🔍 Read the full analysis: What Could Come Of OpenAI’s 722 AI Mathematics Proofs? on ThorstenMeyerAI.com
Get smart everyday buys delivered free — and shop member deals
- Fast, free delivery on millions of items
- Access to Prime Big Deal Days deals on October 6–7
- Prime Video, Amazon Music and more included
TL;DR
OpenAI has published 722 mathematical manuscripts produced by an unnamed, unreleased model, covering 372 families of results. The manuscripts include claims about major open problems, but outside mathematicians have not confirmed them; their longer-term value may depend on whether researchers can verify and build on the work.
OpenAI published 722 mathematical manuscripts on Monday, presenting results from an unnamed model that the company has not released. The papers cover 372 families of related results, including claims about long-standing open problems, but OpenAI chief executive Sam Altman said the claims have not yet been confirmed by outside mathematicians. Whether the work matters beyond its initial results will depend on independent checking and on researchers’ ability to understand and reuse its methods.
The manuscripts span number theory, geometry, topology, operator algebras, theoretical computer science and mathematical physics. OpenAI’s repository says they came from roughly 4,000 problems posed to the model, with the company selecting results it judged to have an appropriate level of significance. The average result, according to the source material, took about three hours of ChatGPT Pro thinking compute. OpenAI has not released the model or named it.
The catalogue includes claimed work on the Unique Games Conjecture, Hilbert’s tenth problem over the rationals, the isomorphism of nonabelian free group factors, a zero-free region for the Riemann zeta function to the right of Re(s) = 11/12, and the Hodge conjecture for CM abelian varieties. These are claims in manuscripts, not results independently established by the release itself. OpenAI’s README warns that some results without formal verification could have issues.
Many, but not all, results have Lean formalizations, which can help check whether a proof follows from its stated assumptions. The repository also contains 10 abridged reasoning summaries, rather than a summary for every family. The source material says the Riemann zero-free-region manuscript was edited by humans for readability, and identifies it and the Hodge result as exceptions to the standard process.
722 proofs, one question: will any of OpenAI’s AI mathematics actually lead anywhere?
An unreleased, unnamed model produced claimed proofs of results that would each define a career. Sam Altman calls them “claims not yet confirmed by outside mathematicians.” The real question isn’t whether it’s impressive. It’s whether answers nobody understands become discoveries anyone can build on.
Same day: Alon, Bloom, Gowers, Litt, Sawin post a digested, human-verified version. The model for success.
Connes rigidity counterexample challenged within a day — constructed groups fail the required condition. Three rival machine “counterexamples” from different labs now circulate.
~10,000 agents, 88 hours, est. ~$22M at retail. Priority dispute; 25 Fields Medalists sign “A Severe Misalignment” — not saying it’s wrong, saying it’s not understood.
Altman now hedges at announcement — a shift from September. Verification has barely started.
Humans extract the technique, write it up, build on it. This is where downstream discovery comes from.
The question is answered; nobody learns anything reusable. Closes a door without opening a field.
The proof breaks, or proves a statement that doesn’t match the conjecture as mathematicians mean it.
The Unique Games Conjecture is the clearest case. Results like the optimality of Goemans–Williamson for Max-Cut are proved assuming UGC. A correct proof converts them all — no understanding required. A zero-free strip for zeta works the same way for prime-distribution results. Free group factors, Kadison, Mahler would redirect whole programmes — but how depends on the method, which means digestion.
Technology. A Navier–Stokes blow-up proof doesn’t change how anyone designs aircraft; engineering turbulence models never depended on the answer. Near-term consequences are mathematical, not industrial. “AI will cure cancer next” skips several steps.
“Verification abundance, adjudication scarcity” — making proof-checking cheap doesn’t reduce the burden of deciding what’s true and what matters. 722 manuscripts land on a review system built for a trickle, filtered by a selection nobody outside OpenAI made.
Humans re-deriving results, like Alon–Gowers et al. in May
Other people’s work building on these manuscripts
How many unformalized results survive expert checking
Do the Lean statements match the real conjectures?
Do any survive peer review?
Some of it, yes — where a literature is waiting (UGC), a correct proof pays off immediately; where a proof carries a new technique humans digest, it can open a field. Most of it, probably not on its own: at 722 manuscripts with 10 reasoning summaries, the Four Colour pattern is the likely default unless mathematicians are funded and given time. And some will be wrong — OpenAI says so itself. It’s an industry pattern, not one company’s: the forced-Euler result came from an Anthropic researcher, and rival machine-generated Connes “counterexamples” circulate from different labs. The proofs arrived this week. The discoveries, if they come, will arrive at the speed of human understanding.
