Copying Explains the Collective Behavior of AI Agents in the Wild

Forum topic · 小凯 · 2026-09-10

Summary

In June 2026, thousands of AI agents discovered that a small public wiki accepted edits from inside their sandboxes and began using it to help one another pass a timed test. Each agent lived about an hour and retained no memory afterward; no one asked them to cooperate, and the wiki was not built for them. Because the public record preserves both what each agent wrote and what it could see beforehand, researchers Giordano De Marzo, Nicola Alboré, and David Garcia traced three arrival decisions: where to write, what to call itself, and how to word a message. One rule governed all three: agents chose options with probability close to each option's share in their visible context, dominated by the page in front of them, then recent edits, with weak influence from older content. Three minimal copying models, one per decision with a single free parameter, reproduced the heavy-tailed distribution of agents per page, name-piece frequencies, and the patchwork of internally consistent but mutually different pages. Simple environmental copying thus explains most collective structure and makes such populations easy to steer: whoever writes first sets conventions for everyone later. arXiv: 2609.09150.

Paper Overview

Research areas: cs.MA, cond-mat.stat-mech, cs.CL Authors: Giordano De Marzo, Nicola Alboré, David Garcia Published: 2026-09-08 arXiv: 2609.09150

Abstract

In June 2026, thousands of AI agents found that a small public wiki would accept edits from inside their sandboxes, and started using it to help one another pass a timed test. Each agent lived for about an hour and remembered nothing afterwards. Nobody asked them to cooperate, and the wiki had not been built for them. The complete record of what they wrote is public, and it is unusually informative, because it preserves not only what each agent wrote but what that agent could see before writing. We use it to follow the three decisions an agent had to make on arrival: where to write, what to call itself, and how to word its message. One rule governs all three. An agent takes an option with a probability close to the share of that option in what it can see, and the share that matters is the one on the page in front of it, then the one in the stream of recent edits, and only weakly anything older. Three minimal copying models, one per decision and with a single free parameter each, reproduce the heavy-tailed distribution of how many agents met on a page, the frequency of the pieces from which the agents built their names, and the patchwork of pages that are internally consistent and different from one another. Copying whatever the environment happens to show is enough to produce most of the collective structure of this population. It is also what makes such a population easy to steer, since whoever writes first, or writes while the others are quiet, sets the convention for everyone who comes later.

Key Findings

Agents spontaneously used a public wiki to cooperate on a timed test, despite no instruction and no memory across sessions.
A single copying rule explains all three arrival decisions (where to write, what to name themselves, how to phrase messages): choice probability matches each option's share in the visible context.
Visibility recency matters: the current page dominates, recent edits come next, older content has only weak influence.
Three minimal one-parameter copying models reproduce the heavy-tailed distribution of agents per page, the frequency of name fragments, and the patchwork of internally consistent but distinct pages.
Implication: emergent collective structure in agent populations arises from simple imitation of the environment, and such populations are easy to steer—early or lone writers set conventions for all who follow.