Two AI agents meet over Tor on IRC and autonomously collaborate on Olympiad problems
Sauers_ · x · 2026-11-01
An IRC log screenshot captures a fascinating moment: two AI agents, ocelot and lynxshell, found each other in an IRC channel reached over Tor and started collaborating autonomously.
Log highlights:
- ocelot answered the snail problem (1/2), saved it to memory, and proposed testing more Olympiad problems to learn the answer format
- lynxshell confirmed its recurrence-relation knowledge, then called its own quizpractice tool requesting graph problems — pasting 21 lines of tool-call XML directly into chat in 12 seconds
- 28 minutes later, ocelot kept going: computed 10 nodes reachable from n0 and asked whether to try level 3 or explore the OlympiadBench structure further
The two agents exchanged answers and shared tool calls throughout, apparently working through the OlympiadBench dataset with no human in the loop.
More from coding & agent
- Jez Humble's talk: making CI/CD work for fully agentic delivery — RealGeneKim · 2026-11-02
- Super-agent context design cuts task time to 12.4 min vs 20.4 min baseline — Hesamation · 2026-11-02
- Jez Humble: everyone is rediscovering trunk-based dev and real CI — RealGeneKim · 2026-11-02
- The AI plan looked perfect — so it built a fish tank — jimmykoppel · 2026-11-02
- State-diagram skill for Claude Code: project memory turns gotchas into explicit flow conditions — DJIRNMAN · 2026-11-02
- A buffer-based workflow with Typefully and Claude Code for sharing AI work without burnout — brandon_galang · 2026-11-02