Olam Labs puts AI swarms in games to test coordination under pressure

Olam Labs and Y Combinator's Labs group reported in an October 2026 report posted October 29 that 50 AI agents attempting to build a Minecraft settlement rationed food after a deadly first night, and that a GPT-based team won 6-1 across seven Mindustry battles. Olam Labs co-founder and CEO Om Buddhdev led the work, which the company frames as a case for simulations as infrastructure for evaluating agent coordination under pressure. The small scale and paired model-harness setup show how much evidence repeatable agent evaluations still need.

Olam Labs puts AI swarms in games to test coordination under pressure Olam Labs and YC Labs report a 6-1 GPT win in Mindustry and a 50-agent Minecraft colony rationing food after a deadly first night. By RuntimeWire Staff https://runtimewire.com/author/runtimewire-staff · Published Primary source: Hacker News https://labs.ycombinator.com/projects/games/index.html Why it matters Olam is making a case for simulations as infrastructure for evaluating agent behavior. The experiments make coordination visible, while their small scale and paired model-harness setup show how much evidence repeatable evaluations still need. Om Buddhdev https://x.com/sensho?ref=runtimewire , co-founder and CEO of Olam Labs https://olamlabs.ai/?ref=runtimewire , teamed with Y Combinator's Labs group to study how AI agents organize when they share a goal and have room to act. In an October 2026 report https://labs.ycombinator.com/projects/games/index.html?ref=runtimewire posted on October 29th, the groups describe 50 agents attempting to build a Minecraft settlement and two model teams adapting across seven Mindustry battles.… When you purchase through links in our articles, we may earn a small commission https://runtimewire.com/affiliate-monetization-standards . This doesn’t affect our editorial independence.