In 27 days, AI systems forced blowup out of the fluid equations — but not the version most people think

Asked:

“events and actors in the navier-stocks breakout Aug-Sep 2026”

This report reconstructs 25 dated events from August 15 to September 11, 2026, drawn from nine source pages on nine independent hosts (research current through September 12, 2026). OpenAI's paper claims finite-time blowup for the smoothly forced 3D Navier–Stokes equations on R³ and T³ — Fefferman alternatives C and D — and leaves the familiar unforced smoothness alternatives A and B untouched.

Timeline: four weeks, five threads

Each dot is one milestone; a larger dot means several sources describe the same milestone. Hover or tap a dot for the full event text and its source. Rows group events by thread; colour matches the thread.

What each event actually claims

Buckmaster and Alpöge's August results concern forced incompressible porous media, Boussinesq and 3D Euler — not Navier–Stokes — Lean-verified by August 22. cims.nyu.edu
OpenAI's internal multiagent system (~10,000 concurrent agents) reached its Navier–Stokes result about 88 hours after launch; GPT-6 Astra spent another 17 hours on Lean verification. mathoverflow.net
Anandkumar's group produced the first PINN-certified self-similar singular profile for the unforced Euler equations on R³ — an independent result released September 7 after encouragement from Terry Tao. terrytao.wordpress.com
The Clay Mathematics Institute said on September 11 the problem had “apparently been settled”, with evaluation and credit assignment deliberately unhurried; OpenAI says it will not claim the $1 million prize. claymath.org

What was actually proved — three distinct results

Forced Navier–Stokes

OpenAI

Theorem 1.1: for every viscosity ν>0 there exist a smooth force f and smooth u, p on R³×[0,1) with bounded kinetic energy and infinite speed as t→1; transported to T³. Establishes Fefferman alternatives C and D only. Not spontaneous unforced blowup — alternatives A and B remain open. mathoverflow.net

Forced IPM, Boussinesq, 3D Euler

Buckmaster & Alpöge

Finite-time blowup with smooth forcing for incompressible porous media, Boussinesq, and 3D incompressible Euler, built on a strategy of Diego Córdoba and Luis Martínez-Zoroa, Lean-verified; not Navier–Stokes itself. A hypo-dissipative Navier–Stokes paper is held pending Lean verification. cims.nyu.edu

Unforced Euler profile

Anandkumar et al.

A self-similar singular profile for the unforced incompressible Euler equations on R³ via physics-informed neural networks, with scaling exponent converging near 0.5 — inviscid Euler, not viscous Navier–Stokes. terrytao.wordpress.com

Actors, accurately distinguished

OpenAI

Internal multiagent system (~10,000 concurrent agents) for the Navier–Stokes proof; GPT-6 Astra for Lean formalization; Sébastien Bubeck among researchers in contact with Buckmaster. Says it will not claim the prize. openai.com

Tristan Buckmaster (NYU) & Levent Alpöge (Anthropic)

Obtained the August forced-blowup results for Boussinesq and 3D Euler using LLMs; went public September 7, roughly 12 hours before OpenAI's announcement. quantamagazine.org

Diego Córdoba & Luis Martínez-Zoroa

Originators of the key mathematical strategy the AI-enabled efforts built on. cims.nyu.edu

Anima Anandkumar & collaborators (Caltech)

Independent PINN-based unforced-Euler singular profile; says mainstream coverage followed OpenAI's press release without acknowledging their work. terrytao.wordpress.com

Terry Tao

Commentator; encouraged the Caltech group's public release and hosted their guest post. terrytao.wordpress.com

Clay Mathematics Institute

Evaluator, not discoverer: the problem is “apparently settled”; evaluation and credit are deliberately unhurried. claymath.org

The dispute — two competing accounts, not adjudicated facts

Buckmaster's account: news of his and Alpöge's progress leaked to OpenAI; his Codex activity may have influenced OpenAI's system; OpenAI proposed publication arrangements excluding Alpöge because he works at Anthropic. cims.nyu.edu
OpenAI's account: employees and agents had no access to their work; its September 10 investigation update says Buckmaster's Codex prompts could not have influenced the system; it offered concurrent release to recognize their forced-Euler priority. openai.com
Lean formalization is strong machine checking, but it is not completed independent human review or final prize assignment; mathematicians are still digesting the 166-page paper. sciencenews.org

