If everyone has an analystInformation advantage used up again, and new asymmetries
In the past, only a few households could afford the time to gather news and analyse fleets, and those who were informed could earn a little at the expense of those who were not. Now every household has a scribe that can read every report in a few minutes. With more people gathering information, prices reflect it faster and more fully, and the benefit of reading reports thins out until it is almost zero. But the race does not end: whoever learns something a moment earlier can still make money from others, so everyone competes to be first, spending resources that create no value at all. With more information than anyone can read, what becomes scarce is attention, and those who write reports learn to write for scribes. The new asymmetry moves to data that is not public.
1. Everyone can read every report
At the step where prices gather information, the share of informed people settled at a point where one more person gathering information would not gain enough to cover the cost. Now only one thing changes: each household's scribe reads the reports and compares the fleets for it, at almost no cost.
Work through how that point moves. As the cost falls, information gathering that used not to pay now pays, and more people gather information. The more people gather it, the more prices reflect it before the news becomes public, and the less is left for each informed person. The new point lies where the benefit equals the very low cost: most households are informed, prices reflect the great majority of what can be read from public reports, and yet each informed person gains almost nothing from it.
Today's counterpart. Research by Lopez-Lira and Tang found that scoring news headlines with a large language model can predict subsequent stock price reactions. Their working paper, revised in 2025, reports that the returns to this strategy declined as large language models spread, which the authors consider consistent with prices becoming more efficient.
Where information gathering stops
The curve is what each informed person gains from public reports: the more people are informed, the smaller the gain. The horizontal line is the cost of analysis. Where the two meet is where the share of informed people settles. Drag the price of cognitive labour and see how that point moves.
Illustrative: forty households. When the informed share is f, the share of information that prices reflect before the news is public is f ÷ (f + 0.1). The part not yet reflected is what the informed earn from the uninformed, totalling 9.6 portions × (1 − share reflected) × (1 − f) a year for the whole village, split equally among the informed. The cost of analysis is originally 0.8 portions per person per year and falls in proportion to the price of cognitive labour. The vertical axis, in portions per year, uses a logarithmic scale.
2. The value of being first
Reading reports no longer gives an advantage, but being first still does. The news that a ship has come into port will always become public, yet whoever knows a moment earlier can buy ship shares before others know and sell them after the news is out. That moment's lead does not make the ship carry one more portion of cargo, nor does it improve anyone's plans. It only moves a sum of money from those who learned later to those who learned earlier.
For each household, being first is worth it; for the village as a whole, the race to be first only consumes resources. Every household hires a faster scribe and stations itself closer to the dock, spending more and more labour, while the sum that changes hands does not grow.
Today's counterpart. Budish, Cramton and Shim (2015) studied the speed race in high-frequency trading. Under continuous trading, a tiny speed advantage yields arbitrage profits, so trading firms compete to invest in faster connections. They propose instead to match orders in a batch at very short intervals, to end this race.
Being a moment first: the money that changes hands stays the same, the money spent keeps growing
News that a ship has come into port becomes public an hour later, and whoever knows first can earn 10 portions from those who learn later during that hour. Drag the number of households racing to be first, and compare the resources the village spends on the race with the value it brings the village.
Illustrative: each piece of news lets 10 portions change hands from those who learn later. The resources each racing household spends on faster scribes and closer positions are set by competitive bidding until the expected gain is zero, so the total approaches 10 portions (following the idea in Hirshleifer, 1971). Amounts in the bars are in portions.
3. The new scarcity: attention
Scribes can read every report, and they can also write more reports for each household: an analysis of every ship, a reminder every day, every possible opportunity. But a principal has only a little time each day to look at these things. The more information there is, the smaller the chance that any one item is seen. What is truly scarce shifts from information itself to the attention needed to look at it.
Once attention is scarce, whoever decides what the principal sees holds a great deal of power. A scribe that selects and ranks items for the principal can put the most important matter first, or it can put first the matter most favourable to the seller.
Today's counterpart. The home screens of news and investing apps are ranked by algorithms, and whether a given item appears first or twentieth makes a large difference to the chance that it is seen.
The more information, the scarcer attention
Each day a scribe writes a number of alerts and analyses for a household, and the principal can read only 8 a day. Drag the number of items written each day, then switch whose interest the scribe ranks by, and see what the principal sees.
