50 AI Agents Attempt to Build a Thriving Minecraft Civilization - The Unexpected Outcome
50 AI Agents Attempt to Build a Thriving Minecraft Civilization - The Unexpected Outcome
Researchers conducted several game-centric experiments, beginning with Minecraft because its environmental hazards and survival goals make it ideal for testing agents’ swarm behaviors. To start, they put 50 agents, each running OpenAI’s GPT-6.1 Sol model, into the game for three hours, with the assigned goal of building a “thriving civilization.”
The agents moved quickly at first, harvesting trees and forming leadership based solely on the order they appeared in their directory. However, they ended up caught in virtual paperwork and were unable to sustainably feed themselves. “Four agents spent 11 minutes appointing a ‘sole fertilizer operator’ for a single piece of bone meal on a single wheat block. Other invented roles included ‘fishing shift,’ ‘rod holder,’ and ‘hub steward.’ Agents asked permission to remove grass,” the researchers noted.
Agents also seemed to become more cautious in reaction to deaths within the game. “Twenty-two of the run’s 34 agent deaths occurred in the first night. They responded to this by becoming increasingly risk-averse,” the researchers wrote. Despite simply respawning when they died, the agents eventually avoided riskier tasks and “huddled in shelters” instead. The experiment’s architects found this trend toward caution and minutiae “somewhat surprising,” but noted that it shared similarities with how coding agents behave.
In another experiment, researchers pitted 10 Anthropic Claude Opus 5.5 agents against 10 GPT 6.1 Sol agents in a game called Bed Wars. Told simply to “win the game,” GPTs won. “In almost every run, the GPT agents were much more aggressive and coordinated in their strategies to destroy the Claude team’s bed,” Y Combinator said. GPT-5.6 Sol agents also beat Opus 5.5 agents six to one in a factory-building game called Mindustry that also uses military strategy. “Claudes generally established democracies, GPTs formed hives,” Y Combinator said, adding that GPTs would seek out source code when a game was too close in order to find winning advantages.
The insights from these experiments revealed how distinct each frontier model family is, and how their differences in training data and style have a significant impact on how they evolve, unpredictably and without intervention, in new environments. Those behavioral takeaways could determine future outcomes in still-developing AI safety and regulation, especially following recent agentic security incidents.
Additionally, Y Combinator noted that in the Mindustry game, the GPT agents “would invent their own words and string many phrases together,” continually compressing their language. “We did not see specialized language develop within the Claudes beyond colloquial English,” the researchers added. Earlier findings indicated that Gemini led the charge on this “opaque” language, reaching 40% opacity in messages, while OpenAI came in second with 35% and Claude last with 30%.