OpenAI announced on October 26, 2026 that its internal AI models produced three research results: a proof in mathematics, a disproof in computer science, and an existence proof in physics. The company calls these manuscripts raw research data and says outside experts will act as shepherds, organizing groups to turn the output into scientific understanding. Each result follows a different release strategy, and the collaborations will be transparent, unpaid, and independent, with no academic credit for OpenAI.
The mathematics result is a proof of the local smoothing conjecture in every dimension. The conjecture, proposed by Christopher Sogge in 1991, concerns how waves behave as they become focused. OpenAI says its models produced the proof on October 18, formalized the next day in Lean, software that checks every step of a proof. A natural language writeup and the Lean proof are public on GitHub.
The computer science and physics results
The computer science result claims that the Exponential Time Hypothesis is false. The widely believed conjecture concerns computational complexity and is related to the P vs. NP problem. This result could stimulate the search for faster computational algorithms. The disproof was produced on October 16 and verified in Lean the next day. All raw data will be public, and OpenAI will fund a conference on the implications.
The physics result is a rigorous existence proof of certain interacting quantum field theories in four dimensions, including quantum chromodynamics, which describes how quarks and gluons interact through the strong force, and a unified completion of the Standard Model. It also includes a proof of pure Yang-Mills theory, which describes gluons interacting with themselves and is part of a Clay Mathematics Institute Millennium Prize problem. Top scientists will review and write up the result before AI -generated sources are released.
If confirmed, each result would be a landmark. Because the proofs are machine-checked, debate shifts to whether the formal statements faithfully capture the original conjectures. The release is also an experiment in how AI discoveries should be verified, shared, and credited, and OpenAI says it welcomes feedback.