SDSignal Desk

From 10 solved problems to Navier–Stokes: how OpenAI’s math run turned into a feud

· Updated · Signal Desk newsroom

Analysis of reporting by The Verge. How we work

Image: The Verge

The Verge traces OpenAI’s path from announcing 10 solved problems in August to claiming a Navier–Stokes solution in September with 10,000 concurrent agents. Historic results have been overshadowed by allegations of scooping, data-provenance questions and a confusing advisory panel.

Key takeaways

  • OpenAI announced 10 solved long-standing problems in August.
  • In September it claimed a Navier–Stokes solution using an internal model and 10,000 concurrent agents.
  • Mathematicians allege scooping and violations of academic norms; claims are disputed.
  • OpenAI acknowledged one August result built heavily on a mathematician’s earlier work.
  • A September advisory panel was called a reasonable first step but left researchers with questions.

The timeline

Over a few months, OpenAI went from announcing solutions to 10 long-standing problems in August to claiming, in September, a solution to the Navier–Stokes problem, one of the seven Millennium Prize Problems, which each carry a $1 million reward. The Verge reports the claim relied on an internal model more powerful than GPT-6 Astra running alongside 10,000 concurrent agents; OpenAI says it began training that model on August 28. Anthropic and other labs have announced their own results. OpenAI says many more results from its unreleased model are coming, and it later posted a large batch (the batch release).

By normal standards the results would be historic. Combining known methods in new ways, linking distant fields and resurfacing buried ideas is much of how mathematics advances, and machines are now doing it on problems that resisted people for decades.

The backlash

The Verge’s reporting describes a company that, after hearing others were progressing on Navier–Stokes, appears to have mounted a last-minute effort to get there first (earlier drama). Mathematicians have raised allegations of scooping and spying and say long-standing academic norms were violated; a Queen Mary University of London professor described the conduct as the kind of thing mathematicians generally do not do. These are allegations, and OpenAI has disputed characterizations of its conduct in other reporting.

Separately, a mathematician whose area underpinned one of the August results publicly questioned whether his and colleagues’ prior chatbot interactions contributed, and OpenAI acknowledged that result built heavily on his earlier work. That raises a practical worry for researchers: why share partial progress, or talk through ideas with a chatbot, if it might feed a competitor with far more compute? The Verge frames the divide simply: mathematicians want to advance the field, and OpenAI wants to win. The goals overlap until credit, provenance and timing are at stake.

The advisory panel and what comes next

The advisory panel announced in September was described by researchers as a reasonable first step, but they asked what it would do, how much influence it would have and whether a small group could speak for the wider community. Its first task, helping coordinate the release of many more results, is the scenario many researchers dread. Disquiet predates the Navier–Stokes claim; an Oxford Fields Medalist told The Verge in August he had spent the year soul-searching about the field’s future.

At scale, proofs could arrive faster than people can check them, careers built on long problems could be upended, and trust between labs and researchers could erode just as human verification matters most. The remedy is not slower math but disclosure of data and prior work used, releases coordinated with verifiers, and credit given as members of the field. Formal verification and peer review of the Navier–Stokes claim remain outstanding. More in our OpenAI coverage, Anthropic coverage and research coverage.

Who feels it

Mathematicians
Faster AI results, unclear credit norms and a coming verification load.
AI labs
Breakthroughs carry reputational risk if releases ignore norms and provenance.
Researchers using chatbots
New reason to worry whether unpublished ideas could benefit the provider.
Science broadly
A preview of how AI-accelerated discovery could strain verification elsewhere.

What to watch

  1. How OpenAI and the panel handle further releases.
  2. Formal verification and peer review of the Navier–Stokes claim.
  3. Disclosure rules on training data and prior work.

FAQ

What did OpenAI claim in September?

A solution to the Navier–Stokes Millennium Prize Problem, using an internal model and 10,000 concurrent agents, per The Verge.

What are the allegations?

Scooping, spying and violations of academic norms, raised by mathematicians; OpenAI has disputed characterizations of its conduct.

Did OpenAI acknowledge building on others’ work?

Yes, for one August result that built heavily on a mathematician’s earlier work.

Is the Navier–Stokes claim verified?

Not yet; formal verification and peer review are pending.

Read the original

Continue at the source.

The Verge

Companies: OpenAI, Anthropic

More on OpenAI

All OpenAI coverage →