OpenAI's Math Claims Draw Skepticism

Episode 1 · OpenAI Says New Model Solved 100+ Open Math Problems, Forms Advisory Group (2026-09-22, 44 posts)

On September 22, OpenAI announced it is partnering with an independent advisory group of mathematicians to guide how it responsibly shares progress at the intersection of AI and mathematics. The group operates independently of OpenAI, its members receive no compensation from the company, and it may proactively offer advice the company didn't request, comment on OpenAI's impact on the math community, and make its recommendations public.

Confirmed

• The advisory group's responsibilities include: how to evaluate and disseminate new mathematical results, maintaining academic and professional standards, and building tools that support mathematical research and learning
• OpenAI said the goal is to place mathematicians at the center, shaping how AI supports mathematical understanding and benefits the broader community
• The group was formed against the backdrop of OpenAI's AI solving more than 100 open math problems, which raised concerns in the math community (as relayed by @RexDouglass)

Why it matters

• AI producing mathematical results at scale could disrupt existing norms for academic publishing and authorship; the independent advisory group is an attempt at industry self-regulation
• The group's independence (unpaid, free to publish advice proactively) is key to its credibility, and is seen externally as a test case for how open OpenAI really is
• @RexDouglass noted some online predicted the group would disband within a year, reflecting skepticism among parts of the public about its effectiveness

• OpenAI forms independent advisory group of mathematicians to guide AI math advances — OpenAI · 2026-09-22
• OpenAI Sets Up Independent Math Advisory Group After AI Solves 100+ Open Problems — Polymarket · 2026-09-22
• OpenAI forms independent mathematician advisory group to vet AI math breakthroughs — patience_cave · 2026-09-22
• Reddit claims OpenAI solved 100 open math problems, links to advisory group post — filterdust · 2026-09-22
• OpenAI says model trained just 24 days ago has solved 100+ longstanding math problems — gaganghotra_ · 2026-09-22
• OpenAI's new model claims Navier–Stokes and 100+ open math problems in weeks — Dr_Singularity · 2026-09-22
• OpenAI's new internal model claims 100+ solved open math problems, incl. Navier–Stokes — Dr_Singularity · 2026-09-22
• OpenAI model reportedly cracked Navier–Stokes Millennium Prize problem plus 100+ open math problems — coherence · 2026-09-22
• OpenAI forms independent mathematician advisory group to guide AI-math advances — soumitrashukla9 · 2026-09-22
• GPT turns 8: model said to have resolved 100+ open math problems — lpachter · 2026-09-22
• OpenAI says its model has resolved 100+ long-standing open math problems, forms advisory group — aran_nayebi · 2026-09-22
• OpenAI claims its model solved 100+ open math problems; mathematicians ask for the list — aran_nayebi · 2026-09-22
• Repost: OpenAI model reportedly resolves Navier-Stokes plus 100+ open math problems — soumitrashukla9 · 2026-09-22
• OpenAI forms independent mathematician advisory group as model reportedly solves 100+ open problems — soumitrashukla9 · 2026-09-22
• OpenAI forms mathematician advisory group amid claims its model solved 100+ open problems — soumitrashukla9 · 2026-09-22
• OpenAI's internal math model said to have solved 100+ long-standing open problems — inductionheads · 2026-09-22
• After Terence Tao urged slower AI math progress, OpenAI formed an advisory group with no pacing power — ns123abc · 2026-09-22
• Terence Tao urges AI to slow down in math; OpenAI forms advisory group 10 days later — silver__tsuki · 2026-09-22
• Claim: a model trained for just 24 days solved 100+ open math problems — rand_longevity · 2026-09-22
• OpenAI forms independent math advisory group; skeptic predicts it dissolves within a year — RexDouglass · 2026-09-22

24 more related posts →

Episode 2 · Rumor: OpenAI's New Model Solved NS Math Problem 11 Days After Training Start (2026-09-22, 2 posts)

An unverified leak from user felpix claims OpenAI started training a new model on August 28, which solved the NS math problem within about 11 days, with math capabilities reportedly reaching 4x astra by September 8.

• Unverified: OpenAI Reportedly Trained a New Model in 1.5 Weeks That Is 4x Better Than Astra at Math — felpix_ · 2026-09-22
• Report: OpenAI Trained New Math Model in ~11 Days, Solved Longstanding Open Problems — felpix_ · 2026-09-22

Episode 3 · Developer Questions Whether Ordinary Users Can Solve Science Problems via Prompts (2026-09-22, 3 posts)

Developer felpix challenged OpenAI's claim that its models solved 100+ open math problems, arguing that ordinary users cannot drive frontier models to crack scientific problems via prompts alone, noting OpenAI itself hired PhD teams and average users rarely tackle open math with frontier models.

• Dev argues prompting alone can't get AI to solve natural science problems — felpix_ · 2026-09-22
• Amassing a PhD team is exactly what OpenAI did, dev notes in AI research debate — felpix_ · 2026-09-22
• Average users aren't throwing frontier models at open math problems, dev observes — felpix_ · 2026-09-22