🔍 Read the full analysis: Could OpenAI’s AI Mathematics Turn 722 Proofs Into Something More? on ThorstenMeyerAI.com
Get hardware and tech essentials 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 published 722 mathematical manuscripts produced by an unnamed, unreleased model, covering 372 families of results. The company says the work includes claims about major open problems, but outside mathematicians have not confirmed the catalogue as a whole; its lasting value will depend on verification and whether researchers can extract useful ideas.
OpenAI published 722 mathematical manuscripts on Monday, presenting work from an unnamed, unreleased model that the company says spans 372 families of results. The collection includes claims involving famous open problems, but OpenAI chief executive Sam Altman said the results have not been confirmed by outside mathematicians, leaving their accuracy and significance unresolved.
OpenAI’s post and GitHub repository describe manuscripts across number theory, geometry, operator algebras, topology, theoretical computer science and mathematical physics. The company says the results were selected from roughly 4,000 problems posed to the model, with an average result taking about three hours of ChatGPT Pro reasoning compute. OpenAI says the manuscripts are published under the Apache-2.0 license.
The catalogue includes claimed results concerning the Unique Games Conjecture, Hilbert’s tenth problem over the rationals, the isomorphism of nonabelian free group factors, the Hodge conjecture for CM abelian varieties, and conjectures in convex geometry. One manuscript claims a zero-free region for the Riemann zeta function to the right of Re(s) = 11/12. These are claims in the released work, not independently established solutions.
OpenAI provides Lean formalizations for many, but not all, results, and its repository warns that some results without formalization could have issues. The release contains ten abridged reasoning summaries for the 372 families. The company selected the problems it considered significant; no outside group made that selection. OpenAI also says the Riemann write-up was edited by humans for readability.
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 Usable Ideas
The importance of the release will depend on more than whether individual statements are true. In mathematics, a proof can matter because it introduces a method that other researchers can apply. If an AI-generated argument is correct but opaque or difficult to reuse, it may settle a question without changing how mathematicians work. If researchers can understand and generalize its techniques, the work could have a wider effect.
The Unique Games Conjecture illustrates the potential stakes. A substantial body of theoretical computer science uses the conjecture to establish limits on approximation algorithms. A verified resolution could prompt reassessment of results built on that assumption. But the manuscript is not a resolution simply because it appears in the catalogue: mathematicians must establish that its proof is sound and that it addresses the conjecture in its accepted form.
The distinction also matters for how AI progress is measured. A large count of generated manuscripts does not establish that the work is correct, useful or understandable. Independent checking, clear exposition and follow-up research will show whether the release contributes knowledge or mainly adds material for experts to evaluate.
As an affiliate, we earn on qualifying purchases.
OpenAI’s Earlier Mathematics Claims
This is OpenAI’s fourth major mathematics release this year, according to the source material. In May, a model produced a counterexample to the Erdős unit-distance conjecture. Five mathematicians — Noga Alon, Thomas Bloom, Tim Gowers, Daniel Litt and Will Sawin — published a human-verified account of the result that same day. That case offers one possible route from machine output to work the mathematical community can assess: researchers translate and check the argument.
OpenAI’s August release, called “Ten Advances,” had a more disputed outcome. A claimed counterexample to Connes’s rigidity conjecture was challenged within a day; the critique said the groups in the construction did not meet a condition required by the conjecture. The September Navier–Stokes announcement, which OpenAI described as a Lean-formalized proof involving about 10,000 concurrent agents over 88 hours, also prompted debate over the role of AI in mathematical research. A group of 25 Fields Medalists signed a declaration criticizing the use of famous problems as benchmarks without human understanding. Their concern was about the purpose and practice of such work, not a declaration that the proof was wrong.
These episodes do not determine whether the new manuscripts hold up. They do show why publication, formalization and independent mathematical review should not be treated as interchangeable forms of confirmation.
formal proof verification software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Independent Checks Still Pending
No outside mathematical review of the full catalogue is reported, and the release does not establish that all 722 manuscripts are correct. Formalization can help check whether a proof follows from stated definitions and assumptions, but many manuscripts are not formalized, and the source material does not specify what independent verification has been completed for each result.
It is also unclear how the 372 families were assessed after the model generated its work, what standards OpenAI used to select the roughly 4,000 problems, or how many manuscripts will lead to follow-up research. The source identifies the Riemann and Hodge manuscripts as exceptions to the standard process, including human editing for readability in the Riemann case, but does not give enough detail to establish how those interventions affected the arguments.
Until experts examine individual manuscripts, it remains possible that some claims are sound, some need correction, and others do not prove the mathematical statements researchers recognize. The collection’s eventual influence cannot be inferred from its size or from the prominence of the problems named in it.
mathematical theorem proving tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Mathematicians Must Test the Claims
The next step is independent review of individual manuscripts, including checking assumptions, proof steps and whether each result matches the problem it claims to solve. For work without Lean formalization, researchers will need to assess the arguments through other methods or develop formal checks where practical. The released summaries may help experts decide which families to examine, but they cover only ten of the 372 families.
OpenAI has not, in the material available here, set out a timetable for external validation or identified which results it expects mathematicians to review first. The clearest measure of progress will be whether researchers publish verified accounts that make the reasoning understandable and reusable. Until then, the manuscripts are a large set of AI-generated mathematical claims awaiting evaluation, not a confirmed list of new theorems.
AI-powered mathematical problem solvers
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What did OpenAI release?
OpenAI published 722 mathematical manuscripts, organized into 372 families and attributed to an unnamed, unreleased model. The work ranges across several fields, including number theory and theoretical computer science.
Have mathematicians verified the results?
Not as a complete collection. OpenAI’s chief executive described the results as claims not yet confirmed by outside mathematicians. The company says many, but not all, results have Lean formalizations, and its repository warns that some unformalized work could have issues.
Does the release prove the Unique Games Conjecture?
The catalogue includes a manuscript claiming a proof of the Unique Games Conjecture. That claim has not been established by the release alone; independent experts must check whether the argument is correct and addresses the conjecture as stated.
Why might a correct AI proof still have limited impact?
A proof can settle a question without providing methods that other researchers can understand or reuse. Its broader value depends partly on whether mathematicians can extract ideas from it and build further results.
What happens next?
Mathematicians will need to examine the manuscripts, verify the arguments and, where useful, produce clearer or formalized versions. OpenAI has not provided a timetable for that review in the source material.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
