🔍 Read the full analysis: Can OpenAI’s AI Mathematics Turn 722 Proofs Into Progress? on ThorstenMeyerAI.com
Get monitors, keyboards and dev gear 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 in 372 related families, generated from roughly 4,000 problems by an unnamed, unreleased model. The catalogue includes claims about major open problems, but OpenAI says outside mathematicians have not confirmed them, and its repository warns that some unformalized results may contain errors. Their lasting value will depend on independent checking and whether mathematicians can extract ideas others can use.
OpenAI published 722 mathematical manuscripts on Monday, reporting that they were produced by an unnamed, unreleased model from roughly 4,000 problems. The collection includes purported solutions or advances on several longstanding problems, but the claims have not been confirmed by outside mathematicians, and OpenAI’s repository cautions that some results may have issues.
The manuscripts are arranged in 372 families of related results, covering fields including number theory, geometry, topology, operator algebras, theoretical computer science and mathematical physics. OpenAI says the average result required about three hours of ChatGPT Pro thinking compute. The files are published under an Apache-2.0 license; many, but not all, results have Lean formalizations, a kind of machine-checkable proof representation.
Among the headline claims are a proof of the Unique Games Conjecture, a resolution of Hilbert’s tenth problem over the rationals, a result concerning whether all nonabelian free group factors are isomorphic, and a zero-free region for the Riemann zeta function to the right of Re(s) = 11/12. The collection also includes claims about the Hodge conjecture for CM abelian varieties and the Mahler conjectures. These are claims described in the released material, not findings established by independent review.
OpenAI says it selected the problems from a pool of about 4,000 prompts for what it judged an appropriate level of significance. That selection was made inside the company. The release includes only 10 abridged reasoning summaries for the 372 result families. The Riemann zero-free-region write-up was edited by humans for readability, and OpenAI identifies the Riemann and Hodge work as exceptions to its standard 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.
From Machine Proofs to Usable Ideas
The central test is not simply whether a result is correct. In mathematics, a proof can matter because its methods give researchers tools for other problems. If mathematicians can understand and adapt methods in these manuscripts, the work could support new lines of research. If a claim is true but its proof offers no reusable insight, it may settle a question without materially changing the field.
The Unique Games claim illustrates the potential reach. The conjecture underpins a substantial body of theoretical computer science, including results about the limits of approximation algorithms. A verified proof could prompt researchers to revisit conclusions that rely on it. But that consequence depends on the exact statement being proved, the proof’s correctness and whether specialists can analyze the argument—not on the headline claim alone.
The collection also sharpens a dispute about what counts as progress. A result may be formally checked yet remain difficult for people to interpret. Mathematicians will need to distinguish correctness, significance and understanding: related but separate judgments. Until that work is done, the size of the release says little by itself about its lasting contribution.
mathematics proof verification software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Earlier Releases Set a Mixed Pattern
This is OpenAI’s fourth major mathematics release of the year, following projects that have drawn different responses. In May, the company’s model produced a counterexample to the Erdős unit-distance conjecture. Five mathematicians—Noga Alon, Thomas Bloom, Tim Gowers, Daniel Litt and Will Sawin—then published a human-verified, digestible account. That episode offers one possible route from machine output to work the field can evaluate.
OpenAI’s August “Ten Advances” release had a more contested example: its claimed counterexample to Connes’s rigidity conjecture was challenged within a day. The critique argued that the constructed groups did not meet the condition required by the conjecture. The episode shows why close examination of the exact definitions and proof steps matters, especially when a result is presented as overturning established mathematics.
In September, OpenAI announced a Lean-formalized result concerning finite-time blow-up for the Navier–Stokes equations, produced with about 10,000 concurrent agents over 88 hours. That announcement coincided with a priority dispute over related work and criticism from 25 Fields Medalists, including Terence Tao, Peter Scholze and Maryna Viazovska. Their declaration objected to using famous problems as benchmarks without human understanding; it did not, according to the supplied account, establish that the proof was wrong. The distinction between checking a result and understanding its mathematical value remains central to the current release.
“A Severe Misalignment of AI in Mathematics.”
— The 25 Fields Medalists who signed the September declaration
As an affiliate, we earn on qualifying purchases.
Verification and Selection Remain Open
No independent review of all 722 manuscripts is reported in the supplied material. It is not yet clear which claims will survive expert scrutiny, how many are fully formalized, or how much the formalizations cover. A Lean formalization can help check a proof encoded in the system, but the release’s warning about unformalized work makes clear that the catalogue does not provide one uniform level of verification.
It is also unclear how the 10 abridged summaries relate to the full arguments across 372 families, and whether researchers outside OpenAI can readily reproduce the work. OpenAI’s selection of problems is another limitation: the company judged which prompts had an appropriate level of significance, so the published set is not a neutral sample of all attempted work. The specific methods, research time and later corrections for each result will need to be assessed individually.
AI-assisted mathematical research books
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Independent Checks Will Determine Impact
The next step is for specialists to examine individual manuscripts, test their claims and, where possible, check or extend their formalizations. Researchers will also need to translate dense machine-produced arguments into explanations that expose the underlying ideas. The earlier Erdős episode suggests that a human-digested account can help make a result accessible, but it does not guarantee that every manuscript will have the same outcome.
For now, readers should treat the most ambitious entries as unverified mathematical claims, not settled breakthroughs. The important milestones will be independent confirmation, clear accounts of any proof gaps or corrections, and evidence that the methods generate further work. Whether this release becomes a source of mathematical progress will depend on what researchers can establish and learn from it.
machine learning 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 grouped into 372 families, attributed to an unnamed model that the company has not released. The work was drawn from roughly 4,000 problems, according to the company’s release.
Have outside mathematicians verified the results?
Not as a collection, based on the supplied reporting. OpenAI’s position is that the claims have not yet been confirmed by outside mathematicians, and its repository says some unformalized results could have issues.
Does a Lean formalization prove a result is correct?
A Lean formalization can allow a proof encoded in Lean to be checked by the system. It does not mean every manuscript in the release is formalized, nor does it by itself establish that a claimed result is the one mathematicians intended to prove. The extent of formalization varies across the collection.
Why does the Unique Games claim matter?
The Unique Games Conjecture is used in theoretical computer science to establish results about approximation algorithms. If a proof of the conjecture is correct, researchers could revisit work that depends on it. That implication remains conditional on independent verification.
What would show that the release leads to progress?
Evidence would include independent confirmation, accessible explanations of the arguments and methods that other researchers can reuse. A correct proof may settle a problem; whether it produces broader mathematical insight is a separate question.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
