Markets
Structure that a crowd produces without any member of it aiming at the structure.
A market looks like the sum of what its participants know: a spread that holds a range, prices that stay bounded, and coordination nobody arranged. It is tempting to read that as intelligence, aggregated. But take the intelligence out, and most of the structure remains. Each of these three panels takes a published market model, removes every trace of judgement, leaving agents that only follow a fixed rule, and checks what survives against a number the original paper computed in advance: the run threshold, the spread’s scaling with order flow, and the location of the phase transition.
Every card here reads the same way, and that is the point of the section rather than a weakness of it. Cut off the flow of orders, of depositors or of traders and there is nothing left: no book, no bank, no game. None of these structures produces its own supply. What separates them is what kind of structure survives that admission. The order book's spread is a statistic of its input, in the way the mean of a thousand dice rolls is a statistic of the dice. The bank run is a genuine second outcome that the same inputs also permit. The minority game's coordinated state is neither: it is a collective phase, with a location on an axis, that the population enters and leaves at a point none of its members can see.
Each card is one published model. Its small tag is an authored reading of that model against the taxonomy map, a teaching device, not a measurement. The kernel says nothing about markets.
Red, natural attractlet · § 7.4: “a naturally-occurring non-recursive structure that exhibits attractor-like behavior under external driving.” All three model cards carry it.
Gold, candidate · reading open: one card, Bitcoin, below the models. Its protocol alone is an attractlet, with its basins written in by its designer. Whether Bitcoin, drawn with the people it pays and recruits, is more than that is left open.
Neither the tag nor the panel reads the simulation. Any program that runs these models is an attractlet model under § 7.4, as on the Lorenz page.
And the direction of the debt matters. Economists and physicists built these three models, published them, and checked them against real exchanges and real bank failures long before this framework existed. The framework did not discover them and takes no credit for them. What it asks is narrower and comes second: what kind of structure is each of these actually producing?
What the panels do claim is arithmetic. Each one reproduces a number from the paper it cites, by an independent route, and shows its working on the page: the optimal deposit contract and the run threshold, the scaling of the spread with the arrival rates, and the location of the phase transition.
The Bank Run
Diamond and Dybvig's bank has two outcomes at identical fundamentals and nothing in its books decides between them. The contract that makes every depositor better off is the contract that can be destroyed by the belief that it will be. The safest bank is the useless one. Runnable now.
The Order Book
Every order placed at random. Nobody forecasts, nobody reads anything, nobody has a strategy. The book still has a bid-ask spread of a definite size, and the arrival rates predict that size before you run it. Runnable now.
The Minority Game
Traders who cannot talk to each other, sharing one public record too short to say anything useful. They coordinate anyway, six times better than coin flipping, and stop doing so at a transition whose location was calculated before this simulation existed. Runnable now.
The open reading
a published design, a live simulation, and a question left openBitcoin
A currency whose creator disappeared and which kept going. Every 2016 blocks it reads its own timestamps and resets its difficulty; every copy of the ledger refuses forged payments. The panel reproduces the white paper's attacker odds, then runs the race and finds the paper's formula was an approximation. Runnable now.
A reading from outside the models
a live event, no simulation, and a different tagData Center Return on Investment
Spending on AI data centers is running at roughly twenty times the revenue of the companies whose products justify it. AI labs, cloud companies, chip suppliers and utilities each keep the others going, and none of them is running the loop. The page asks what kind of structure that is, and what number would show whether each added watt still pays. A reading, no simulation.
The Crash of 2008 Through the Lens of RAPT
The same shape as the data-center reading, but finished. A self-feeding loop in housing finance ran on one belief, that prices would keep rising, looked strongest right before that belief broke, and could not restart on its own when it did. The page tells the crisis plainly first, then reads it as a structure. A reading, no simulation.
The Autocatalytic Sets of the Financial Markets
Four loops that make more of themselves, written out member by member: collateral and leverage, liquidity, deposits, and fund flows. Each one gets its circuit, the outside supply it runs on, the member whose loss stops it, and a dated episode when that member was in fact removed. A reading of published work, no simulation.
The Autocatalytic Sets of the Financial Markets, Two
What happens to a loop that already closes when one of its members is taken out and a different one put in. The rate that meters credit, the funds that supply short-term money, and the counterparty on the other side of a Treasury trade, each with dates. One of the three is still running. A reading of published work, no simulation.
Box and Papers
The secondary market in Rolex watches written out as a loop: the waiting list, the grey dealers, the price trackers, the counterfeiters, and the warranty card the street turned into proof of authenticity. Then the maker taking that proof back with its own certified resale. A reading of published work, no simulation.
The two ends of the axis: Living Structures → and Attractlets →.
← Back to the Laboratory