Anthropic and OpenAI both say they refuted the Exponential Time Hypothesis with a 2^o(n) algorithm for 3-SAT, found on the same day. Before anyone rewrites their SAT solver: Anthropic's own bound at five clauses per variable doesn't beat brute force until n > 10^10000. The Lean proof checks out, but it was written and reviewed by AI, and no human has checked the statement says what ETH says.https://takschdube.substack.com/p/anthropic-and-openai-say-they-refuted#ComplexityTheory #AI
Anthropic and OpenAI both say they refuted the Exponential Time Hypothesis with a 2^o(n) algorithm for 3-SAT, found on t...
