🔍 Read the full analysis: OpenAI’s AI Mathematics And The Possibilities After 722 Proofs on ThorstenMeyerAI.com
Get business pricing on monitors, keyboards and dev gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
OpenAI says an unreleased, unnamed model produced 722 mathematical manuscripts across 372 families, selected from about 4,000 problems. The catalogue includes claims involving major open problems, but the work has not been independently confirmed as a whole, and the company warns some unformalized results may have issues. Its impact will depend on verification and whether mathematicians can extract ideas others can use.
OpenAI published 722 mathematical manuscripts on Monday, saying they were generated by an unreleased, unnamed model after it was given roughly 4,000 problems. The manuscripts, organized into 372 families, include claims about prominent open problems, but OpenAI chief executive Sam Altman said the claims have not been confirmed by outside mathematicians.
The collection spans number theory, geometry, operator algebras, topology, theoretical computer science and mathematical physics. OpenAI says the average result used about three hours of ChatGPT Pro thinking compute. The company selected the problems and filtered the results for what it considered an appropriate level of significance; no outside group made that selection.
Among the manuscripts are claims of 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 Riemann zeta function and the Hodge conjecture for CM abelian varieties. These are descriptions of what the manuscripts claim, not independently established resolutions of those problems.
OpenAI says many, but not all, results have Lean formalizations. Its repository README cautions that some unformalized results could have issues. The release also includes ten abridged reasoning summaries, rather than summaries of all 372 families. The Riemann-related write-up was edited by humans for readability, according to the source account.
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 Will Set the Value
The scale and ambition of the release matter, but publication alone does not establish that its results are correct. If mathematicians verify some of the claims, the consequences could extend beyond answering individual questions. A proof of the Unique Games Conjecture, for example, could affect theoretical computer science, where many results about the limits of approximation algorithms are established under assumptions involving that conjecture.
For mathematics, however, a correct proof is not always the same as a useful new method. The longer-term value often comes when researchers can understand a proof, identify its underlying ideas and apply them elsewhere. The release therefore raises two distinct tests: does each proof hold up? And, if it does, can people learn from it and build on it?
Some machine-produced proofs may settle a question without yielding techniques that travel to other problems; others may contain ideas that become productive after mathematicians recast them. That difference cannot be assessed from the number of manuscripts alone. The work of checking and explaining the results will shape whether the catalogue becomes a source of new mathematical tools or mainly a set of answers to verify.
As an affiliate, we earn on qualifying purchases.
Earlier Releases Offer Cautions
This is OpenAI’s fourth major mathematics release this year, according to the source material, following work on the Erdős unit-distance conjecture, a set of ten claimed advances, and a Navier–Stokes result. The earlier examples show why independent review and human interpretation matter.
In May, OpenAI’s model produced a counterexample to the Erdős unit-distance conjecture. Five mathematicians later posted what they described as a digested, human-verified version, translating the machine output into a form they could evaluate. That process offers one possible path from generated work to a result the mathematical community can assess.
The August release was more contested: a claimed counterexample to Connes’s rigidity conjecture faced a critique that said the constructed groups did not meet the conjecture’s required condition. In September, OpenAI announced a Lean-formalized Navier–Stokes proof produced by about 10,000 concurrent agents over 88 hours. That announcement prompted debate about benchmarks and mathematical understanding. A group of Fields Medalists criticized the practice of targeting famous problems without human understanding; that criticism was not itself a finding that the proof was false.
formal verification tools for mathematicians
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Which Manuscripts Will Hold Up
No outside verification of the full catalogue is reported in the supplied material. It is not yet clear which of the 722 manuscripts will be checked first, how many will survive scrutiny, or whether any of the headline claims will be accepted by specialists. The presence of Lean formalizations may help with checking some results, but formalization is not reported for every manuscript.
The selection process also leaves unanswered how representative the release is: OpenAI chose the problems and filtered the output for significance, while only ten families received abridged reasoning summaries. The materials provided do not explain the criteria in enough detail to assess rejected attempts or compare the published results with the full set of roughly 4,000 problems. Nor do they establish whether the model’s work will lead to reusable methods.
AI research notebooks for mathematics
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Independent Review Comes Next
The immediate next step is for mathematicians to examine individual manuscripts, check the arguments and, where useful, convert machine-generated reasoning into forms specialists can follow. Independent verification will determine whether any of the claims amount to accepted proofs; further work may also be needed to identify the ideas behind a valid result.
OpenAI’s release does not, on the information provided, set a timetable for that review or name an external body responsible for it. Readers should treat the manuscripts as claims under examination rather than a list of settled discoveries. The most meaningful measure of what follows will be which results are verified and whether researchers can use their methods in work beyond the original problems.
mathematical problem solving software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What did OpenAI publish?
OpenAI published 722 mathematical manuscripts grouped into 372 families. The company says an unreleased, unnamed model generated them from a pool of roughly 4,000 problems.
Have the claimed proofs been independently confirmed?
Not as a collection, according to the supplied material. Altman described the results as claims not yet confirmed by outside mathematicians, and OpenAI’s repository warns that some unformalized results could have issues.
What major problems do the manuscripts address?
The catalogue includes claims involving the Unique Games Conjecture, Hilbert’s tenth problem over the rationals, nonabelian free group factors, the Riemann zeta function and the Hodge conjecture for CM abelian varieties. Their inclusion does not establish that the claims are correct.
Why does human review matter if some results are formalized?
Formalization can support checking, but the source says not all results have Lean formalizations. Reviewers also need to establish whether a proof addresses the intended problem and whether its reasoning offers ideas other researchers can understand and reuse.
Source: ThorstenMeyerAI.com
Evergreen bestsellers Picks
bestsellers
As an affiliate, we earn on qualifying purchases.
