🔍 Read the full analysis: Can OpenAI’s AI Mathematics Move Beyond The Proofs? 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, including claims about major open problems. The claims have not been confirmed by outside mathematicians, and the main question is whether researchers can verify and use the work to develop new ideas.
OpenAI published 722 mathematical manuscripts on Monday, presenting work from an unnamed, unreleased model across fields including number theory, geometry and theoretical computer science. The collection includes claims about several prominent open problems, but outside mathematicians have not confirmed the results, leaving verification and the work’s usefulness to the field unresolved.
The manuscripts are organized into 372 families of related results, selected from roughly 4,000 problems posed to the model. OpenAI says an average result took about three hours of ChatGPT Pro thinking compute. The company filtered the problems for what it considered an appropriate level of significance; that selection was made internally, not by independent mathematicians.
The catalogue ranges from geometry and topology to operator algebras, mathematical physics and theoretical computer science. Among its claims are a proof of the Unique Games Conjecture, a resolution of Hilbert’s tenth problem over the rationals, results concerning free group factors and the Hodge conjecture for CM abelian varieties, and a zero-free region for the Riemann zeta function to the right of Re(s) = 11/12. These are claims in the released manuscripts, not independently established breakthroughs.
OpenAI released the collection under the Apache-2.0 license. Its repository includes Lean formalizations for many, but not all, results. The README cautions that some unformalized results could have issues. The company also supplied only 10 abridged reasoning summaries for the 372 families. The write-up about the Riemann zeta function was edited by humans for readability, according to the source material.
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 Impact
The release’s significance will depend less on the number of manuscripts than on whether mathematicians can check the proofs, understand the methods and reuse them. A valid solution to a famous problem may settle a question; a method that researchers can explain and extend may also change how they work across a field.
The Unique Games claim illustrates the potential stakes if it holds. The conjecture is central to hardness of approximation in theoretical computer science, and many results about the limits of approximation algorithms rely on it. But the manuscript’s claim alone does not change those results: researchers would need to verify that the proof establishes the conjecture as stated and then determine what follows from it.
The release also raises a practical challenge for the research community: hundreds of technical claims require expert attention, while the published reasoning summaries cover only a fraction of the families. Formal verification can help establish that a formalized proof follows specified rules, but it does not by itself show that a result is important, that the formal statement matches the intended conjecture, or that the mathematical ideas are broadly useful.
mathematics problem solving software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Earlier AI Math Claims Offer Caution
This is described in the source material as OpenAI’s fourth major mathematics release this year. In May, the company’s model produced a counterexample to the Erdős unit-distance conjecture. Five mathematicians posted a human-verified, digested version the same day, providing an example of how machine-generated work can become assessable through human review.
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 said the constructed groups did not meet the condition required by the conjecture. The episode shows why a proof can fail to establish the claim researchers care about even when its reasoning appears substantial.
In September, OpenAI announced a Lean-formalized result about finite-time blow-up in the Navier–Stokes equations, produced, the company said, using about 10,000 concurrent agents over 88 hours. That announcement prompted a dispute over research priority and criticism from 25 Fields Medalists, including Terence Tao, Peter Scholze and Maryna Viazovska. Their declaration argued that using famous problems as benchmarks without human understanding conflicts with the purposes of mathematics. The disagreement concerns not only correctness but also how mathematical progress should be judged.
“A Severe Misalignment of AI in Mathematics.”
— The 25 Fields Medalists who signed the September declaration
As an affiliate, we earn on qualifying purchases.
Independent Checks Remain Pending
No independent confirmation is reported for the collection’s headline claims. It is not yet clear which manuscripts will withstand expert scrutiny, whether the formalized versions cover the central claims in each paper, or how many results will prove useful beyond the problems they address.
The selection process also leaves open how representative the published set is. OpenAI chose which problems and results met its significance threshold, and only 10 abbreviated reasoning summaries were provided for 372 families. The source material does not give a timetable for outside review, a list of independent reviewers, or a confirmed assessment of the claims by subject-area experts.
Even if individual proofs are correct, their broader effect remains uncertain. Some could supply methods that other mathematicians adopt; others could settle statements without yielding reusable techniques. The collection may also contain arguments that need correction or that establish a result different from the intended conjecture. Those outcomes can only be distinguished through detailed mathematical review.
AI-based mathematical research tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Mathematicians Must Test the Manuscripts
The immediate next step is independent examination of the papers, including checking formalizations where available and comparing each result with the exact statement of the relevant problem. Researchers will need to determine not only whether a proof is valid, but whether its reasoning can be translated into a form others can inspect and build on.
OpenAI’s prior releases suggest that some manuscripts may be recast into human-readable, verified accounts, while others may attract corrections or disputes. The company has not provided a public schedule for such reviews in the source material. Until that work is done, the 722 papers should be treated as a large set of mathematical claims—not as a verified list of solved problems.
The longer-term measure will be what researchers can do with the results. If they yield new techniques, further theorems or clearer proofs, the collection could contribute to research beyond its initial claims. If they are correct but difficult to interpret, or fail scrutiny, their impact will be narrower. That judgment remains ahead.
formal proof assistants for mathematicians
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 from an unnamed, unreleased model. They are grouped into 372 families and came from roughly 4,000 problems posed to the model.
Have mathematicians verified the results?
The source material reports that outside mathematicians have not confirmed the claims. OpenAI’s repository also warns that some unformalized results could have issues.
Does a Lean formalization prove a result is important?
A formalization can help check that a proof follows the specified rules. It does not, on its own, show that the result is significant, that the formal statement matches the intended conjecture, or that the method will be useful to other researchers.
Why is the Unique Games claim drawing attention?
The Unique Games Conjecture is central to parts of theoretical computer science, including work on the limits of approximation algorithms. If the claimed proof is correct, researchers would still need to verify it and assess its consequences.
When will the claims be confirmed or rejected?
No review timetable is specified in the source material. Independent mathematicians will need to examine the manuscripts; the current status of individual claims remains unsettled.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
