🔍 Read the full analysis: Can 722 Proofs Help Chart A Course For OpenAI’s AI Mathematics? on ThorstenMeyerAI.com
Get the latest gadgets 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, grouped into 372 families, generated by a model it has not named or released. The collection includes claims about long-standing problems, but the company says they have not been confirmed by outside mathematicians; the key test is whether experts can verify and build on the work.
OpenAI published 722 mathematical manuscripts on Monday, grouping them into 372 families of related results generated by a model the company has not named or released. The papers include claims about several major open problems, but OpenAI says outside mathematicians have not yet confirmed them, leaving verification and the usefulness of the proofs as the central questions.
According to OpenAI’s post and its GitHub repository, the manuscripts cover number theory, geometry, operator algebras, topology, theoretical computer science and mathematical physics. The work was selected from about 4,000 problems posed to the model. OpenAI filtered that pool for what it considered an “appropriate level of significance”; no outside group made that selection. The company says the average result used about three hours of ChatGPT Pro thinking compute.
The collection makes claims on topics including the Unique Games Conjecture, Hilbert’s tenth problem over the rationals, the isomorphism of nonabelian free group factors, and a zero-free region for the Riemann zeta function to the right of Re(s) = 11/12. It also includes claims concerning the Hodge conjecture for CM abelian varieties and the Mahler conjectures in convex geometry. These are claims in manuscripts, not independently established resolutions.
OpenAI says many, but not all, results have Lean formalizations, which can help check mathematical arguments in a proof assistant. The repository warns that “some of the unformalized results could have issues.” The company supplied abridged reasoning summaries for just 10 of the 372 families. It also says the Riemann write-up was edited by humans for readability; the Riemann and Hodge results were exceptions to the usual process described for the collection.
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 Shape the Value
The scale and ambition of the release make independent checking a substantial undertaking. But the number of manuscripts does not establish how many results are correct, significant or usable. Each claim must be assessed on its own, and formal verification can check whether a formalized argument follows within a system without, by itself, showing that the result is important or that its ideas will help mathematicians solve other problems.
The distinction matters because a proof can settle a question without producing a method others can use. The mathematical value of landmark proofs often lies in the techniques they introduce and the work they make possible. The Four Colour Theorem, proved with computer assistance, is often cited as a case where a result was established through extensive checking but did not open a comparable body of new theory. By contrast, a machine-generated proof that mathematicians can understand and adapt could become a starting point for further discoveries.
For readers outside mathematics, the practical question is not simply whether AI can produce long proofs. It is whether experts can verify them, identify new reasoning inside them, and use that reasoning elsewhere. Those outcomes are not yet known for this release.
mathematical proof assistant software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
OpenAI’s Earlier Math Releases
This is OpenAI’s fourth major mathematics release this year, according to the supplied account, and earlier work illustrates why review matters. In May, the company’s model produced a counterexample to the Erdős unit-distance conjecture. Five mathematicians, including Noga Alon, Thomas Bloom, Tim Gowers, Daniel Litt and Will Sawin, then published a human-verified, digestible version. That episode offers one possible path from machine output to accepted mathematical work: experts first translate and examine the result.
An August release called “Ten Advances” had mixed results. A claimed counterexample to Connes’s rigidity conjecture was challenged within a day: a critique said the constructed groups did not meet the condition required by the conjecture. In September, OpenAI announced a Lean-formalized result concerning finite-time blow-up in the Navier–Stokes equations, a Millennium Prize problem. The release also prompted a dispute over priority in related work and a declaration signed by 25 Fields Medalists, including Terence Tao, Peter Scholze and Maryna Viazovska. Their stated concern was that optimizing for famous benchmark problems without human understanding could conflict with the aims of mathematics.
That record does not settle the status of the new manuscripts. It does show why claims, formal checks and expert assessment should not be treated as interchangeable. A result may be correct but hard to interpret; it may also contain a flaw or address a slightly different statement from the one mathematicians intended.
““Digested, human-verified.””
— Five mathematicians’ description of their May review
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 independent confirmation of the new claims is provided in the source material. It is not clear which of the 372 result families will survive outside review, how long that review will take, or whether the claimed results will be accepted by specialists. OpenAI has released only 10 abridged reasoning summaries, limiting what readers can quickly assess across the full catalogue.
It is also unclear how many manuscripts contain ideas that mathematicians can reuse, even if their conclusions prove correct. The model’s identity, training and technical details have not been disclosed in the supplied material, and the company’s significance filter was internal. The release therefore does not establish that the selected work represents a comprehensive or independently ranked survey of mathematical progress.
AI-based mathematical research tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Expert Review and Follow-Up Work
The next step is for mathematicians to inspect individual manuscripts, test their arguments and compare each statement with the exact problem it claims to resolve. Where a proof has a Lean formalization, specialists can examine the formal proof; for other work, the repository’s warning makes further scrutiny especially important. Independent verification, not the size of the release, will determine which claims stand.
For results that hold up, the further test will be whether researchers can explain the methods and use them in subsequent work. OpenAI has not, in the supplied material, set a timetable for outside review or announced a process for reporting corrections. The status of the collection is therefore developing, with no confirmed schedule for a broader assessment.
As an affiliate, we earn on qualifying purchases.
Key Questions
What did OpenAI publish?
OpenAI published 722 mathematical manuscripts, organized into 372 families of related results and generated by an unnamed model. The work was selected from about 4,000 problems posed to the system.
Have mathematicians confirmed the results?
Not according to the supplied source material. OpenAI described the manuscripts as claims not yet confirmed by outside mathematicians, and the repository warns that some unformalized results could have issues.
What are some of the major problems covered?
The manuscripts include claims concerning the Unique Games Conjecture, Hilbert’s tenth problem over the rationals, nonabelian free group factors, and a zero-free region for the Riemann zeta function. Their inclusion does not mean the claimed results have been accepted.
Does a Lean formalization prove a result is important?
A Lean formalization can help check that a proof follows within a proof-assistant system. It does not by itself show that a result is significant, conceptually useful or likely to lead to further discoveries.
What would count as a meaningful outcome?
Beyond a correct proof, a meaningful outcome would be mathematicians understanding its reasoning and using its techniques in other work. Whether any of the 722 manuscripts will do that remains unknown.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
