OpenAI’s Sebastien Bubeck Says He Tried To Coordinate Release Of Navier Stokes-Related Proofs With Buckmaster & Alpoge But Was Rebuffed

OpenAI scientist Sebastien Bubeck has released a detailed account of his interactions with NYU mathematician Tristan Buckmaster and Anthropic researcher Levent Alpöge over the Navier-Stokes proof dispute, denying that he tried to strip Alpöge of authorship and revealing text messages in which he proposed a coordinated release. Bubeck says the conflict arose when he learned the pair had solved the Euler equations, not the full Navier-Stokes problem, and that he floated the idea of Buckmaster leading a rewrite of OpenAI's proof, which would have been 'simpler' if Alpöge were not an Anthropic employee. He also claims he faced 'unfounded accusations' and a threat to go to the press, and that his 'why would you risk your career' remark was an ill-chosen phrase he retracted immediately.

Sebastien Bubeck, the OpenAI scientist at the center of the Navier-Stokes credit dispute, has posted his promised fuller account of what happened between him, NYU mathematician Tristan Buckmaster, and Anthropic researcher Levent Alpöge — pushing back on the characterization that he tried to strip Alpöge of authorship, while also surfacing a text message exchange that shows him reaching out to Alpöge directly to propose a coordinated release. Bubeck opens by pointing to a screenshot of his own texts to Alpöge as evidence of good faith. In the messages, sent on a Sunday, he tells Alpöge he wants to be as transparent as possible and lays out that an OpenAI internal model had produced a proof of Navier-Stokes with minimal human involvement — but stresses that credit for the result should go to Alpöge and Buckmaster, and offers to share everything OpenAI has, including the prompts used. Alpöge’s terse reply asks what exact theorem is being claimed; Bubeck responds that it’s the existence of forced blow-up in R^3 and T^3. From there, Bubeck’s account tries to walk back the most damaging line from Buckmaster’s original statement — that Bubeck twice pushed to have Alpöge dropped from authorship because of his ties to Anthropic. Bubeck says he never asked for Alpöge to be removed from his own work. He says the confusion arose because, partway through the call with Buckmaster, he learned for the first time that the pair had only solved the Euler equations rather than the full Navier-Stokes problem. That prompted a scramble to figure out what to do next, and one idea floated was having Buckmaster lead-author a rewrite of OpenAI’s own Navier-Stokes proof — a scenario in which, Bubeck says, it would have been “simpler” if Alpöge weren’t an Anthropic employee, since he felt it would be improper for a rival company’s staffer to co-author OpenAI’s work, especially once it came out that Anthropic’s internal models had been used in the Euler proof. He says a second idea he wanted to raise but didn’t get to was offering the pair access to OpenAI’s internal model so they could try to close the gap between Euler and full Navier-Stokes themselves. Bubeck’s account gets more pointed on the substance of the call itself. He says that rather than being received as an attempt to celebrate the pair’s work, he was met with what he describes as a barrage of unfounded accusations and a threat to go to the press if OpenAI proceeded with its own announcement. It’s here that he places the now-infamous line — “why would you risk your career” — saying it was an ill-chosen remark made in frustration while trying to salvage a chance to give the mathematicians full credit, and that he retracted it immediately on the call. He calls it a poor choice of words that conveyed the opposite of what he meant. He also takes aim at how the dispute played out logistically, saying it was Buckmaster who represented Anthropic’s side throughout, since Alpöge himself declined to join any of the calls despite repeated invitations — something Bubeck frames as awkward given that, as OpenAI’s Sholto Douglas has said, the two labs are going to need to find a way to work together going forward. On the technical origin story, Bubeck reiterates a version of the account OpenAI has already given saying the company’s team began looking at the Millennium Prize problems after viral rumors claimed Anthropic had already resolved two of them, and that the effort was partly a test of how far its internal model’s recent capability gains could go. He again insists that neither OpenAI’s researchers nor its agents saw any of Buckmaster and Alpöge’s work before it was released publicly, arguing that the two teams’ proofs — and even the exact problems they solved, forced versus unforced Euler — turned out to be meaningfully different. He adds that OpenAI’s side of the effort was staffed largely by people with academic backgrounds in math, physics and computer science, but with no research-level expertise in fluid dynamics specifically, which he says explains why the humans involved couldn’t meaningfully shape the mathematical content themselves. Bubeck also offers a new detail on how OpenAI first zeroed in on Buckmaster and Alpöge: he says it was Alpöge who reached out unprompted to an OpenAI employee earlier in the week, and that a string of Alpöge’s cryptic tweets, combined with the earlier viral rumors, left OpenAI believing the project was at least partly an Anthropic effort and that a Millennium Prize result might be imminent. The post is Bubeck’s attempt to close out a dispute that pulled in researchers from both companies trading barbs on X https://officechai.com/ai/anthropic-openai-researchers-spar-over-what-appears-to-be-credit-for-progress-towards-solving-navier-stokes-equation/ after his first, brief denial calling Buckmaster’s allegations “false and inflammatory.” https://officechai.com/ai/openais-sebastien-bubeck-calls-tristan-buckmasters-claims-of-trying-to-take-credit-for-fluid-dynamics-proofs-false-and-inflammatory/ It comes on the same day OpenAI published its own account denying it had seen any of the pair’s work before publication https://officechai.com/ai/openai-says-it-didnt-see-any-of-buckmaster-and-alpoges-work-while-resolving-navier-stokes/ , and shortly after the company released its Navier-Stokes solution https://officechai.com/ai/openai-shares-solution-to-navier-stokes-problem-created-by-a-model-significantly-more-capable-than-astra/ attributed to a model it says is significantly more capable than its predecessor. Buckmaster and Alpöge’s original claims, laid out in a lengthy statement, first surfaced in a report on the mathematician’s allegations that OpenAI pressured him over credit https://officechai.com/ai/mathematician-tristan-buckmaster-says-he-cracked-a-fluid-dynamics-problem-with-ai-accuses-openai-of-trying-to-take-credit/ for the pair’s finite-time blow-up proofs. Neither Buckmaster nor Alpöge has publicly responded to Bubeck’s latest post as of this writing.