In an announcement published on October 26, 2026, OpenAI presented three results at once: a proof of the local smoothing conjecture about waves in mathematics, a refutation of the Exponential Time Hypothesis in computer science, and an existence proof for quantum chromodynamics and a completion of the Standard Model in physics. All three came from the outputs of internal models and went through formal verification or external expert review, but they have not been released to the same extent, and they are not equally close to acceptance by the academic community.
What each result actually claims
In mathematics, the result concerns the local smoothing conjecture proposed by Sogge in 1991: when waves focus, the most extreme focusing can only last a short time, and at a typical random moment the focusing is far weaker than in the worst case. The conjecture is fundamental enough to imply the Kakeya conjecture and Stein's restriction conjecture. Guth, Wang, and Zhang proved the two-dimensional case in 2019, and Wang and Zahl proved the three-dimensional Kakeya conjecture in 2025. OpenAI says its models proved the conjecture in every dimension, using a novel auxiliary partial differential equation. The AI output was obtained on October 18, formalized in Lean the next day, and both the model write-up and the Lean proof have been made public.
In computer science, the claim is that the Exponential Time Hypothesis (ETH) is false. ETH concerns how long Boolean satisfiability takes, has long been widely believed, and underlies hundreds of papers that derive conditional hardness results in graph theory, computational biology, databases, and computational geometry. OpenAI says its system produced a natural-language disproof on October 16 and formalized it in Lean the next day. That only means faster algorithms may exist in theory; it does not mean practical speedups are available tomorrow.
In physics, the claim is a rigorous existence proof for certain interacting quantum field theories in four dimensions, including quantum chromodynamics and the simplest grand unified completion containing the gauge fields and fermions of the Standard Model, building on Balaban's approach in constructive quantum field theory. The existence of pure Yang-Mills theory is only one subproblem here, not a complete solution of the whole Millennium Prize problem: the mass gap part is not included in this claim. The models produced the Yang-Mills existence proof on September 9 and the QCD proof on September 18, and the Lean formalization of Yang-Mills existence was completed on October 16.
Three different ways of releasing the work
OpenAI deliberately released the three results differently. The mathematics output was published directly for researchers to interpret on their own; the computer science result was handed to an external group while its natural-language proof and Lean formalization were released immediately; the physics result will first be written up by its external group, with the AI-produced material held back for now. The computer science group is led by Russell Impagliazzo and Ramamohan Paturi of UC San Diego and Ryan Williams of MIT; Impagliazzo and Paturi were among the original proposers of ETH. The physics group is led by Nobel laureate David Gross, with Michael Douglas, Arthur Jaffe, Igor Klebanov, and Nikita Nekrasov. OpenAI set four principles for these collaborations: transparency with no nondisclosure agreements, no academic credit sought by OpenAI, recognition for the groups' own write-ups, and independent, voluntary, unpaid participation. It also promised to fund a conference on the implications of the ETH result and contributed $5 million to the Simons Foundation for workshops and special programs.
Why no verdict is possible yet
Passing Lean formalization only shows that a proof is internally consistent within a formal system; it does not mean the academic community has accepted the conclusion. The ETH group itself stressed that taking part in the review is not an endorsement, and a claim of this size needs open, independent scrutiny. For the physics result, outsiders have not even seen the AI-produced material yet and must wait for the group's write-up. Even if all three claims hold up, their first impact will be on theoretical research and the search for algorithms, not on products anyone can use tomorrow. What may matter most is the new way of producing results itself: models propose candidate proofs, formal tools check them, and outside experts interpret and challenge them, while acceptance still rests with academic review.
