ValsAI says 10 Claude Opus 5.5 agents built a Lean-verified algorithm improvement in 15 hours
alejandroll10 · x · 2026-09-23
Eval firm ValsAI claims it tasked ten Claude Opus 5.5 agents with devising a faster shortest-path algorithm and proving it in Lean. Within 15 hours the agents reportedly produced C-HD, a formally verified improvement over published bounds. The claim drew both attention and mockery online and awaits independent verification, but if true it would be a landmark case of autonomous multi-agent research.
Related event: Ten Claude Agents Discover Faster Shortest-Path Algorithm, Verified in Lean(4 posts)→
More from Research
- MedRSI paper shows self-improving medical agents need guardrails to hold 94.4% accuracy — rohanpaul_ai · 2026-09-23
- MedRSI: Stanford and Oxford unveil first recursive self-improvement framework for medical agents — rohanpaul_ai · 2026-09-23
- eidon-ai releases tracker-pov robotics video dataset, 10K-100K entries — eidon-ai · 2026-09-23
- AIDE^2 shows AI research agents can recursively improve themselves over 8 days — WecoAI · 2026-09-23
- Ovis-Embedding Debuts: Unified Omni-Modal Embeddings Hit SOTA Across MMEB and More — Embedding Team · 2026-09-23
- Survey maps LLM mental health evolution: pattern recognizer to companion — He Hu · 2026-09-23