What each page answers

The migrating bottleneck

Which input is scarce now, and where does the shortage move next?

• The scarcest part gets paidWhy does the scarcest input collect the profit?
• It has already movedWhere has the shortage been so far, and how did it move?
• Watch what the labs buyWhat do the labs' purchases reveal about what they find scarce?
• The scorecardHow tight is each input tonight, and which cannot be scored?
• What nobody publishesWhich possible shortages have no public data?
• What binds now and nextWhat binds now, what binds next, and what would prove each wrong?
• The predictions, re-tested every nightAre those predictions holding?

The value chain

Who are the companies at each step, and which hold lasting power?

• Compute & physicalWho makes the chips, clouds and power, and who holds power there?
• ModelWhich labs build the models, and can they keep their margins?
• Training inputWho supplies the data and expert work models learn from?
• Serving & orchestrationWho runs and routes models for others?
• Deployment & applicationWho builds the products and puts AI to work in firms?
• AdoptersWhat do the firms using AI gain?
• Labour & consumersWhat reaches workers and consumers?
• The market mapWhich companies sit in each finer category, and where does coverage stop?

Businesses that could be built

Where is the chain short of something a new business could supply?

• The rent rule, drawnWho would keep each business's profit, and how large would it be?
• The businesses at a glanceWhere do they sit on the chain, how crowded is it there, and which kinds of firm would need them?
• The businessesWhat is each business, what profit would it keep, and what would prove it wrong?

What to build, future by future

Which businesses are worth building, and which hold whatever happens?

• Test a business of your ownWhat would the site's records say about a business that is not on the list?
• The chain, and what is scarce on itHow short is each part of the chain now, in the transition and at the mature state, and what does each future do to it?
• A call on each businessWhich should be built, which only on a condition, and what does each future do to them?
• The ground under today's companiesWhich parts of the chain get scarcer, which hold and which ease, and which companies sit on each?
• Where the shortage goes nextWhich shortages does the site predict, how is each prediction doing tonight, and which businesses ride it?
• What a durable business buildsWhich lasting things does each business build up?
• The planning assumptionsWhat do the calls on this page take for granted, and what would weaken each assumption?
• Every business in every futureWhich businesses does each future strengthen, weaken or break?

The data market

What are the people who train AI buying?

• How this is readWhere does a record of what model builders buy come from, when they publish none?
• Who pays whomWhich way does the money run, and which prices are published?
• The signalsWhat do the vendors' own pages show their buyers asking for, and how many vendors show each?
• What the pages do not showWhat can this reading not tell you?

The data market: vendors

Who sells training data, and what do each vendor's own pages show?

• Sellers with evidence of fast growthWhich vendors are growing fast, and which signals do their pages show?
• Sellers with weak evidence of growthWhich vendors are listed on thinner evidence?
• Model buildersWhat do model builders say they train on, collect or buy?

The data market: prices and pay

What do vendors pay for data and the people who make it, in their own published figures?

The data market: every kind of record

Which kinds of record could train AI, and is anyone already selling or using each?

• Every kind of record as one markHow much of the field is already taken?
• The records, family by familyWho holds each kind of record, and what was found?

The data market: dossiers

Which kinds of record have been examined closely, and what did each turn up?

One page per kind of record: what it is, who holds it, who is already there, what stands in the way, and a labelled judgement.

The data market: how it was made

How was this section made, what did each check change, and what is wrong with it?

• From candidates to dossiersHow many kinds of record were looked at, and how many closely?
• What each check changedHow much did a second reading change?
• What is left outWhat is not on these pages, and why?
• What is wrong with itWhere is this work weak?

The stack

What are the layers and sub-layers, and which companies does the tracker follow in each?

One page per sub-layer: its companies, its readings and the venture money going in.

Circular financing ledger

How much of AI's demand is paid for by its own suppliers?

What happens to each kind of firm

What does AI do to an accounting practice, a hospital or a freight operator, and what will such firms need to buy?

• TodayHow much of each kind of firm's payroll passes the census screen, and what holds the rest?
• The stagesWhat is known now, what is judged next, and what is extrapolated for a world with robots?
• Kind by kindWhat does each kind of firm look like at each stage, and what would prove that wrong?
• What an AI-run firm looks likeWhat shape does this site judge these kinds of firm are heading towards?
• Where the opportunity isWhat will AI-run firms need to buy, and which business would supply it?

Compare

Does the same evidence favour the normal-technology view or the fast one?

Atlas

What do named writers expect to change in each part of life?

• The mapWhere do the sourced works expect change, and by when?
• Eight parts of lifeWhat is expected in money, work, meaning, power, war, science, health and an ordinary day?

One page per part of life: each writer's expectation, in this site's words, with its source.

Futures

What did fiction imagine, has it been built, and who would profit?

• When each idea was imaginedIn which decade was each idea first imagined?
• When they may arriveWhen might the unbuilt ones arrive?
• Kinds of technologyWhat kinds of technology are there, and who would keep the profit from each?
• A galleryWhat might these inventions look like?

One page per decade: the ideas imagined or expected then.

One page per kind of technology: its ideas and the dated forecasts about it.