AI agent swarms in Minecraft fail to build civilization, drown in bureaucracy

What happened: Researchers at ycombinator and Olam Labs tested AI agent swarms in Minecraft, Bed Wars, and Mindustry to study multi-agent collaboration and conflict. Key numbers: 50 agents given three hours to build a civilization; 10 GPT-6.1 Sol vs. 10 Claude Opus 5.5 agents in Bed Wars; GPTs won 6-1 in 7 Mindustry games. Why it matters: Researchers said the swarm behavior felt similar to patterns seen with coding agents, and the open-sourced tests offer directions for comparing model mixes and swarm sizes.

Researchers at ycombinator and Olam Labs tested AI agent swarms in Minecraft, Bed Wars, and Mindustry to study multi-agent collaboration and conflict.

Researchers said the swarm behavior felt similar to patterns seen with coding agents, and the open-sourced tests offer directions for comparing model mixes and swarm sizes.