1\section{Conclusion}2\label{sec:conclusion}34When AI agents become economic agents, the economy acquires a new source of aggregate fluctuations that has nothing5to do with technology shocks: the mistakes of a few shared models, acting through millions of delegated decisions.6This paper has tried to say precisely what that means and, just as precisely, what it does not.78It means that a market for delegation sets the error-covariance structure of the economy. The delegation share and9the effective number of independent models are sufficient statistics for how much capital, wages, interest rates and10asset prices load on model noise, and the loading has a closed form. It means that whoever owns the compute that11delegated decisions run on captures a new kind of rent, and that better decisions are themselves a scale-dependent12return that concentrates wealth. And it means that markets populated by agents sharing one model are quieter and13more correlated at the same time, so that their fragility shows up conditional on a false shared signal, through14collateral, and depends on how execution turns correlated orders into a marked price.1516It does not mean that delegation is inefficient. In the benchmark economy the private and social values of17delegation coincide exactly; the externality that policy discussions take for granted exists only under18identified frictions, and we have given its size and sign under each. Model diversity is a policy instrument only19where those frictions bind; a delegation tax and a diversity mandate are substitutes there, and both are20unnecessary where they do not. Nor does it mean that language-model agents behave like people: our laboratory shows21what the architecture does to a market, under constraints that no model can override.2223Three extensions follow directly from the code. A Krusell--Smith version of Model B would put the delegation game24inside each period of an economy with aggregate risk and let the decision-quality channel and the model-noise channel25interact. A firm-side version would let managers delegate input choices and quantify the productivity effect of26advisory AI through reduced misallocation. And the empirical roadmap of Section~\ref{sec:calibration}---adoption,27transaction and broker data linked to the identity of the model behind each agent---would turn the effective number28of models from an illustrative parameter into a measured one. The last is the most important: the quantity that29matters for the economy is not how many foundation models exist, but how many of them make independent mistakes.30