1\section{Mechanisms: what the evidence says about the theory}2\label{sec:mechanisms}34Table~\ref{tab:predictions} listed six predictions. Reading them against Section~\ref{sec:results} gives5a mixed scorecard that is more informative than a clean confirmation would have been.67\paragraph{H1 (informativeness falls in the shared-model share): not supported on levels, supported on dynamics.}8The stylized model predicts that a common error component in the signals of a growing share of traders9leaves a persistent gap between price and value. The simulator shows no such gap: the tracking error10is, if anything, lower with LLM agents at every share. The reason is visible in the prompts and the11decision logs. The LLM agents are given the public present value and told how it moves with the next12dividend; most of them hold no private signal (they inherit the information set of the slot they replace,13and only fundamentalists are informed); their modal decision is a limit order near the public value.14Their ``common error'' is therefore not a biased signal about the fundamental but a common15\emph{anchor} on the public value, and a population anchored on the public value prices the stock at the16public value---which is close to the latent value except in the window between a latent shock and its17revelation. In that window the prediction holds: the shock is absorbed more slowly as the share rises18(Section~\ref{sec:results_dose}), and the persistence of returns (variance ratio 2.1, autocorrelation190.32 at 95\,\%) is the signature of a price that moves toward new information in many small steps against20a wall of stale limit orders. The misleading-news cell of the classical design shows what a genuinely21biased common signal does (Section~\ref{sec:results_classical}); the LLM cells did not inject one.2223\paragraph{H2 (coordination rises with the share): supported for synchrony, rejected for flow correlation.}24Decision synchrony is near its ceiling at every share, consistent with a single model answering a single25snapshot. Pairwise flow correlation falls with the share because the agents trade with one another; the26stylized model, in which every trader submits a market order against a common market maker, has no room27for this. The classical analogue shows that rule-based agents of one type are more correlated than28LLM agents, which qualifies the common intuition that language models are the more homogeneous traders:29in this economy they are more homogeneous than a mixed classical population and less homogeneous than any30single classical rule.3132\paragraph{H3 (bubbles and crashes rise with penetration): rejected; H3$'$ (leverage interaction): reversed.}33No bubble, crash or halt occurs in 112 live runs. With leverage present, the LLM agents reduce the34instability the leveraged traders create. Proposition~\ref{prop:spiral} is about an exposed cohort whose35positions are same-signed and leveraged; the LLM agents in this design are unleveraged and their36positions lean against the exposed cohort, so the proposition's premise is not met and its conclusion does37not apply. A design in which the LLM agents themselves carry leverage is the natural next test.3839\paragraph{H4 (diversity restores informativeness): supported in sign, small in size.}40Diverse prompts and the two-model mix lower coordination and shorten the absorption of the latent shock41by about four ticks, as Proposition~\ref{prop:diversity} predicts; they also lower liquidity provision,42which the proposition does not consider because its traders do not supply liquidity.4344\paragraph{H5 (communication raises synchrony): supported.}45The channel raises pairwise correlation to the highest level of all cells and the tracking error by a46small but precisely estimated amount; memory does the same, less strongly.4748\paragraph{H6 (safeguards trade depth for delay): not testable at these shock sizes.}49The breaker never fires; the position limit binds and removes liquidity without removing any50instability, because there was none to remove.5152\paragraph{The unifying mechanism.}53The one regularity that runs through every arm is liquidity provision by value-anchored agents who act54infrequently. It explains the tighter spreads and deeper books, the lower volume, the slower absorption of55latent information, the predictable returns, the trend followers' higher profits, the dampening of56leverage-driven fire sales, the liquidity cost of diversity (more market orders) and of position limits,57and the behaviour of the frozen library at $N = 1{,}000$. It is not the mechanism Section~\ref{sec:theory}58set out to test---correlated errors in a market-order population---but it is the mechanism the data59identify, and it suggests a different stylized model: a population of patient, public-value-anchored60limit-order traders whose aggregate order book slows price discovery of private information while61stabilizing prices against non-fundamental shocks. Whether that is good or bad for a market depends on the62mix of information and noise in its order flow, which the information-quality and misleading-news cells63begin to map.64