Source and claims guide — nine hosts compared

SourceTitleDispute framingValidation framing
nature.comOpenAI claims huge maths breakthrough on a famed ‘Millennium Problem’OpenAI focused resources on Navier-Stokes after hearing rumours that Levent Alpöge and Tristan Buckmaster had solved some version of the fluid-motion puzzle using Anthrop…Announced by OpenAI via a press briefing and online statement; accompanied by the release of a paper by Alpöge and Buckmaster and a release by Anandkumar and collaborator…
cims.nyu.eduPublic statement (PDF)Tristan alleged that information about his and Levent's progress was passed to OpenAI, after which OpenAI's internal research model reportedly produced a proof of finite-…Three results made public (IPM, Boussinesq, 3d incompressible Euler) with Lean verification completed for them (though Boussinesq and Euler write-ups described as unpolis…
quantamagazine.orgAI Has Solved One of Math’s $1 Million Millennium Prize Problems | Quanta MagazineOpenAI admits their work was inspired by rumors that Alpöge and Buckmaster had solved a Millennium Prize problem (though OpenAI claims priority for the Navier-Stokes resu…The OpenAI result for Navier-Stokes has been formally checked in the programming language Lean (and formalized by another AI model after 17 hours). Buckmaster and Alpöge …
mathoverflow.netProof outline and discussion: OpenAI's claimed forced-blowup construction for Navier–Stokes (Fefferman alternatives C/D)Tristan Buckmaster's public statement gives an account of contact with OpenAI in the preceding days that OpenAI's own post doesn't fully corroborate, alongside simultaneo…The paper features a Lean 4 formalization alongside it (with 17 hours spent on Lean verification). OpenAI's post states they are not claiming the Clay Prize for it, and a…
sciencenews.orgAI may have solved one of math’s biggest puzzles, raising controversyDisputes arose regarding credit attribution, provenance, and coauthorship. Tristan Buckmaster alleged that discussions with OpenAI researchers about how to present the tw…OpenAI's solution has been confirmed by Lean, though mathematicians are still digesting its 166-page paper and human verification is not yet complete. OpenAI stated in a …
terrytao.wordpress.comStable singularity of the Euler equations on R^3Mainstream media religiously followed OpenAI's press release, which fails to acknowledge the author's work even after being informed.
claymath.orgNavier-Stokes Announcement - Clay Mathematics InstituteThe process for evaluating what has been achieved and for assigning credit is described by the rules governing the prizes as deliberately unhurried, with updates to be pr…
openai.comOn the Navier–Stokes Millennium Prize ProblemOpenAI heard rumors on September 1st regarding Millennium Prize resolutions, which turned out to relate to work by Levent Alpöge and Tristan Buckmaster. OpenAI offered a …Validation included an analytical proof and a formalization in Lean (checked and verified via GPT-6 Astra taking 17 hours, completed by September 6, 2026). OpenAI explici…
science.orgHow an AI math breakthrough ignited a controversyA fierce dispute arose over who deserves credit for the solution. Tristan Buckmaster and Levent Alpöge had made significant progress over the prior month using AI tools, …The solution was formalized in Lean (a programming language that automatically verifies mathematical proofs). The announcement has not yet been officially confirmed by in…

Method: 25 dated event descriptions (Aug 15 – Sep 11, 2026) extracted from nine source pages on nine independent hosts, inspected through September 12, 2026; the timeline groups same-day accounts of one milestone from independent sources into a single dot (dot size = number of accounts). Dispute and validation texts in the table are truncated for space; full texts are at the linked pages. Framing differs across sources; the C/D forced-blowup scope follows the MathOverflow mathematical outline.

This report was generated automatically by Keenable SELECT at a user's request, from publicly available web sources linked herein. Keenable does not review, verify, or endorse its contents and makes no representation as to accuracy, completeness, or timeliness; AI-based extraction may contain errors. Nothing in this report is investment, legal, financial, or other professional advice. All trademarks and referenced content remain the property of their respective owners; no affiliation or endorsement is implied. To report an error, rights concern, or request removal: legal@keenable.ai.

Keenable SELECTAsk your own question
Made with Keenable SELECT