Important and relevant to the principal's planPaid promotion by sellersOtherThe 8 items the principal sees
Illustrative: about one in ten items written is important and relevant to the principal's plan, about two in ten are paid promotions by sellers, and the rest are general news; each day the principal reads only the top 8 items (the scarcity of attention described by Simon, 1971).
4. Those being read change too
Shipowners discover that their reports are read first by scribes and only then by principals. Scribes judge prospects by counting the negative words in a report, so shipowners begin to weigh their wording: the same loss, described differently, earns a higher score from the scribe. Reports become easier for scribes to read and better at pleasing them. When the reader changes, the writer changes along with it.
Cao, Jiang, Yang and Zhang (2023) studied the annual and quarterly reports of U.S. listed companies. The more these filings are downloaded and read by machines, the more companies write them to be easy for machines to process and avoid words that count as negative in sentiment dictionaries. The change is especially marked after a dictionary is published and after new language models are released.
Today's counterpart. Corporate disclosure now faces two kinds of readers, people and machines. Those being read study those who read them, and the readers in turn must see through the new ways of writing.
Reports written for scribes
The same shipowner writes up the same quarter's loss as two reports. The scribe scores a report by its negative words. Switch between the two versions and see how the words and the score change, and whether the facts do.
Illustrative: the wording of both reports is made up; the scribe's score = 100 − number of negative words × 15. Cao, Jiang, Yang and Zhang (2023) found that the more machine readers a U.S. listed company has, the more it avoids words that count as negative in sentiment dictionaries.
5. A new asymmetry: data that is not public
Once the advantage of public reports is used up, what advantage remains comes from things the reports do not contain: the voyage records of one's own fleet, informants at the dock, data sold only to a few. Other people's scribes cannot read this information, so prices do not reflect it in advance, and those who hold it can still gain at others' expense. The more widespread scribes become, the less public information is worth and the more exclusive data is worth.
Katona, Painter, Patatoukas and Zeng (2025) studied satellite images of U.S. retailers' car parks. Investors who could buy this data profited by short selling companies with poor results before those companies released their earnings, while individual investors were buying during the same period. The authors conclude that unequal access to data can widen information asymmetry without immediately improving price discovery.
Today's counterpart. Alternative data such as satellite images, card transaction records and web traffic is bought mainly by a small number of large institutions.
The advantage moves from public reports to exclusive data
Divide the information that can give an advantage into two kinds: what can be read from public reports, and what only a few people hold. Drag the price of cognitive labour and see how much advantage each kind still gives.
Illustrative: the advantage of public reports is calculated with the model in the figure "Where information gathering stops", as the total portions earned each year by the village's informed households. The advantage of exclusive data is 3 portions a year, shared among the data holders, who make up one tenth of households, and does not change with the price of cognitive labour. Amounts are in portions.
6. The arithmetic does not change
Scribes make active selection much cheaper, but they do not change one thing: all the ship shares held by everyone, taken together, are all the ship shares there are. Before costs, active selectors as a group can only get the average. The extra cost they pay to select becomes smaller, so the margin by which they fall behind narrows, but it does not turn into a lead.
Today's counterpart. Stock selection tools built on large language models have greatly reduced the cost of active selection, but index products that hold everything in proportion to market share remain the cheaper choice.
Selection got cheaper; the arithmetic did not change
The average annual return on all ship shares is 6%. Those who passively hold everything pay only 0.1% in fees. Drag the cost of active selection and see the return of active selectors as a group after costs.
Illustrative: by the arithmetic of Sharpe (1991), the return of active selectors as a group before costs equals the market average. The bars start at 5% so that the differences are visible.
7. A market where everyone has an analyst: Civilization Simulator, version 16
The simulator unlocks "Analysts for all". Forty households decide whether to have a scribe analyse public reports for them; the cost of analysis falls with the price of cognitive labour. You can drag the price of cognitive labour and switch on or off "a few households hold exclusive non-public data", then watch how many households choose to analyse, how much public information prices reflect in advance, and how much the analysts and the holders of exclusive data each earn.
Civilization Simulator · Chapter 16: a market where everyone has an analyst
Forty households decide whether to have a scribe analyse public reports for them. The two lines show, as the price of cognitive labour falls from its original level to one hundredth of it, the households willing to analyse and the public information that prices reflect in advance; the dots mark the current price. Then switch on "A few households hold exclusive non-public data" and see where the advantage goes.
