Mind · In / Out · In · X 长帖

Tristan Buckmaster 的声明

Statement by Tristan Buckmaster

Tristan Buckmaster · NYU 个人页 PDF · 2026-09-08

一位流体数学家同时宣布了两件事:LLM 助推做出三维 Euler 的爆破解,和一场与 OpenAI 的争议。

Indigo 的结论

数学层是「可验证的领域」这条线迄今最硬的一击:LLM 助推做出三维 Euler 爆破这种著名难题,并经 Lean 验证。争议层是分量很重的一手陈述,但只有一方在场,只能当「未经对方回应的说法」读。

怎么读这篇 一手文件,但对争议来说是一面之词。分两层读:数学结果有 Lean 验证,论文和形式化即将公开,可以独立复核,是硬事实;与 OpenAI 的往来全部是 Buckmaster 本人的陈述,对方缺席、没有回应。他自己也写了:「我没有指控任何人」。

需要记住的几件事

  1. 生成一侧的硬里程碑:LLM 助推做出流体方程的新爆破解,含三维 Euler,经 Lean 验证;一个数学家加一个 LLM,一个月做完。
  2. 机械终审器让「先信任、后理解」成立:第一份证明读不懂,但 Lean 先验证通过,人再慢慢读懂。
  3. 署名、审稿、培养学生都受冲击:他呼吁学界认真、不慌地讨论,什么值得一个人用一生去关注。
  4. 一场未经对方回应的争议:据其声明,OpenAI 一方的说法在通话中被推翻,还施压要去掉合作者署名。全是一面之词。
  5. 归属上的诚实是强信号:思路归 Córdoba 和 Martínez-Zoroa,自贬书面稿是 AI 垃圾,明写不指控任何人。

拆解 · 3 步

  1. 01

    LLM 助推,一个月做出三个爆破解

    三个光滑外力下的有限时间爆破,含三维 Euler;思路归 Córdoba 等人,LLM 把纲领推到终点,8 月 22 日经 Lean 验证。 读这一段原文 →

  2. 02

    争议:据其声明,OpenAI 也说证出来了

    据其声明,他主动联系 OpenAI 后被告知,内部模型证出了有外力 Navier-Stokes 的爆破;起初说几乎没有人工参与,后来改口。 读这一段原文 →

  3. 03

    两个方案、一句重话,和他没说的

    据其声明,对方提议分先后发布,或只署他一人、承认 OpenAI 模型解决;他都拒绝了,并强调不指控任何人。 读这一段原文 →

对 Rewired Index 意味着什么

两种读法都要谨慎:一是 AI 做前沿科研的能力实证,利好 AI 做科研的叙事,以及评测和形式化工具链;二是若声明属实,对 OpenAI 的声誉和治理是负面信号。没有直接对应的标的。

什么会让我改口

论文和 Lean 形式化公开后,独立复核推翻了这些爆破结果。

怎么读这篇

一手文件,但对争议来说是一面之词。分两层读:数学结果有 Lean 验证,论文和形式化即将公开,可以独立复核,是硬事实;与 OpenAI 的往来全部是 Buckmaster 本人的陈述,对方缺席、没有回应。他自己也写了:「我没有指控任何人」。

拆解 · 3 步
  1. LLM 助推,一个月做出三个爆破解
  2. 争议:据其声明,OpenAI 也说证出来了
  3. 两个方案、一句重话,和他没说的
01

LLM 助推,一个月做出三个爆破解

三个光滑外力下的有限时间爆破,含三维 Euler;思路归 Córdoba 等人,LLM 把纲领推到终点,8 月 22 日经 Lean 验证。

今天,Levent Alpöge 和我公开了三个结果:在光滑外力作用下,不可压缩多孔介质方程、Boussinesq 方程和三维不可压缩 Euler 方程的有限时间爆破。

我们相信,我们也得到了低耗散(hypo-dissipative)Navier-Stokes 方程的爆破。那篇论文今天不发:和上面几项不同,它的 Lean 验证还没有完成,我们也还没有任何称得上能拿出来的书面稿。我提到它,是因为它暗示了一条通向无外力 Euler 方程的路。

这项工作所属的研究纲领,不是我们开创的,也不是大语言模型提出的。这个纲领的基本思想,功劳归于 Diego Córdoba 和 Luis Martínez-Zoroa,他们多年来一直在探索如何构造有外力的爆破。我们以他们的工作为起点,借助大语言模型把他们的纲领推到了终点。

具体来说,Levent 和我做的,是拿 Córdoba 和 Martínez-Zoroa 的纲领(它在粗糙外力下得到了爆破结果),在 LLM 的大量帮助下,把它推进到光滑外力,并推进到不可压缩 Euler 方程。让这条进攻路线成为可能的想法,归功于 Córdoba 和 Martínez-Zoroa。让我把私下对同行说过的话公开说清楚:鉴于这一整套工作,我认为 Luis Martínez-Zoroa 配得上菲尔兹奖。

我和 Levent 的合作纯属个人之间的合作,没有任何机构协议,我们双方的雇主也都没有正式参与。整个过程中我们用了几种 LLM:Anthropic 的 Claude,OpenAI 的 Codex(尤其是配合 GPT-5.6 Sol),以及最近的 Astra。Astra 只用于撰写书面稿和审查我们的论证。

过去一年的大部分时间里,进展很慢。我们把文献过了一遍,改进了各种初步结果,一直做到得出不可压缩多孔介质方程(带光滑外力)的有限时间爆破。直到大约一个月前,我们才有了真正的突破:8 月 15 日,我们得到了 Boussinesq 和 Euler 两者在光滑外力下的爆破结果。我可以说,Levent 发给我的第一份 LLM 生成的证明,是我读过最可怕的;8 月 22 日,我们在 Lean 上验证了它。从那时起,我们一直在昼夜不停地理解这份证明,把它变成可读的东西。

