OpenAI Releases Three New Science Findings, Adds Expert Review

OpenAI published three new findings in mathematics, computer science and physics on Monday, less than a month after it released hundreds of manuscripts claiming solutions to long-standing math problems, The Verge reported.

The company posted the latest results to a GitHub repository the same day. It called the work "major" discoveries, according to The Verge, which first reported the release late Monday.

What did OpenAI release this time?

The batch covers one result each in math, computer science and physics. The math entry is a proof involving local smoothing for the Euclidean wave equation. OpenAI says the proof uses an auxiliary partial differential equation to establish Sogge's local smoothing for waves in every spatial dimension of two or higher. The same proof also yields the Kakeya maximal and dimension conjectures and Stein's spherical restriction estimates across the interior of the conjectured range, per the GitHub posting cited by The Verge. OpenAI said it is releasing that model output directly, the same approach it used in its October 6 release.

Why did OpenAI bring in outside reviewers this time?

The company says it consulted external experts before publishing the computer science and physics results. The Verge reports OpenAI describes those outside reviewers as "shepherds" who guided how the findings were framed and released.

OpenAI describes the outside reviewers guiding the computer science and physics results as "shepherds."

For the math result, those advisers took a different approach. Rather than shepherding the framing, they recommended OpenAI release the raw AI output and let researchers work through their own explanations of the proof, according to The Verge.

How does this compare to the October 6 release?

OpenAI's October 6 release involved hundreds of manuscripts the company said contained solutions to long-standing mathematics questions, The Verge reported. Monday's release is smaller — three results instead of hundreds — but adds the outside-expert review step that was not part of the earlier release. The math result in Monday's batch follows the same direct-release format as October 6: the model's output goes out without an intermediate human rewrite.

What details are still missing?

The Verge's report does not spell out the specific computer science or physics findings beyond noting that external shepherds reviewed them before release. OpenAI has not published additional explanatory material beyond the GitHub posting referenced in the report.

What should researchers watch for next?

• Whether independent mathematicians publish verification or critique of the wave-equation proof, including its claimed links to the Kakeya conjectures and Stein's restriction estimates.
• Whether OpenAI names the "shepherd" reviewers or describes their vetting process in more detail.
• Whether future releases keep the split approach — direct AI output for math, expert-shepherded framing for computer science and physics — or move toward one standard.
• Whether academic institutions formalize the collaboration model OpenAI is testing with outside experts.

OpenAI has not announced a timeline for its next release of this kind, based on the material reviewed.