AI BriefingClaude Mythos Solves OpenAI's Landmark Erdős Problem with a 'Cute, Simple Proof'
Key point
Anthropic's **Claude Mythos** solved the **Erdős** math problem that OpenAI had solved, using a simple and clear proof method.
Details
Anthropic engineer Sholto Douglas revealed that Claude Mythos solved the Erdős unit-distance conjecture, a combinatorial geometry problem that had been unsolved since 1946, with a "cute, simple proof." This suggests strong potential for AI to lead mathematical discovery.
This achievement was made through a testing system utilizing Claude Code instances. The system operates as follows:
- Individual Claude Code instances with Mythos privileges receive the problem and develop solution paths.
- One instance summarizes the solution and distributes it to other instances working independently.
Mythos frequently found different paths than OpenAI's model, and mathematician Daniel Litt evaluated the result as "slightly less satisfying" than OpenAI's while also noting that Mythos had found OpenAI's solution as well.
Meanwhile, Google DeepMind's AlphaProof also solved a similar problem, but differs in that it relies on Lean, a formal proof language. Additionally, Claude Code operates as an agentic harness rather than a pure LLM.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Anthropic Pacing to $100B ARR; Rumors of OpenAI Solving Navier-Stokes Spark Controversy
Claude Code Fails to Load AGENTS.md When Telemetry Is Disabled, With WorkaroundOpenAI and Anthropic's RSI Strategy and the Uncertainty of AI Profitability