🔍 Read the full analysis: OpenAI’s AI Mathematics After 722 Proofs: A Direction Still To Find on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
OpenAI published 722 mathematical manuscripts produced by an unreleased, unnamed model, covering 372 families of results. The manuscripts include claims about major open problems, but the claims have not been confirmed by outside mathematicians; whether they yield reusable ideas remains unknown.
OpenAI published 722 mathematical manuscripts on Monday, presenting results from an unnamed, unreleased model across 372 families of related work. The collection includes claims involving major open problems, but OpenAI CEO Sam Altman said the results have not been confirmed by outside mathematicians, leaving their validity and research value unsettled.
OpenAI’s post and GitHub repository describe work spanning number theory, geometry, operator algebras, topology, theoretical computer science and mathematical physics. The company says the manuscripts were drawn from roughly 4,000 problems submitted to the model and filtered for what it considered an appropriate level of significance. That selection was made by OpenAI, not by an external mathematical panel.
The source account says the average result used about three hours of ChatGPT Pro thinking compute. Many, but not all, results have Lean formalizations, a computer-checkable representation of mathematical proof. OpenAI’s repository warns that some unformalized results could have issues. The release includes only ten abridged reasoning summaries for the 372 families, and the Riemann zero-free-region manuscript was edited by humans for readability.
Among the collection’s claims are a proof of the Unique Games Conjecture, a resolution of Hilbert’s tenth problem over the rationals, and results concerning nonabelian free group factors, the Hodge conjecture for CM abelian varieties and Mahler’s conjectures. One manuscript claims a zero-free region for the Riemann zeta function to the right of Re(s) = 11/12. These are claims reported in the release, not results independently established by the material provided here.
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.
Verification Is Only the First Test
The immediate question is whether the proofs hold up. But mathematical value does not end with a correct answer: researchers often judge a proof by whether its methods can be understood, reused and applied elsewhere. A result that settles a problem without yielding a transferable technique may have a different effect on a field from one that opens new lines of work.
The distinction matters because the collection includes claims tied to questions that underpin large bodies of research. In theoretical computer science, for example, many results are proved on the assumption that the Unique Games Conjecture is true. If a proof of that conjecture were validated, researchers would need to examine what it changes for those results. The current release does not establish that consequence; it creates a substantial body of work for experts to scrutinize.
Scale also raises a practical issue. Reviewing 722 manuscripts requires time and subject expertise, while only a small portion of the reasoning summaries is available in abridged form. The publication count alone cannot show how many results are correct, useful or new in a way mathematicians can build on.
mathematical proof assistant software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Earlier Releases Offer Caution
This is described in the source material as OpenAI’s fourth major mathematics release of the year. In May, the company’s model produced a counterexample to the Erdős unit-distance conjecture. Five mathematicians later posted a human-verified, digested account, illustrating how machine output can be turned into work the field can evaluate.
An August release, called “Ten Advances,” had a more contested result: a claimed counterexample to Connes’s rigidity conjecture was challenged within a day. The critique, according to the source, argued that the constructed groups did not meet a condition required by the conjecture. That episode underscores why a claim’s description is not a substitute for independent checking.
In September, OpenAI announced a Lean-formalized result on finite-time blow-up for the Navier–Stokes equations, produced with about 10,000 concurrent agents over 88 hours, according to the source account. The announcement also prompted debate about research priorities and whether using famous problems as benchmarks serves mathematical understanding. These earlier episodes provide context, but do not determine whether any manuscript in the new collection is correct.
The Proofs Await Outside Review
Independent verification is not established for the collection’s headline claims in the supplied source material. It is also unclear how many manuscripts have full Lean formalizations, how reviewers will prioritize the 372 families, and whether the human-edited Riemann manuscript differs in substance from the model’s original output.
The release does not settle whether the claimed proofs are correct, whether they address precisely the statements mathematicians regard as open, or whether their techniques can be reused. Those questions require scrutiny of the individual arguments. It is also unknown which results, if any, will be developed into human-readable work that other researchers can evaluate and extend.
Mathematicians Must Test the Claims
The next step is independent examination of the manuscripts, including checks of formalized proofs and detailed review of arguments that lack formal verification. Researchers will need to compare each result with the precise problem it claims to solve and identify any assumptions or gaps. The source material gives no timetable for that review.
OpenAI’s release provides a catalogue, not a final verdict on its contents. The clearest measure of its longer-term impact will be whether mathematicians verify the claims and can extract methods or results that support further work. Until then, the collection’s size and ambition should not be mistaken for confirmation.
Key Questions
What did OpenAI publish?
It published 722 mathematical manuscripts, grouped into 372 families and produced by an unnamed, unreleased model, according to the company’s post and repository.
Have outside mathematicians verified the results?
The supplied source material says the claims have not yet been confirmed by outside mathematicians. OpenAI’s repository also cautions that some unformalized results could have issues.
What major problems do the manuscripts claim to address?
The reported claims include the Unique Games Conjecture, Hilbert’s tenth problem over the rationals, the Hodge conjecture for CM abelian varieties and a zero-free region for the Riemann zeta function. Their correctness remains unconfirmed in the source material.
Why does it matter whether the proofs are understandable?
A proof can settle a question, but methods that researchers understand and reuse can also support further discoveries. The release has not yet shown which, if any, of its results will have that wider effect.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
