OpenAI Claims Three Science Breakthroughs, Including a Shocking Math Result

What's different this time is how OpenAI is handling release. The company says external experts it calls "shepherds" are guiding the computer science and physics results, while for the math problem it says advisers told it to release the raw model output so researchers could work through their own explanations.

In computer science, OpenAI says an independent working group led by Russell Impagliazzo, Ramamohan Paturi and Ryan Williams will examine and explain the proof. At the group's request, the materials are being made public immediately so others can scrutinize them. Impagliazzo and Paturi are at UC San Diego and were part of the group that originally proposed ETH in 1999; Williams is an MIT professor.

For physics, OpenAI says David Gross will lead a group that includes Michael Douglas, Arthur Jaffe, Igor Klebanov and Nikita Nekrasov to shepherd the work. The company admits this release approach could have its own problems. Just a day earlier, an AGMAI panel of math advisers recommended that AI companies not adopt the practice of designating certain people to receive output first, citing unequal access, borrowed reputation and biased selection. The panel instead recommended releasing results on scholarly repositories, with Lean-formalized proofs, information on how problems were chosen and funding for people working to understand the results.

The Exponential Time Hypothesis underpins large parts of complexity theory and algorithm design. If it is false, many results built on the assumption that 3-SAT cannot be solved in subexponential time would need rethinking.

But the claim is only as strong as its verification. OpenAI itself notes that the ETH result will be checked by the expert group, and that it released the write-up and a Lean proof on GitHub so others can examine them. The group also published an open letter asking the community for input on how to evaluate the claim and how to set norms for AI-generated breakthroughs, including questions about early access to results, credit for checking and explaining them, and support for students and researchers.

For now, this is a research story, not a product one. Nothing in OpenAI's announcement points to a new app, subscription or feature readers in Tunisia can use today, and there is no announced availability, pricing or regional rollout.

The practical takeaway is about verification. If the ETH result holds up under review, it could reshape how computer scientists think about hard problems — the kind of work that eventually feeds into faster algorithms and more capable software. Until independent researchers confirm it, treat the claims as unverified, even though OpenAI has published the materials for anyone to check.