Back to news
researchOpenAI2026-10-26

OpenAI says AI proved the local smoothness conjecture and disproved the Exponential Time Hypothesis

OpenAI published a paper on working with academic communities on Oct 26 describing model results across mathematics, computer science and physics, with artifacts uploaded to GitHub and all results awaiting independent verification. Mathematically, the models produced a proof of the local smoothness conjecture in all dimensions, resolving the Kakeya conjecture and Stein restriction conjecture, based on a new auxiliary partial differential equation, with the model output on Oct 18 and Lean formalization the following day. In computer science, OpenAI says the models disproved the Exponential Time Hypothesis, the widely held assumption that 3-SAT has no sub-exponential general solution, in a paper titled The Exponential Time Hypothesis is False, with a natural-language counterexample found on Oct 16 and formalized in Lean the next day. Russell Impagliazzo and Ramamohan Paturi of UC San Diego and Ryan Williams of MIT will examine the result. In physics, OpenAI claims a rigorous mathematical proof of existence for four-dimensional gauge field theory with fermions, completing the Yang-Mills existence proof that forms part of a Clay Millennium Prize Problem, with Nobel laureate David Gross invited to lead verification. The Lean-verified time bound differs somewhat from the paper's strongest result.

OpenAI挂谷猜想指数时间假设杨-米尔斯理论科学发现Lean