OpenAI claims three new science breakthroughs and recruits expert shepherds to verify them

OpenAI has announced three new results it describes as major discoveries in mathematics, computer science and physics, paired with a new process for putting AI-generated claims in front of outside academics. According to The Verge, the announcement arrives less than a month after the company publicly released hundreds of manuscripts that it said contained solutions to long-standing mathematics problems.

Three results across math, CS and physics

The mathematics result is a claimed proof of local smoothing for the Euclidean wave equation. As The Verge reports, the argument uses an auxiliary partial differential equation to establish Sogge's local smoothing for waves in every spatial dimension of two or more, and it also implies the Kakeya maximal and dimension conjectures plus Stein's spherical restriction estimates across the interior of the conjectured range. On the advice of its advisers, OpenAI released the raw model output directly, as it did in its October 6 release, so that researchers can work through their own explanations of it.

The computer science paper makes a startling claim: that the Exponential Time Hypothesis (ETH), widely believed since it was proposed in 1999, is false. According to The Verge, it presents a deterministic algorithm that exactly counts satisfying assignments of fixed-width CNF formulas in subexponential time and polynomial space, with the 3-SAT case amounting to a refutation of ETH.

The physics result has not been released in full. OpenAI describes it as a mathematically rigorous existence proof for certain interacting quantum field theories containing gauge fields and fermions in four dimensions, including Quantum Chromodynamics and a unified completion of the Standard Model of particle physics.

A new verification model built on named experts

What distinguishes this release is how review is being handled. For the computer science claim, an independent working group consisting of Russell Impagliazzo and Ramamohan Paturi of UC San Diego, who were part of the group that originally proposed ETH in 1999, together with MIT's Ryan Williams, will examine and explain the proof. At the group's request, the write-up and a Lean-formalized proof were published on GitHub immediately so that others can inspect them too.

For the physics work, OpenAI says David Gross will lead a group including Michael Douglas, Arthur Jaffe, Igor Klebanov and Nikita Nekrasov to shepherd the result, in the company's own terminology.

The computer science group also published an open letter asking the community for input on norms, raising questions about expectations for early access to AI-generated results, disclosure of conflicts of interest and use of unpublished material; how credit should recognize both new discoveries and the work of checking, fixing, simplifying, explaining and extending them; and how to support students and researchers while preserving the understanding needed to question results.

OpenAI's own advisers pushed back the day before

The shepherding approach is contentious. The Verge notes that AGMAI, the panel of math advisers OpenAI assembled, stated just one day before this announcement that it recommends AI companies not adopt the practice of designating certain people to receive output first, citing unequal access, borrowed reputation and biased selection. The panel suggested instead that labs release output on scholarly repositories with Lean-formalized proofs, information about how the problems were chosen, and funding for people working to understand the results. OpenAI itself acknowledges that its new arrangement could carry problems.

A Fields Medal backdrop

The timing adds an extra layer. NYU professor Hong Wang received a Fields Medal from the International Mathematical Union a few months ago for resolving the Kakeya conjecture in three dimensions. The new AI output claims results that cover all dimensions, extending well beyond that landmark human proof.

Why it matters

Each claim, if it survives scrutiny, matters on its own. A confirmed refutation of ETH would upend assumptions underlying much of computational complexity theory, where the hypothesis serves as a widely used foundation for hardness arguments. The math result would push past what human proofs have achieved on problems central to harmonic analysis.

But the deeper story is procedural. AI systems are producing claims at a pace that traditional peer review cannot absorb, and the field is improvising verification arrangements in public. This release shows the friction in real time: OpenAI chose named expert shepherds for two of the three results, while its own advisory panel publicly recommended against exactly that model a day earlier, favoring open repository releases and formalized proofs instead. How discovery, verification and credit get divided between machines and mathematicians has become an open research-policy question, not just a technical one.