🔍 Read the full analysis: Are 722 Proofs A Sign Of Progress For OpenAI’s AI Mathematics? 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 has published 722 mathematical manuscripts produced by an unnamed, unreleased model, covering 372 families of results selected from about 4,000 problems. The manuscripts include claims about major open problems, but the company says outside mathematicians have not confirmed them, and some results lack formal verification. Their lasting value will depend on whether mathematicians can check, understand and build on the work.
OpenAI published 722 mathematical manuscripts on Monday, presenting work by an unnamed, unreleased model across 372 families of related results. The collection includes claims involving famous open problems, but OpenAI chief executive Sam Altman said the results have not been confirmed by outside mathematicians, leaving their correctness and broader significance unsettled.
OpenAI’s post and project repository describe manuscripts spanning number theory, geometry, topology, operator algebras, theoretical computer science and mathematical physics. The company says the collection was selected from roughly 4,000 problems it posed to the model. The average result used about three hours of ChatGPT Pro reasoning compute, according to the source material. OpenAI grouped the selected work into 372 families and released it under the Apache-2.0 license.
The catalogue includes claims about the Unique Games Conjecture, Hilbert’s tenth problem over the rationals, nonabelian 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 descriptions of what the manuscripts claim, not independently established mathematical results. OpenAI’s repository says some results have Lean formalizations, but not all; its README cautions that some unformalized results could have issues.
Only 10 abridged reasoning summaries accompany the 372 families, and OpenAI chose which problems to include after assessing their significance. The source says two manuscripts received nonstandard treatment: the Riemann-related write-up was edited by humans for readability, and the Hodge result was also an exception to the usual process. The release does not, by itself, show that independent experts have checked the proofs or that the model’s reasoning can be readily followed by other researchers.
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 Proofs Become Useful
The central question is not only whether a proof reaches a correct conclusion. In mathematics, a result’s wider value often depends on whether other researchers can understand its methods and reuse them. A verified proof that introduces a new technique may reshape a field; a correct result that offers no accessible insight may settle a question without generating further work.
The Unique Games Conjecture illustrates why verification could matter beyond pure mathematics. A substantial body of theoretical computer science uses the conjecture as an assumption when establishing limits on approximation algorithms. If the manuscript’s claim survives expert review, it could affect how researchers assess those results. But until specialists confirm exactly what has been proved and examine the argument, the potential consequences remain conditional.
There is also a practical distinction between a machine-produced proof and a mathematical discovery that a community can use. The collection’s size is an output count, not evidence on its own of correctness, novelty or downstream impact. Researchers must check whether each argument works, whether it proves the statement under discussion, and whether its ideas can be extracted and applied elsewhere.
As an affiliate, we earn on qualifying purchases.
Earlier Releases Set Expectations
The Monday collection follows three earlier OpenAI mathematics releases described in the source material. In May, the company’s model produced a counterexample to the Erdős unit-distance conjecture. Five mathematicians later posted what they called a digested, human-verified version, offering a concrete example of how machine-generated work can become assessable and useful through expert review.
OpenAI’s August release, “Ten Advances,” had a more disputed reception. A claimed counterexample to Connes’s rigidity conjecture was challenged within a day; the critique said the constructed groups did not meet a condition required by the conjecture. That episode shows why the distinction between a claim and a checked result matters, particularly when a proof depends on technical definitions or conditions.
In September, OpenAI announced a Lean-formalized proof concerning finite-time blow-up in the Navier–Stokes equations, a Millennium Prize problem. The source material also describes disagreement about research priority and a declaration signed by 25 Fields Medalists criticizing the use of famous problems as AI benchmarks when the work does not support human understanding. The concerns are not identical to a finding that a proof is wrong: they raise questions about how mathematical progress should be judged.
As an affiliate, we earn on qualifying purchases.
Proof Checks Still Pending
The main unresolved issue is whether the manuscripts’ claims withstand independent mathematical review. The release does not establish how many of the 722 papers have been checked by specialists, whether each stated result follows from its proof, or whether the claimed results are genuinely new. Formalization can help verify a proof within a defined system, but the source says formal versions exist for many results, not all.
It is also unclear how OpenAI assessed the roughly 4,000 problems before selecting the published work. The company controlled the selection funnel, while only 10 reasoning summaries were provided for 372 result families. The source material does not identify the model, give a full account of the selection criteria, or describe a schedule for external review. No conclusion about the collection’s overall success or failure can yet be drawn from the number of manuscripts alone.
As an affiliate, we earn on qualifying purchases.
Independent Review Will Decide
The next meaningful development will be expert scrutiny of individual manuscripts: checking the arguments, confirming that they address the stated problems, and determining whether their methods can be understood and reused. Results that survive that process may be corrected, rewritten or developed into work other mathematicians can build on. Others may be narrowed, challenged or rejected.
For now, the release is best treated as a large set of research claims rather than a confirmed list of solved problems. The collection’s significance will depend on what independent mathematicians can verify and learn from it, not simply on how many manuscripts it contains.
As an affiliate, we earn on qualifying purchases.
Key Questions
What did OpenAI release?
OpenAI published 722 mathematical manuscripts, grouped into 372 families and selected from about 4,000 problems posed to an unnamed, unreleased model.
Have mathematicians verified the results?
The source material says the claims have not been confirmed by outside mathematicians. Some results have Lean formalizations, but not all, and OpenAI’s repository warns that unformalized results could have issues.
Does the collection prove that OpenAI’s model solved the listed conjectures?
No. The manuscripts make claims about major problems, but publication does not establish that the arguments are correct or that each paper proves the conjecture as mathematicians understand it. Independent review is still needed.
Why does the Unique Games Conjecture claim matter?
Many theoretical computer science results use the conjecture as an assumption when describing limits on approximation algorithms. If a proof is confirmed, researchers could need to revisit some conclusions that rely on it.
What would make these manuscripts valuable beyond solving problems?
Their wider value would depend on whether mathematicians can understand and reuse the methods. A correct result may settle a question; an accessible proof that provides new techniques could also lead to further discoveries.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
