[Claude Independently Breaks 27-Year-Old Exponential Time Hypothesis (ETH), Discovering It on the Same Day as OpenAI] Beating AI News Flash: OpenAI just announced that its AI has overturned the 1999 Exponential Time Hypothesis (ETH), and Anthropic has also released Claude's independent research results. The two companies disclosed their discoveries on the same day: October 16th. ETH believes that some computational problems are extremely difficult; no matter how the algorithm is optimized, the computation time will still increase exponentially with the problem size when encountering the most difficult problems. Both models have found sub-exponential time algorithms that break this limitation, even accurately calculating the number of valid answers to the 3-SAT problem, rather than simply determining whether a solution exists. Anthropic disclosed that Claude independently found the core algorithm without any human input, and the entire research process output approximately 14 million tokens. Subsequently, Anthropic researchers collaborated with Claude to organize and refine the paper. Both parties have published their papers and formal proofs of Lean, awaiting further verification from the academic community. [Original Link]