OpenAI models solve major math and physics conjectures, results handed to academic groups

OpenAI is framing these manuscripts as raw research data rather than finished papers, and it is testing three distinct release strategies chosen with external experts. For the math result, the AI output is released for researchers to organize their own explanations. For the computer science result, an external group will explain the work while the AI output is made public. For the physics result, an external group will write up the work itself and decide when to release materials. All three collaborations follow principles of transparency, no academic credit for OpenAI, recognition for the groups, and independence.

The results themselves span three fields. The math proof resolves the local smoothing conjecture in every dimension and, as a consequence, also settles the Kakeya and Stein restriction conjectures. The computer science result refutes the Exponential Time Hypothesis, a widely believed conjecture about the hardness of Boolean formula satisfiability that has underpinned hundreds of conditional hardness results. The physics result gives a rigorous existence proof for certain interacting quantum field theories in four dimensions, including QCD and a unified completion of the Standard Model, and includes a proof of pure Yang–Mills existence as a subproblem.

OpenAI acknowledged tradeoffs in its approach: expert shepherds can help unpack and contextualize the results, but choosing initial shepherds also shapes who gets involved first and creates information asymmetry. The collaborations complement a recently announced $5 million contribution to the Simons Foundation that will support workshops and conferences to help researchers build on AI-produced results.