Information in pricesParameter: price of cognitive labourPersonal agentGuaranteeing agentsAnalysts for allAgents as market participants · Ch. 17
Households willing to analysePublic information reflected in prices in advance
Illustrative model: the same as in the figure "Where information gathering stops". The cost of analysing public reports is originally 0.8 portions per person per year and falls in proportion to the price of cognitive labour; the informed share settles where each person's gain exactly equals the cost. Exclusive data brings an advantage of 3 portions a year, shared by one tenth of households; the money comes from households that trade without this data, and it does not change with the price of cognitive labour. The horizontal axis is the price of cognitive labour on a logarithmic scale.
8. What this step leaves open
Once everyone has an analyst, several questions remain unresolved:
Analysts that resemble each other. Many scribes come from the same few operators, read the same reports and use the same methods. Will they reach the same judgment at the same moment?
The race to be first. Once the speed of understanding news converges, the race shifts to who is faster and who is closer to the source of the news. How can a market stop rewarding knowledge that is merely a moment earlier?
The power to rank. In whose interest does a scribe that selects and ranks items for the principal do its ranking?
Exclusive data. Non-public data is becoming ever more valuable. Who holds it, and to whom should it be disclosed?
What this step leaves open
Four questions that remain unresolved once everyone has an analyst. Click a question to see the assumption it hides.
Analysts for allPublic information is no longer an advantage
9. Implications for products
Analysis tools that everyone can use are turning the reading of reports into a basic capability. Three implications follow. First, drawing conclusions from public information is no longer a product's advantage; a product's value must shift elsewhere: helping users stick to their plans, checking agents, and controlling risk and cost. Second, when a product selects and ranks information for users, what it holds is the users' attention; ranking by the user's goals rather than by who pays is the precondition for such a product to be trustworthy. Third, do not promise users that they can beat the market with public information. Real information advantage comes from non-public data and greater speed, and ordinary users usually have neither.
Chapter summary: if everyone has an analyst
Four conclusions, each corresponding to one mechanism in the chapter.
Rerun: Grossman–StiglitzReading reports is no longer an advantage
Analysis is almost free, the number of informed people rises sharply, and prices reflect the great majority of public information.
Private and social valueBeing first moves money; it creates nothing
Knowing a moment earlier can earn money from others, but racing to be first only consumes resources.
AttentionWhat is scarce is time to look at information
The more information there is, the greater the power to rank; those being read also learn to write for scribes.
New asymmetryThe advantage moves to exclusive data
The arithmetic of active management does not change; who wins and who loses within the group depends on who has non-public data and who is faster.
References
Budish, E., Cramton, P., & Shim, J. (2015). The high-frequency trading arms race: Frequent batch auctions as a market design response. Quarterly Journal of Economics, 130(4), 1547–1621.
Cao, S., Jiang, W., Yang, B., & Zhang, A. L. (2023). How to talk when a machine is listening: Corporate disclosure in the age of AI. Review of Financial Studies, 36(9), 3603–3642.
Grossman, S. J., & Stiglitz, J. E. (1980). On the impossibility of informationally efficient markets. American Economic Review, 70(3), 393–408.
Hirshleifer, J. (1971). The private and social value of information and the reward to inventive activity. American Economic Review, 61(4), 561–574.
Katona, Z., Painter, M. O., Patatoukas, P. N., & Zeng, J. (2025). On the capital market consequences of big data: Evidence from outer space. Journal of Financial and Quantitative Analysis, 60(2), 551–579.
Lopez-Lira, A., & Tang, Y. (2025). Can ChatGPT forecast stock price movements? Return predictability and large language models. Working paper, arXiv:2304.07619.
Nobel Prize Outreach. The Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel 1978. NobelPrize.org.
Nobel Prize Outreach. The Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel 2001. NobelPrize.org.
Sharpe, W. F. (1991). The arithmetic of active management. Financial Analysts Journal, 47(1), 7–9.
Simon, H. A. (1971). Designing organizations for an information-rich world. In M. Greenberger (Ed.), Computers, Communications, and the Public Interest. Baltimore: Johns Hopkins Press.