🔍 Read the full analysis: What’s The Point Of 722 Proofs For OpenAI’s AI Mathematics? on ThorstenMeyerAI.com
TL;DR
OpenAI published 722 mathematical manuscripts, organized into 372 families, from work by an unnamed model. The manuscripts include extraordinary claims, but the company says outside mathematicians have not confirmed them, and some results are not formally verified. Their value will depend on independent checking and whether researchers can understand and build on the methods.
OpenAI published 722 mathematical manuscripts on Monday, grouped into 372 families of related results and generated by a model the company has not named or released. The catalogue includes claims about major open problems, but the results have not been confirmed by outside mathematicians, leaving their accuracy and scientific value unresolved.
OpenAI says the manuscripts came from roughly 4,000 problems posed to the model, with the company selecting results it considered significant. The average result reportedly used about three hours of ChatGPT Pro thinking compute. The papers span areas including number theory, geometry, topology, operator algebras, theoretical computer science and mathematical physics. OpenAI published them under the Apache-2.0 license.
The most striking manuscripts claim results concerning 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. Other claims concern the Hodge conjecture for CM abelian varieties and conjectures in convex geometry. These are claims in papers, not established solutions.
OpenAI’s repository includes Lean formalizations for many, but not all, results. The README cautions that some unformalized work could contain issues. The company also supplied ten abridged reasoning summaries for the 372 families. According to the source material, the Riemann manuscript was edited by humans for readability; the company itself chose which problems and results to feature.
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.
When a Proof Becomes Useful
The release matters because it puts a large set of potentially important mathematical claims into public view, but publication is not the same as verification. Independent mathematicians must establish whether each argument is correct and whether its conclusions match the original problem. Formalization can help check a proof against specified rules, but it does not by itself show that the result has been independently reviewed or that its ideas are useful to researchers.
In mathematics, a result’s longer-term value often comes from methods other researchers can reuse. The source material contrasts OpenAI’s May result on the Erdős unit-distance conjecture, which five mathematicians described in a digested, human-verified form, with a disputed August claim about Connes’s rigidity conjecture. In the latter case, a critique said the constructed groups did not meet the conjecture’s required condition. These episodes show why translation, scrutiny and correction by people in the field are part of assessing AI-generated work.
If a claim such as the one concerning the Unique Games Conjecture holds up, it could affect a substantial body of theoretical computer science that relies on the conjecture when analyzing the limits of approximation algorithms. But no such consequences follow from the catalogue alone. Researchers first need to check the proof, clarify what it establishes, and identify any techniques that can be carried into other problems.
AI research tools for mathematicians
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
OpenAI’s Math Releases So Far
This is described in the source material as OpenAI’s fourth major mathematics release this year. In May, its model produced a counterexample to the Erdős unit-distance conjecture, and five mathematicians published a human-verified account the same day. That episode offered a concrete route from model output to a result researchers could evaluate.
OpenAI’s August collection, called “Ten Advances,” had a more mixed reception. A claimed counterexample to Connes’s rigidity conjecture was challenged within a day, with critics saying the construction did not satisfy a required condition. The source material says similar machine-generated counterexamples from other groups were also circulating.
In September, OpenAI announced a Lean-formalized result concerning finite-time blow-up for the Navier–Stokes equations, produced using about 10,000 concurrent agents over 88 hours. That announcement coincided with a dispute over priority involving separate work on forced Euler equations by Levent Alpöge and Tristan Buckmaster. Three days later, 25 Fields Medalists signed a declaration objecting to AI mathematics focused on famous problems as benchmarks without human understanding. Their criticism, as described in the source material, was about the purpose and practice of the work, not a declaration that the Navier–Stokes proof was false.
formal proof verification software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Which Claims Will Hold Up
Independent assessments of the 722 manuscripts are not established in the supplied reporting. It is unclear which claims will survive expert review, how long checking will take, or whether all papers will receive equal attention. The ten abridged reasoning summaries cover only a portion of the 372 families, and many results lack Lean formalizations.
It is also unknown how much of the work will lead to reusable mathematical ideas. A proof may be correct yet difficult to interpret or unproductive for later research; it may also contain an error or address a version of a conjecture that differs from the question experts intended. The catalogue’s selection was made by OpenAI, so the published set does not independently establish that these are the most significant results among the roughly 4,000 problems posed.
As an affiliate, we earn on qualifying purchases.
Independent Review Comes Next
The immediate next step is for mathematicians to examine the manuscripts, reproduce key arguments and compare each statement with the problem it claims to solve. Where formalizations exist, specialists can inspect what the code encodes; where they do not, the arguments require other forms of expert scrutiny. No independent review timeline or final assessment is specified in the supplied material.
The more consequential test will come after verification: whether researchers can extract explanations and techniques that help with other questions. The May Erdős episode offers one possible model, in which mathematicians turn machine output into an account the field can understand and assess. Until comparable work is done across these papers, the catalogue is best treated as a collection of substantial but unconfirmed mathematical claims.
AI-assisted mathematical research books
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, arranged into 372 families, based on work by an unnamed model. The company says the model was given roughly 4,000 problems.
Have mathematicians verified the claimed results?
Not according to the supplied reporting. OpenAI’s publication includes a warning that some unformalized results could have issues, and the claims have not been confirmed by outside mathematicians.
What is Lean formalization?
Lean is a proof assistant used to encode mathematical arguments in a form that can be checked by software. OpenAI says many, but not all, results have Lean formalizations; that status does not replace broader expert assessment of the claims.
Why does the Unique Games claim matter?
The Unique Games Conjecture is used as an assumption in work on the limits of approximation algorithms. If a claimed proof were correct, it could affect that research, but the manuscript remains unverified.
What happens after publication?
Researchers must check the arguments, confirm what each paper proves and determine whether its methods can be reused. The source material gives no schedule for completing that review.
Source: ThorstenMeyerAI.com
