OpenAI announced on October 26, 2026, that its models had produced three results in harmonic analysis, theoretical computer science and mathematical physics. The company described a different release and examination path for each, and said its Exponential Time Hypothesis (ETH) result was verified in Lean. That formal check is separate from an independent expert review. OpenAI’s announcement lays out the claims and their publication paths.
NeoTeo previously covered OpenAI’s reported plan to publish mathematical results on GitHub. The October 26 announcement describes three specific workstreams and how the company says their results should be examined.
Three fields, three release paths
OpenAI says its models produced a proof related to wave behavior, an algorithmic result about ETH, and existence results for certain quantum field theories. The materials were handled differently in each case.
| Research area | Result OpenAI describes | Publication and examination path |
| Harmonic analysis | A proof of Sogge’s local smoothing conjecture for waves in every spatial dimension of at least two, with a claimed Kakeya-related consequence. | OpenAI released the natural-language output and Lean proof for researchers to examine. |
| Complexity theory | A deterministic subexponential-time algorithm for counting satisfying assignments of fixed-width CNF formulas, which OpenAI says refutes ETH in the 3-SAT case. | OpenAI made the natural-language work and Lean formalization public. Russell Impagliazzo, Ramamohan Paturi and Ryan Williams lead an external examination. |
| Mathematical physics | Existence results for certain interacting four-dimensional quantum field theories, including QCD and a unified completion of the Standard Model, as well as pure Yang–Mills theory. | The group led by David Gross chose to withhold the AI-produced artifacts while preparing its own write-up. |
The table captures OpenAI’s claims and the arrangements it announced, not a shared verdict on whether the results have been independently accepted. In the physics case, the group’s decision to prepare its own write-up also means its materials followed a different path from the work released for direct examination.
What Lean verification means for the ETH claim
Lean is a proof assistant used to check formalized mathematics. OpenAI says its system’s ETH result was verified in Lean, and separately announced an external examination led by Russell Impagliazzo, Ramamohan Paturi and Ryan Williams. The three organizers have said that their participation should not be treated as certification of the claim.
So does “verified in Lean” mean that the external group endorsed the theorem? No. It describes the formal-checking step OpenAI reported; the expert examination is a separate process. A status account checked through October 30, 2026, reported no published verdict from the group by that date.
What an ETH refutation would and would not imply
OpenAI’s ETH claim does not establish P = NP. A subexponential running time is not necessarily polynomial time, so the claimed result does not settle the P-versus-NP question. OpenAI has also said that finding algorithms and translating theoretical advances into practical improvements will require further research.
That distinction matters for anyone wondering whether the result would immediately make everyday computing faster. The claim concerns a theoretical complexity result; practical applications would require additional work.
The scope of OpenAI’s physics claim
OpenAI says its work establishes the existence of certain interacting four-dimensional quantum field theories, including QCD and a unified completion of the Standard Model. It also claims an existence result for pure Yang–Mills theory, which is part of the Clay Mathematics Institute’s Millennium problem. The announcement does not claim the separate mass-gap condition, so it does not claim to resolve the entire Yang–Mills problem.
The group led by David Gross chose to withhold the AI-produced physics artifacts while preparing its own write-up. For background on the strong-force context and the distinction between Yang–Mills existence and the mass gap, World of Math offers this explainer:
