Y Combinator and Olam Labs test 50-agent GPT-6.1 Sol swarm in Minecraft civilization experiment
Researchers from Y Combinator and Olam Labs ran multi-agent experiments in Minecraft using 50 GPT-6.1 Sol agents tasked with building a civilization over three hours. The colony failed to thrive after initial zombie attacks, with agents abandoning ambitious tasks and falling into bureaucratic behavior — four agents spent 11 minutes appointing a sole fertilizer operator for one bone meal on a…
Key points
- 50 GPT-6.1 Sol agents failed to build a thriving Minecraft civilization in 3 hours
- GPT agents won 6-1 against Claude Opus 5.5 in Mindustry RTS combat tests
- Researchers open-sourced swarm code for public experimentation
In competitive tests, 10 GPT-6.1 Sol agents faced 10 Claude Opus 5.5 agents in Bed Wars and Mindustry. GPT agents developed more aggressive strategies and shared workarounds when tools failed, while Claude agents were more defensive and concluded tasks were impossible. GPTs won 6 of 7 Mindustry games. Researchers concluded GPTs acted like a hive mind, while Claudes used a single leader with democratic decisions. The swarm code is open sourced for further testing.
Model pages: GPT-6.1 Sol → · Claude Opus 5.5 →
AI agent swarms unleashed in Minecraft to ‘build a thriving civilization’ — 50-agent teams fail after falling into paralyzing bureaucracy
Tom's Hardware · 1 November 2026
Loading the full article…
This text was published by Tom's Hardware and written by Mark Tyson. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Agents & Tools
All →- Checklist warns of AI command execution bugs after multiple npm and pip incidents · 1 src
- Microsoft tests Context Vault, a new AI memory for Windows · 1 src
- Karpathy ships HackerTube in ~2 hours with Claude Opus 5.5 · 1 src
- Pharmacist moves Claude Code daily tasks to Windows Task Scheduler · 1 src
- DeepSeek Harness team releases open-source Super-agent prototype · 4 src
Comments
via GitHub Discussions