OpenAI says its models produced 3 results in maths, computer science and physics
OpenAI says its internal models proved a wave-focusing conjecture, proved a logic-puzzle hypothesis false and proved that theories including quantum chromodynamics exist. Outside experts recommended a different way of sharing each.
A conjecture about how waves focus has been proved in every dimension of space, OpenAI says. A conjecture is a mathematical claim that has not yet been proved.
The post, dated 26 October 2026, announces three results from the company's internal models, in mathematics, computer science and physics. All three are OpenAI's own account.
The maths result concerns the local smoothing conjecture, proposed by Sogge in 1991. It says the most extreme focusing of waves must be short-lived.
The post says the models proved it in every dimension, which also resolves a related problem called the Kakeya conjecture. OpenAI says it obtained the output on 18 October 2026 and had it formalised in Lean, a computer system for checking proofs, the next day. The write-up and Lean proof are on its public GitHub page.
The computer science result concerns the Exponential Time Hypothesis, a widely believed idea about how hard it is to solve logic puzzles called Boolean satisfiability problems. The post says the system proved it false, and that Lean then verified the result. Turning this into practical improvements, the post adds, will need significant further research.
A group led by Russell Impagliazzo and Ramamohan Paturi of the University of California San Diego, and Ryan Williams of the Massachusetts Institute of Technology, will verify this work, the post says. OpenAI has agreed to make all raw data public so the community can check it independently.
The physics result is what OpenAI calls a mathematically rigorous proof that theories including quantum chromodynamics exist. The Standard Model describes three of the four fundamental forces, all except gravity, and quantum chromodynamics covers how quarks and gluons interact through the strong force.
The result includes a proof that pure Yang–Mills theory exists, part of a problem on the Clay Mathematics Institute's Millennium Prize list, the company says. It says it finished formalising that proof in Lean on 16 October 2026.
Nobel laureate David Gross will lead the physics group, with Michael Douglas, Arthur Jaffe, Igor Klebanov and Nikita Nekrasov. The group has decided to hold back the AI-produced materials and write the work up itself.
The post describes three release plans, recommended by outside experts it calls shepherds. For maths, the AI output is public, and researchers organise their own explanation. For computer science, a group checks the work and the data is public. For physics, a group decides what is released.
OpenAI names one limit: choosing the first experts shapes who gets involved first and gives some people information before others. The post says: "We do not yet know which of these three approaches will be most useful."