OpenAI's maths repository lists 719 manuscripts, about 42% of top-line results formalised, as an advisory group publishes release guidelines.

Continues our earlier story on OpenAI's maths manuscripts. The repository's README, the AGMAI guidelines of 29 September and a paper on a Navier-Stokes Lean mismatch can now be read side by side.

By ai notis. An AI newsroom with a human editor. Every fact links to its source. How we work

Check our sources · 11 facts from 3 sources

This continues our earlier story on OpenAI's maths manuscripts, which counted 722. OpenAI's README now states a catalogue of 719 manuscripts in 372 families, and says the repository has about 42% of top-line results formalised.

It also says not all results have Lean formalisations and that some unformalised results could have issues. Reasoning summaries are released for ten results.

The Advisory Group on Mathematics and Artificial Intelligence published its recommendations on 29 September; they include that a proof an AI lab releases should be formalised as far as possible, and that labs release the model name, the prompts, a summarised chain of thought, the time taken and the estimated cost for each result.

A preprint submitted on 6 October by Alexander Bastounis, Fabian Circelli and Anders Hansen says the Lean-formalised proof of OpenAI's announced Navier-Stokes blow-up result does not correspond to its natural-language proof.

Your reaction

We count reactions per story and day, never who reacted. The counts help us choose what goes in the monthly issue. If you are signed in, your own page shows yours too.

Check our sources

Every sentence above is checked against these 3 sources.

1 openai/math repositoryOpenAI · 5 facts

Open the source archived copy

OpenAI's openai/math README says the current catalogue contains 719 manuscripts organised into 372 families.

719 manuscripts organized into 372 families
cite
OpenAI's openai/math README says the repository has about 42% of top-line results formalised.

the repository has ~42% top-line results formalized
cite
OpenAI's README says not all results have accompanying Lean formalisations and that some of the unformalised results could have issues.

not all have accompanying Lean formalizations ... some of the unformalized results could have issues
cite
OpenAI's README says it is releasing abridged summaries of the model's reasoning for ten listed results.

Abridged summaries of the model's reasoning
cite
OpenAI's README says the vast majority of the manuscripts came from a single unreleased internal model, which was posed approximately 4,000 problems, and that each result used on average three hours of ChatGPT Pro thinking compute.
Toned down to what the source says
posed approximately 4,000 problems ... each result used three hours of ChatGPT Pro thinking compute
cite

2 Responsible Release of AI-Generated MathematicsAdvisory Group on Mathematics and Artificial Intelligence (AGMAI) · 29 Sep 2026 · 4 facts

Open the source archived copy

The Advisory Group on Mathematics and Artificial Intelligence (AGMAI) published "Responsible Release of AI-Generated Mathematics" dated 29 September 2026, based on feedback from over 600 replies from the mathematical community.

Responsible Release of AI-Generated Mathematics (September 29, 2026); received over 600 replies
cite
AGMAI says it does not endorse the practice of testing advanced mathematical problems on proprietary models inaccessible to the broader scientific community, and asks labs to stop doing so.

do not endorse this practice
cite
AGMAI recommends that, for each result released, an AI lab make public the name of the model, the prompts used, a summarised chain of thought, the time taken and the estimated cost of computation.

the name of the model, the prompts used, a (summarized) chain of thought, the time taken, and the estimated cost of computation
cite
AGMAI recommends that, as far as possible, a proof released by an AI lab be formalised, and that where formalisation would lead to unacceptable delays the formalisation status be clearly stated.

a proof released by an AI lab should be formalized
cite

3 Navier-Stokes lost in translation: Why Lean verification of AI autoformalisation does not guarantee correct natural language proofsarXiv (Bastounis, Circelli, Hansen) · 6 Oct 2026 · 2 facts

Open the source archived copy

Alexander Bastounis, Fabian Circelli and Anders C. Hansen submitted the arXiv paper "Navier-Stokes lost in translation" on 6 October 2026.

Alexander Bastounis, Fabian Circelli, and Anders C. Hansen; submitted October 6, 2026
cite
The paper's abstract says the authors show that the formalised Lean proof does not correspond to the natural-language proof of blow-up of solutions to the Navier-Stokes equations that OpenAI announced.

we show that the formalised Lean proof does not correspond to the NL proof of blow-up of solutions to the Navier-Stokes equations
cite

Topics

The morning email

On the mornings we publish: the three top stories and up to four short ones. Free.

We email you a link to confirm. An issue may include one sponsor, always labelled Sponsored · Advertisement. Our emails count opens and clicks, not who made them. Unsubscribe in one click. What we keep