🔍 Read the full analysis: Could 722 Proofs Be A Starting Point For OpenAI’s AI Mathematics? 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 manuscripts containing claimed mathematical results produced by an unnamed, unreleased model, grouped into 372 families. The claims include solutions to long-standing problems, but external mathematicians have not confirmed them, and the release does not show whether the work will lead to reusable ideas.
OpenAI published 722 mathematical manuscripts on Monday, presenting work by an unnamed, unreleased model across 372 families of results. The collection includes claims about major open problems, but the company and its own repository caution that outside mathematicians have not confirmed the results.
OpenAI says the manuscripts came from roughly 4,000 problems posed to the model, with results selected by the company for what it judged an appropriate level of significance. The average result used about three hours of ChatGPT Pro thinking compute, according to the source material. The manuscripts cover fields including number theory, geometry, topology, operator algebras, theoretical computer science and mathematical physics, and are published under the Apache-2.0 license.
The catalogue includes claims involving the Unique Games Conjecture, Hilbert’s tenth problem over the rationals, free group factors, 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 manuscripts, not independently established solutions. OpenAI’s repository says some results lack formal verification and warns that unformalized results could have issues. Many results have Lean formalizations, but not all; the source says the Unique Games, Riemann-region and free-group-factor manuscripts are among those with formalizations.
Only 10 abridged reasoning summaries were provided for the 372 families. The Riemann-region manuscript was edited by humans for readability, and the Hodge result and Riemann write-up were exceptions to the standard process, according to the source material. The selection and presentation were controlled by OpenAI; the release does not establish that independent mathematicians have checked the full set.
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.
From Machine Proofs to Usable Ideas
The scale and ambition of the release matter, but a mathematical claim is not the same as a mathematical discovery. The field must determine whether the arguments are correct and whether they offer methods that other researchers can understand and reuse. Verification is the immediate test; the longer-term test is whether the work changes what mathematicians can prove.
The source material frames three possible outcomes: researchers may digest a result into a clear argument and build on it; a proof may be correct but yield little reusable insight; or it may fail, or address a statement different from the one mathematicians intended. Those possibilities make the collection a test not just of AI’s ability to produce answers, but of human review and mathematical explanation.
The Unique Games claim could be consequential if verified because many theoretical computer science results rely on the conjecture to establish limits on approximation algorithms. But the manuscript’s existence does not itself settle those implications. Until the proof is checked and its assumptions are understood, researchers cannot treat those downstream results as changed.
As an affiliate, we earn on qualifying purchases.
OpenAI’s Recent Math Results
The release follows three earlier mathematics announcements described in the source material. In May, OpenAI’s model produced a counterexample to the Erdős unit-distance conjecture. Five mathematicians subsequently posted a human-verified, digestible version, offering an example of how machine-generated work can become assessable by the field.
In August, OpenAI announced “Ten Advances.” One 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 Navier–Stokes blow-up proof generated using about 10,000 concurrent agents over 88 hours. That announcement prompted debate over the use of famous problems as AI benchmarks and over whether technically checked answers are enough without human understanding. The current release extends that debate across a much larger set of claims.
formal verification tools for mathematicians
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What Independent Checks Will Show
No outside confirmation of the 722 manuscripts is established in the supplied material. It is unclear how many claims will survive expert review, how long checking each will take, or whether the formalizations cover the key reasoning rather than only parts of an argument. A Lean formalization can support verification, but the release’s caveat about unformalized work means the collection does not have a single uniform verification status.
It is also unclear how OpenAI chose the problems and then narrowed roughly 4,000 attempts to the published set. The selection was made inside the company, and only 10 reasoning summaries were published. The supplied material does not identify the model, provide a complete account of the review process, or report independent assessments for the headline claims. Until those details and checks emerge, the number of manuscripts should not be treated as a count of confirmed discoveries.
As an affiliate, we earn on qualifying purchases.
Mathematicians Must Test the Claims
The next step is independent examination of the manuscripts, with researchers checking whether each proof is valid, whether it proves the stated result and whether its reasoning can be understood and reused. Formalized arguments may make some checks more direct, while unformalized manuscripts may require additional scrutiny. No review timetable is given in the supplied material.
For readers, the most informative developments will be clear assessments from mathematicians and any corrected, formalized or independently reproduced proofs. OpenAI’s release makes the claims available for that process; it does not settle them. The longer-term measure will be whether researchers can extract new techniques or results from the work, rather than simply confirm that a statement has been proved.
As an affiliate, we earn on qualifying purchases.
Key Questions
What did OpenAI publish?
OpenAI published 722 mathematical manuscripts grouped into 372 families, based on roughly 4,000 problems posed to an unnamed model.
Have the claimed results been confirmed?
Not in the supplied material. OpenAI’s stated caveat is that the claims have not yet been confirmed by outside mathematicians, and its repository warns that some unformalized results could have issues.
What is a Lean formalization?
Lean is a proof assistant used to encode mathematical arguments in a form a computer can check. The source says many, but not all, results have Lean formalizations; that does not mean every manuscript has the same verification status.
Why does the Unique Games claim matter?
The conjecture is used in theoretical computer science to establish limits for approximation algorithms. If a proof is correct, it could affect work that depends on the conjecture, but the claim must first be independently checked.
What would make the release valuable beyond solving problems?
Researchers would need to understand the proofs and identify methods they can reuse. A result may be correct yet offer little new insight, while a clear, general technique could support further discoveries.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
