Tristan Buckmaster

Sources#

Summary#

Mathematician at NYU's Courant Institute, working on singularities of the Euler and Navier–Stokes equations and on convex integration, the technique he says he "built my whole career on". He was a student of László Székelyhidi, coauthored the Buckmaster–Vicol convex-integration work, and had a prior collaboration with Google DeepMind. With Levent Alpöge (Anthropic, working on it as a side project), he produced the Lean-verified forced-Euler blow-up that OpenAI concedes priority on. Alpöge built the agentic system they used, and Buckmaster also used OpenAI's Codex. They released on 2026-09-07, 12 hours before OpenAI's Navier–Stokes announcement (The Navier–Stokes AI Claim).

Until 2026-10-16 the corpus held him only through quotation: his statement PDF, as Quanta and De Toffoli–Duede quote it ("a Deep Blue–Kasparov moment", "the most horrendous [proof] I have ever read", "AI slop"). His Numberphile interview with Brady Haran, uploaded 2026-09-29 (The Buckmaster Interview (Navier-Stokes) - Numberphile, practitioner-opinion, auto-captions), is the first first-hand source from the rival side. It came three weeks after OpenAI's 09-10 "could not have influenced" update.

What he claims, first-hand#

All of it is one interested party's account, and OpenAI had not given Numberphile an interview at upload time.

On the result: Navier–Stokes "has been solved"; OpenAI's paper "is not … in a readable form but … the ideas can be digested".
On the dispute: an Anthropic-internal leak started the rumour. He reached out to OpenAI first, to de-escalate. On the call, OpenAI's researchers presented a one-prompt solve that "evaporated within minutes", and Bubeck agreed they began only after the rumour. His influence case rests on unforced Euler, not Navier–Stokes. Detail and the contradictions with OpenAI's account are on The Navier–Stokes AI Claim.
On authorship: he says he was offered sole authorship of OpenAI's paper if he dropped Alpöge, because Alpöge is an Anthropic employee (Cross-Lab Pre-Release Review).
On the mathematics: the step from Euler to Navier–Stokes combines convex integration with the Euler growth mechanism to correct a forced non-solution. He calls it "a smart idea" that OpenAI's own introduction does not describe (Logical vs Intelligible Proof).
On the field: mathematicians now hide open problems from the labs, and he is withholding further results of his own (Proof Abundance).

Connections#

The Navier–Stokes AI Claim — the rival claimant whose forced-Euler priority OpenAI concedes, and whose first-hand account contests OpenAI's "Concurrent work" section on contact, timing and influence
Cross-Lab Pre-Release Review — the individual researcher caught between two labs. A reported authorship offer that depends on a collaborator's employer is the rivalry acting on a paper's byline
Proof Abundance — an independent report, two weeks before Tao's, of the same chilling effect on sharing open problems
Logical vs Intelligible Proof — his verdicts on his own team's AI-written proofs, and his claim that the Navier–Stokes paper misdescribes its own key idea

Sources#

The Buckmaster Interview (Navier-Stokes) - Numberphile — Numberphile interview (Brady Haran), uploaded 2026-09-29, 46 min, practitioner-opinion, YouTube auto-captions. The primary source for everything under "What he claims, first-hand".
AI Has Solved One of Math's $1 Million Millennium Prize Problems — Quanta, 2026-09-08: his statement quoted second-hand, and the release dates.
After Math — the "Deep Blue–Kasparov moment" quotation.
On the Navier–Stokes Millennium Prize Problem — OpenAI's priority concession and its account of contact with him.