OpenAI has published a 67-page paper claiming its AI models have disproved the Exponential Time Hypothesis, a computer-science conjecture from 1999. The news briefly confused crypto traders on X, who initially reacted as though the announcement concerned Ethereum's ETH ticker. One user joked that algo traders had slaughtered $ETH, while another quipped that Vitalik Buterin had gone quiet. However, the ETH in question has no connection to Ethereum, its price, or its co-founder.
The Exponential Time Hypothesis was proposed by computer scientists Russell Impagliazzo and Ramamohan Paturi with Francis Zane. It states that a classic logic puzzle known as 3-SAT cannot be solved much faster than by trying every possible combination. OpenAI says its AI found a deterministic algorithm that counts every satisfying assignment of such puzzles in subexponential time, meaning the work grows more slowly than any exponential in the number of variables. Counting solutions is considered harder than deciding whether one exists, so the harder result implies the easier one.
OpenAI said the AI made the discovery on October 16, and the result was formalized in Lean, a software tool that mechanically verifies every logical step, on October 17. Both the paper and the formalization are available on GitHub. Impagliazzo, Paturi, and MIT's Ryan Williams are leading an independent examination, but they cautioned that their involvement should not be taken as certification of the claim. They said a result of this significance requires careful, independent scrutiny.
Anthropic also entered the discussion by posting a paper credited to Claude titled "A 2^o(n) Time Algorithm for 3-SAT," dated October 26. The repository says an internal research Claude model discovered the algorithm and that the result is purely asymptotic. OpenAI stressed that the finding does not show that every NP problem can be solved in subexponential time, so the P versus NP question remains open. OpenAI also announced a $5 million contribution to the Simons Foundation to fund workshops, conferences, and special programs examining AI-produced results.
