On October 26, 2026, OpenAI announced three results produced by its internal models in mathematics, computer science and physics, covering harmonic analysis, complexity theory and quantum field theory. The company described the outputs as raw research data that still needs expert interpretation, and said it had brought in outside academics to decide how each result should be explained and released.
The mathematics result concerns the local smoothing conjecture, proposed by Sogge in 1991, which holds that the most extreme focusing of waves is short-lived. OpenAI says its models produced a proof in every dimension, which would also resolve the Kakeya conjecture and Stein's restriction conjecture in all dimensions. Guth, Wang and Zhang proved the two-dimensional case in 2019, and Wang and Zahl proved the Kakeya conjecture in three dimensions in 2025. In computer science, OpenAI reports that its system proved the Exponential Time Hypothesis is false, a result it says was verified in Lean. The hypothesis has been widely accepted and used in hundreds of papers to derive conditional hardness results. These claims come from OpenAI's own announcement, and the article does not report independent peer review. The release approach differs by result: the mathematics output is being published directly, the computer science write-up and Lean proof are being published on GitHub by an external collaboration, and the physics materials are being held back while that collaboration writes up its own account.
For Australian businesses and professionals, these results change nothing in the short term. They are frontier research with no product, pricing, or compliance impact, and the Privacy Act is not engaged. The most relevant audience is researchers and engineers in algorithms and computational complexity, whose published work may rely on conditional hardness results built on the Exponential Time Hypothesis and may need rechecking as expert review proceeds. Otherwise, the practical step is to watch the public GitHub release and the expert write-ups before drawing conclusions.
Key Points
OpenAI says its internal models produced a proof of the local smoothing conjecture in every dimension, which would also resolve the Kakeya and Stein restriction conjectures.
The company claims its system proved the Exponential Time Hypothesis is false, with the result verified in Lean.
OpenAI is releasing each result differently, based on advice from external expert shepherds, with the physics materials held back by the collaboration.
Outside collaborations operate under transparency, no-academic-credit, recognition and independence principles, and OpenAI does not claim credit for the discoveries.
For most Australian businesses, there is no immediate impact; researchers relying on ETH-based hardness results may need to review their work.