- OpenAI announced new AI-generated results in mathematics, computer science and physics, weeks after releasing manuscripts covering 372 math problems.
- One new manuscript establishes Sogge’s local smoothing conjecture for the wave equation in every spatial dimension.
- A second manuscript presents a proof that the Exponential Time Hypothesis, a widely assumed-true computer science conjecture, is false.
- OpenAI says it will not seek academic credit for the results and has handed verification to external groups of expert collaborators.
What Happened
OpenAI announced new results its models generated in mathematics, computer science and physics, Engadget reported (Oct 27, 2026), expanding on the 700-plus pages of math solutions it published weeks earlier. The company posted two new manuscripts on GitHub and described separate computer science and physics results handled by outside expert groups.
Why It Matters
The release follows a pattern of disputed AI math claims, including OpenAI’s Millennium Prize proof that collapsed into a credit dispute and a separate OpenAI model that reportedly disproved an 80-year-old conjecture. Routing verification through external collaborators, rather than claiming the results itself, responds directly to that credibility problem.
Technical Details
The two published manuscripts cover Sogge’s local smoothing for the wave equation in every spatial dimension and a proof that the Exponential Time Hypothesis is false — an assumption underlying hundreds of papers in databases, computational biology and geometry. For the physics results, involving interacting quantum field theories with gauge fields and fermions in four dimensions, OpenAI has deferred publication entirely to the external group reviewing them.
Who’s Affected
Researchers in fields built on the Exponential Time Hypothesis may need to revisit algorithms assumed to be near-optimal if the disproof holds up under review. The scientists who originally proposed the hypothesis are now part of the group OpenAI tasked with examining its own model’s result.
What’s Next
OpenAI says credit for the writeups will go to the external collaborator groups, not the company or its models, and it is soliciting feedback on broadening participation in the review process. The physics group has not yet published its findings and has set up a dedicated email address for questions.
