Math researchers split over whether AI labs should release math results immediately

A debate erupted over whether AI labs should immediately publish their math research results, triggered by a question from DeepMind researcher Christian Szegedy (colleague of Noam Brown): if large AI labs truly care about math, should they release findings directly or wait for gatekeeper approval? Positions split sharply: one side calls for ending the gatekeeper era and dumping everything openly, while others invoke Clay Institute verification rules and worry about marketing-driven one-upmanship.

Confirmed

• Google researcher Markus Rabe argued labs should focus on problems they genuinely care about; recent math announcements are legitimate capability milestones, but he fears future ones could become prestige contests between labs. Szegedy mocked the framing that solving conjectures is a "peacocking contest."
• Commenters argued for direct release: the point of the AI era is ending gatekeeping of specialized knowledge by a few; useful results will be applied to new problems, and solving one problem does not solve all.
• Chris Brannigan replied "don't wait for the high priests, human needs are enormous," favoring immediate release.
• Mathematician Elliot Glazer backed OpenAI "slop dropping" all its math findings without curation, arguing the consequences of hiding them are worse than unrefined publication.
• Blogger panickssery half-jokingly proposed an open letter from mathematicians demanding OpenAI release all solved problems and solutions, lest humans waste effort re-solving what models already have.
• Mathematician littmath said results, human or AI, should in principle be public, but a delay of months is fine—he has results of his own that take months to write up properly. Community members jokingly besieged him to immediately release his isomonodromy foliation arithmetic work with Lam.
• DimitrisPapail offered a "100-year perspective": judged by the AI capabilities scientists will have a century from now, the importance of this decision may look different.
• JFPuget cited Clay Institute rules: millennium-prize claims require publication in a renowned journal plus a two-year waiting period for community verification.
• Blogger PAstynome pushed back against "AI companies becoming Bell Labs" narratives: proofs may take over a year to verify, and he is tired of unreviewed, unverified claims.
• Stanford professor Anshul Kundaje criticized an unnamed AI company for rushing to announce progress on open math problems (possibly scooping earlier-conceived work by others) and then touting responsible sharing—a pattern of acting irresponsibly first and branding responsibility after.

Unconfirmed

• The specific company and result Kundaje criticized, and the alleged scooping details, are not identified in the source posts.

Why it matters

• Set against a wave of math announcements from AI labs, the debate goes to the heart of publishing norms in the AI era: whether traditional peer review and verification (e.g., Clay's journal-plus-two-years rule) still fit AI-generated math, and whether lab announcements devolve into marketing contests. Glazer's "slop drop" and littmath's "months are fine" mark opposite ends of the openness spectrum, while PAstynome's verification-cost critique and Kundaje's priority-dispute accusation sharpen the stakes.

2026-09-22 ~ 2026-09-24 · 14 related posts

• Should AI labs publish math results or wait for gatekeepers? A 100-year view — DimitrisPapail · 2026-09-22
• [source] Google researcher warns AI labs' math result announcements could become a pissing match — MarkusNRabe · 2026-09-22
• Solving conjectures a 'pissing match'? AI labs' math announcements spark banter — ChrSzegedy · 2026-09-22
• AI researcher: math results should be open, but a few months' delay is fine — littmath · 2026-09-22
• Mathematician littmath responds to mob demanding immediate release of isomonodromy work — littmath · 2026-09-22
• Should AI labs publish math results directly or wait for gatekeepers' approval? — TinfoilTricorn · 2026-09-22
• Should AI labs release math results directly or wait for gatekeepers? — TinfoilTricorn · 2026-09-22
• [source] Stanford professor slams AI lab for scooping math results, then preaching responsibility — anshulkundaje · 2026-09-23
• AI labs claiming millennium-problem wins must follow Clay Institute rules first — JFPuget · 2026-09-23
• [source] Mathematician Elliot Glazer argues OpenAI should "slop drop" all its math results rather than hide them — burny_tech · 2026-09-23
• Satirical open letter pitch: mathematicians should demand OpenAI release all solutions — panickssery · 2026-09-23
• Mathematicians should demand OpenAI release all its math solutions, argues post — panickssery · 2026-09-23
• Skepticism Grows Over AI Labs' Unverified Math Breakthrough Claims — PAstynome · 2026-09-24

1 near-duplicate retellings: Chris_Brannigan