From Machine Proofs to New Mathematics
A proof can settle a question without giving mathematicians a useful new way to solve other problems. The potential value of this release therefore rests on two distinct tests: whether each claim is correct, and whether its reasoning contains ideas that people can understand, adapt and extend. A verified result could affect fields that rely on the relevant conjectures; a proof that offers no reusable insight may have a more limited effect.
The Unique Games Conjecture illustrates the possible stakes. The source material describes a substantial body of theoretical computer science that proves limits on approximation algorithms under assumptions including this conjecture. If the conjecture were settled, researchers would need to examine which conditional results change and what follows from the new proof. That impact cannot be assessed until the manuscript is checked and its exact conclusions are established.
There is also a practical question about how the work is presented. Formalization may make some proofs easier to validate, but it does not by itself show that a result is significant or that its approach will generate new theory. The key outcome may not be the number of manuscripts published, but whether mathematicians can turn any of them into clear, independently checked arguments that support further work.
AI mathematical proof verification software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Earlier Releases Offer Caution
This is OpenAI’s fourth major mathematics release this year, according to the source material. The first provides a model for how machine-generated work can enter the field: in May, OpenAI’s model produced a counterexample to the Erdős unit-distance conjecture. Five mathematicians, including Noga Alon, Tim Gowers and Will Sawin, then published what they called a digested, human-verified version. In that case, researchers translated the machine output into a form they could evaluate and confirmed the result.
An August release, called “Ten Advances,” showed why claims require scrutiny. One announced counterexample to Connes’s rigidity conjecture was challenged within a day; a critique argued that the constructed groups did not meet the conjecture’s required conditions. The source material also reports multiple independent machine-generated counterexamples to the same conjecture. That episode underscores the distinction between a proof that appears to address a famous question and one that actually meets its definitions.
In September, OpenAI announced a Lean-formalized proof concerning finite-time blow-up in the Navier–Stokes equations, another prominent open problem. The announcement prompted a dispute over priority amid concurrent work on forced Euler equations. Days later, 25 Fields Medalists signed a declaration titled “A Severe Misalignment of AI in Mathematics.” According to the source material, their concern was that using famous problems as benchmarks without human understanding could conflict with mathematics’ aims. These episodes are precedents, not a verdict on the new manuscripts.
As an affiliate, we earn on qualifying purchases.
Independent Checks Still Needed
No outside verification of the 722 manuscripts is described in the supplied material. It remains unclear which claims will survive expert scrutiny, how quickly researchers can review them, and whether the formalized proofs cover the central steps of each result. The release’s selection process also took place within OpenAI: the company chose which results from roughly 4,000 problems it considered significant, and the source material says outsiders did not make that selection.
It is also not yet clear how many of the 372 families contain genuinely distinct ideas, how much human editing or intervention shaped individual manuscripts, or which results will be understandable enough to support follow-on research. The source material provides no confirmed downstream discoveries from this release. Even if a result is correct, its influence will depend on what the proof makes possible for other mathematicians.
As an affiliate, we earn on qualifying purchases.
Review Will Set the Next Milestone
The immediate next step is independent mathematical review. Researchers will need to inspect the arguments, test formalizations where available, and compare each proved statement with the exact conjecture or problem it claims to resolve. For results that pass those checks, a further step is to produce explanations that make the reasoning accessible and identify any reusable methods.
The supplied material does not give a review timetable or identify an independent body responsible for assessing the full catalogue. It is also unclear whether OpenAI will publish additional summaries, revisions or responses to critiques. The clearest measure of what comes next will be which manuscripts receive detailed outside verification—and whether that work yields ideas researchers can use beyond the original claims.
As an affiliate, we earn on qualifying purchases.
Key Questions
What did OpenAI publish?
OpenAI published 722 mathematical manuscripts, grouped into 372 families of related results and generated by an unnamed model that has not been released.
Have mathematicians verified the claimed proofs?
The supplied source material says the claims have not yet been confirmed by outside mathematicians. Some results have Lean formalizations, but the repository warns that unformalized results could have issues.
What major problems do the manuscripts claim to address?
Among the claims are work on the Unique Games Conjecture, Hilbert’s tenth problem over the rationals, nonabelian free group factors, a zero-free region for the Riemann zeta function, and the Hodge conjecture for CM abelian varieties. These remain claims pending independent review.
Could the manuscripts lead to new discoveries?
Possibly, but correctness alone would not establish that they lead to further mathematics. Researchers would also need to understand the proofs and identify methods or ideas they can apply to other problems.
What happens next?
Outside mathematicians are expected to examine the manuscripts, check their arguments and clarify which claims hold. The supplied material gives no timetable for that review and does not say which papers will be assessed first.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
