefore this desk there was “Keeping Up With AI”, a Renaissance-themed news edition. Kept here, as it was.https://manasbihani.netlify.app/work/aura/https://manasbihani.netlify.app/work/aura/Tue, 22 Sep 2026 00:00:00 GMTNoteWalter Benjamin thought reproduction would dissolve the aura of the original. It made originals more expensive instead.https://manasbihani.netlify.app/work/rented-attention/https://manasbihani.netlify.app/work/rented-attention/Sun, 20 Sep 2026 00:00:00 GMTNoteMost “distribution advantages” are leases on someone else’s audience, and leases get repriced.https://manasbihani.netlify.app/work/convergence-test/https://manasbihani.netlify.app/work/convergence-test/Thu, 17 Sep 2026 00:00:00 GMTNoteA repeatable measurement of how alike AI products are becoming. Pilot run of eight products; results so far are specimens.https://manasbihani.netlify.app/work/shelf/https://manasbihani.netlify.app/work/shelf/Sat, 12 Sep 2026 00:00:00 GMTCollectionBooks, essays and papers, most of them written long before modern AI, that turn out to be about it.https://manasbihani.netlify.app/work/anatomy/https://manasbihani.netlify.app/work/anatomy/Tue, 08 Sep 2026 00:00:00 GMTNoteA cross-section of a typical AI company, from interface to bedrock, ranked by how hard each layer would be to copy.https://manasbihani.netlify.app/work/expertise-shocks/https://manasbihani.netlify.app/work/expertise-shocks/Wed, 02 Sep 2026 00:00:00 GMTNoteEach time a layer of expertise got cheap, value moved up a level. The gaps between shocks are getting shorter.https://manasbihani.netlify.app/work/nineteen-public-keys/https://manasbihani.netlify.app/work/nineteen-public-keys/Tue, 01 Sep 2026 00:00:00 GMTEssayA swarm of AI agents rebuilt the internet’s communication stack in four days. They got to the last layer and hit the same wall we did.https://manasbihani.netlify.app/work/value-machine/https://manasbihani.netlify.app/work/value-machine/Fri, 28 Aug 2026 00:00:00 GMTModelDrag the price of intelligence down three orders of magnitude and watch where the value goes. A toy model with its assumptions showing.https://manasbihani.netlify.app/work/aaa-rated-gpus/https://manasbihani.netlify.app/work/aaa-rated-gpus/Wed, 19 Aug 2026 00:00:00 GMTEssayWall Street just agreed to insure Nvidia’s downside. Once you see the shape of the trick, you start seeing it everywhere else too.https://manasbihani.netlify.app/work/moats/https://manasbihani.netlify.app/work/moats/Wed, 19 Aug 2026 00:00:00 GMTVisualizationTen sources of competitive advantage, scored before and after intelligence got cheap. The ones in blue are draining.https://manasbihani.netlify.app/work/rhinoceros/https://manasbihani.netlify.app/work/rhinoceros/Tue, 11 Aug 2026 00:00:00 GMTNoteDürer drew the most famous rhinoceros in Europe without ever seeing one.https://manasbihani.netlify.app/work/scarcity/https://manasbihani.netlify.app/work/scarcity/Tue, 04 Aug 2026 00:00:00 GMTEssayEvery time something becomes abundant, the value moves next door. A search for the next door.https://manasbihani.netlify.app/work/ai-has-passed-every-exam-it-has-never/https://manasbihani.netlify.app/work/ai-has-passed-every-exam-it-has-never/Sat, 01 Aug 2026 00:00:00 GMTEssayThe machine has passed every exam we can build and discovered nothing.https://manasbihani.netlify.app/work/every-ai-bet-is-the-same-bet/https://manasbihani.netlify.app/work/every-ai-bet-is-the-same-bet/Tue, 21 Jul 2026 00:00:00 GMTEssayNuclear reactors and downloadable models look like opposite strategies. They’re two sides of one wager on a question nobody will say out loud: does intelligence have a ceiling?https://manasbihani.netlify.app/work/indias-carbon-markets-a-new-test/https://manasbihani.netlify.app/work/indias-carbon-markets-a-new-test/Wed, 08 Jul 2026 00:00:00 GMTEssayPart one of three on carbon credits, India’s compliance market, and where the money will actually go.https://manasbihani.netlify.app/work/the-electricity/https://manasbihani.netlify.app/work/the-electricity/Sat, 04 Jul 2026 00:00:00 GMTEssayWhy Google may be trying to make intelligence disappear.https://manasbihani.netlify.app/work/the-wall-that-wasnt-yours/https://manasbihani.netlify.app/work/the-wall-that-wasnt-yours/Sat, 13 Jun 2026 00:00:00 GMTEssayFable 5 is Anthropic’s best model, but Friday’s letter showed why 'best' doesn’t mean valuablehttps://manasbihani.netlify.app/work/it-was-never-the-gpu/https://manasbihani.netlify.app/work/it-was-never-the-gpu/Tue, 09 Jun 2026 00:00:00 GMTEssayHow AI keeps solving the wrong bottleneckhttps://manasbihani.netlify.app/work/the-speed-of-being-wrong/https://manasbihani.netlify.app/work/the-speed-of-being-wrong/Fri, 05 Jun 2026 00:00:00 GMTEssayWhether AI levels you up or quietly hollows you out comes down to a single variable almost no one is naming. It is not your skill.https://manasbihani.netlify.app/work/uber-burned-a-year-of-ai-budget-in/https://manasbihani.netlify.app/work/uber-burned-a-year-of-ai-budget-in/Sun, 31 May 2026 00:00:00 GMTEssayBig Tech turned AI usage into a metric and a 124-year-old bounty scheme in colonial Hanoi explains exactly what happened next.https://manasbihani.netlify.app/work/finding-a-flat-in-india-is-broken/https://manasbihani.netlify.app/work/finding-a-flat-in-india-is-broken/Sun, 10 May 2026 00:00:00 GMTEssayOn fake listings, misaligned incentives, and why the right home might already be three streets away from you.https://manasbihani.netlify.app/work/the-internet-feels-worse-heres-why/https://manasbihani.netlify.app/work/the-internet-feels-worse-heres-why/Sat, 21 Mar 2026 00:00:00 GMTEssayHow AI advertising is driving culture underground and why the places that feel authentic today are only safe until someone notices them.https://manasbihani.netlify.app/work/gen-z-is-going-offline/https://manasbihani.netlify.app/work/gen-z-is-going-offline/Sun, 15 Mar 2026 00:00:00 GMTEssayWhat looks like a generational preference for offline socializing is a market signal about the failure of retention-based business models. Gen Z isn’t rejecting digital tools. They’re leaving platform