🔍 Read the full analysis: 722 Proofs, No Clear Destination? OpenAI’s AI Mathematics In Focus on ThorstenMeyerAI.com
Get office and shipping supplies 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, covering 372 families of results drawn from about 4,000 problems. The claims include solutions to major open problems, but independent mathematicians have not confirmed them; what the work may contribute beyond individual answers remains unsettled.
OpenAI published 722 mathematical manuscripts on Monday, presenting work attributed to an unnamed, unreleased model and spanning 372 families of related results. The collection includes claims about several prominent open problems, but OpenAI chief executive Sam Altman said they have not been confirmed by outside mathematicians, leaving the central question of their correctness unresolved.
OpenAI’s post and repository describe work across number theory, geometry, operator algebras, topology, theoretical computer science and mathematical physics. The results were selected from roughly 4,000 problems posed to the model. OpenAI said it filtered those results for what it considered an appropriate level of significance; that selection was made within the company, not by an external panel.
The source report says an average result used about three hours of ChatGPT Pro thinking compute. Many results have Lean formalizations, computer-checkable versions of proofs, but not all do. OpenAI’s repository warns that some unformalized results could have issues. The collection includes 10 abridged reasoning summaries, far fewer than the 372 result families.
Among the manuscripts are claims concerning the Unique Games Conjecture, Hilbert’s tenth problem over the rationals, isomorphism of nonabelian free group factors, a zero-free region for the Riemann zeta function to the right of Re(s) = 11/12, the Hodge conjecture for CM abelian varieties, and conjectures in convex geometry. These are claims described in the collection, not independently established breakthroughs. The Riemann manuscript was edited by people for readability, according to the source report; it and the Hodge result were exceptions to the usual process.
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.
Why Verification Matters Beyond the Claims
If even some of the manuscripts withstand scrutiny, they could affect fields that rely on the problems they address. The Unique Games Conjecture, for example, underpins many results in theoretical computer science about the limits of approximation algorithms. A verified proof could prompt researchers to revisit conclusions built on that assumption. The consequences would depend on the exact result and proof, and cannot be inferred from OpenAI’s announcement alone.
Correctness is only part of the issue. Mathematicians often value a proof for the methods it makes available, not just for settling a statement. A result that can be checked but whose reasoning yields no reusable technique may answer a question without changing how researchers work. Whether these manuscripts produce new tools, are simply verified answers, or fail review will shape their importance to mathematics.
The release also tests how the field can evaluate a large volume of machine-generated work. With 722 manuscripts and limited summaries, outside researchers face a substantial review task. Formal verification can help establish that a formalized proof follows specified rules, but it does not by itself establish that the result addresses the intended conjecture or that the argument is mathematically illuminating.
mathematics problem solving software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Earlier Releases Offer Mixed Evidence
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, according to the report. Five mathematicians then published what they called a digested, human-verified account. That episode offers one possible route for machine-generated work to become useful: researchers translate the output into a form they can examine and confirm.
OpenAI’s August release, called “Ten Advances,” had a more contested result. A claimed counterexample to Connes’s rigidity conjecture was challenged within a day, with critics arguing that the constructed groups did not meet the conjecture’s required condition. The report also describes OpenAI’s September announcement of a Lean-formalized Navier–Stokes result, generated using about 10,000 concurrent agents over 88 hours. These prior episodes do not determine whether the new manuscripts are correct, but they show why outside checking matters.
The debate is not limited to verification. After the September announcement, 25 Fields Medalists signed a declaration titled “A Severe Misalignment of AI in Mathematics,” according to the source report. Their stated concern was that treating famous problems as benchmarks, without human understanding, could conflict with mathematics’ aims. The disagreement concerns the purpose and value of the work as well as its correctness.
As an affiliate, we earn on qualifying purchases.
Which Manuscripts Will Survive Review
No independent confirmation is reported for the collection’s headline claims. The available material does not identify external reviewers who have checked the manuscripts, provide a timetable for peer review, or say which results mathematicians have already examined in detail. It is also unclear how many of the 372 families have formal Lean proofs and how much those formalizations cover.
The selection process is another open issue. OpenAI says it chose results for an appropriate level of significance from about 4,000 problems, but the material does not give an external standard for that judgment or explain how unsuccessful attempts were assessed. The 10 abridged summaries provide only a small window into the reasoning behind the larger collection.
Even if a claim is correct, its longer-term mathematical value cannot yet be judged. Researchers will need to determine whether proofs are sound, whether statements match the problems mathematicians intended to solve, and whether the reasoning offers reusable ideas. Those outcomes remain unknown; the number of manuscripts alone does not answer them.
AI-based mathematical research tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Independent Checks Will Set the Record
The next step is independent mathematical review of individual manuscripts. Researchers will need to check the statements and proofs, scrutinize any formalizations, and establish whether a result solves the problem as posed. For the unformalized work, OpenAI’s own warning makes careful examination particularly relevant.
As that review proceeds, clearer summaries and human-readable accounts could help specialists assess the results. The earlier Erdős episode, as reported, involved mathematicians producing a digested and verified version of machine output; whether a similar process develops for any of these 372 families is not yet known. There is no review schedule or confirmed next milestone in the source material.
For now, the release is evidence that OpenAI says its model produced a large body of mathematical work, not proof that the headline claims are correct or that the collection will reshape research. What happens next depends on what outside mathematicians can verify and learn from it.
As an affiliate, we earn on qualifying purchases.
Key Questions
What did OpenAI release?
OpenAI published 722 mathematical manuscripts grouped into 372 families, attributed to an unnamed model that has not been released, according to the source material.
Have mathematicians verified the claimed results?
The source material reports no outside confirmation of the collection’s claims. Altman described them as claims not yet confirmed by outside mathematicians.
Do all the manuscripts have formal proofs?
No. The report says many, but not all, have Lean formalizations. OpenAI’s repository warns that some unformalized results could have issues.
Why does the Unique Games Conjecture claim matter?
Many theoretical computer science results about the limits of approximation algorithms rely on the conjecture. If a proof is verified, researchers could revisit work that depends on it; the consequences remain conditional on review.
What will determine whether the release is important?
Independent checks must establish whether the proofs are sound and address the intended problems. Researchers will also assess whether the arguments provide reusable methods, rather than only answers to specific questions.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