这个故事还有另一部分,老实说,我非常希望自己不必为它操心。我对这些论文的呈现质量并不满意。理想情况下,我们本想花上几周,把 LLM 生成的证明从第一页起就改写得可读。这些问题、以及投身于这些问题的学界,值得这样的用心。尤其是 Boussinesq 和 Euler 的几份书面稿,更接近模型在人的指导下产出的东西,而不像一个人写的论文。尤其是 Euler 那份,只能用「AI 垃圾」来形容。对此我很抱歉。原因写在下面,涉及我们受到的外部压力。我这样说不是找借口,而是解释。

我原本打算在宣布成果时说:结果本身不是重点。重点在于,一个数学家加一个 LLM 模型,现在一个月就能做完这一切。这对我们如何培养学生、如何分配功劳、如何审稿、如何决定什么值得一个人用一生去关注,意义怎么说都不为过。这是一个「深蓝对卡斯帕罗夫」的时刻。学界需要认真、不慌不忙地讨论接下来往哪里走。可我没能写这些极其重要的进展,却发现自己在写另一件事。我想尽可能直白地把发生了什么讲清楚。

02

争议:据其声明,OpenAI 也说证出来了

据其声明,他主动联系 OpenAI 后被告知,内部模型证出了有外力 Navier-Stokes 的爆破;起初说几乎没有人工参与,后来改口。

9 月 3 日星期四,有传言说 Anthropic 解决了一个重大公开问题,Levent 也收到消息,说我们进展的信息被传给了 OpenAI。于是我写信给 OpenAI 一位知名的数学家。我全文引用这封邮件,因为我宁愿大家读原文,而不是读我的概括:

「你好 [……],我们没见过面,但我是 [……] 的讲者之一。我写信是因为,有个传言似乎传得很快,说 Anthropic 解决了一个重大公开问题。我不能确定传言说的是不是我,但昨天 Courant 的一位同事发邮件跟我提了这事,他是听英国一位分析师说的,而那位分析师又是从更上游的某处听来的,所以大概可以认为说的就是我。我听说科技圈里也在流传把 Levent Alpöge 和某个解答联系在一起的版本。不过,对于到底解决了哪个问题,似乎有很多混淆。

我还应该强调,这不是机构层面的工作。这完全是我们两人之间的个人合作,背后没有任何正式协议。我团队用的工具是我用自己的科研经费付的钱,其中包括付给 OpenAI 的一大笔账单。我以前有过一次业界合作,是和 DeepMind,那次有正式的机构安排。

我们在这个方向上确实有有把握的工作,很快会把论文和形式化证明一起发出来。我们有意决定不抢先放出一份 Lean 证书配一份没打磨的预印本。我坚持认为,任何人读到的第一样东西,应该是以常规方式呈现的数学论证,而不只是一份形式化证书。

我直接写信给你,而不是公开说什么,是为了让你掌握事实,好在你们那边处理这件事。祝好,Tristan」

他当天回复:「如果你愿意提供任何细节,会有助于我们在这件事上避免竞争;总的来说,看到数学家用我们的模型取得进展,我们总是很高兴。另外,如果在算力方面 OpenAI 这边能提供什么,我们很乐意。」

我提出下周再谈。9 月 4 日星期五,对方问我能不能当天见面;我又说下周。9 月 6 日星期日 12:45,对方问我能不能「今天任何时候」见面。Sebastien Bubeck 加入了。那天下午我们三个人谈了两次。Levent 不在通话里。

我被告知,OpenAI 的一个内部模型给出了有外力 Navier-Stokes 方程有限时间爆破的证明。Levent 发短信询问精确的表述,得到的回答是:「在 R³ 和 T³ 中存在有外力的爆破」,并且「外力函数光滑,对应 Fefferman 表述里的选项 c 和 d」。我被告知证明大约 100 页。我没有见过它。

这里我应该说明,我为什么会那样理解他们的说法,也就是下面要谈的那种理解。经由光滑外力通向克雷问题的路线,也就是 Fefferman 对这个问题的表述里的选项 c 和 d,正是 Luis 和 Diego 开辟、Levent 和我悄悄选定去攻的路线。据我所知,几乎没有别人在做。这不是给模型一个问题陈述、几天之内就能走到的方向。我听到「有外力」时,那是一面刺眼的红旗。

他们给我看了一个提示词,告诉我内部研究模型只是拿到了问题陈述。Sebastien 告诉 Levent,用了「极少的人工输入」。后来证明并非如此。在通话过程中,随着他们团队成员通过内部聊天给 Sebastien 发去更正和细节,情况逐渐清楚:一整个团队一直在做这个问题;这只是他们尝试的多件事之一;工作最初是从无外力问题开始的;团队先让模型做了更容易的问题,包括 Euler;就连给我看的那个提示词,也是通过给 Codex 写提示生成的;而且用了疯狂多的算力。

我问他们的第一个提示是什么时候发出的。OpenAI 有一段时间没有直接回答这个问题。最后双方认定,它是在过去几天里发出的,也就是在我们工作的信息传到 OpenAI 之后。

我问这个模型是否用我们在 Codex 里的会话训练过,或者能否访问这些会话;整个项目期间,我们把所有草稿都放在那里。我被告知模型不会查阅用户数据。我又问了一次训练的事,没有得到回答。

03

两个方案、一句重话,和他没说的

据其声明,对方提议分先后发布,或只署他一人、承认 OpenAI 模型解决;他都拒绝了,并强调不指控任何人。

对方向我提了两个方案。第一个是我们发布 Euler 的结果,OpenAI 第二天发布它的 Navier-Stokes 结果。第二个是在发布 Euler 之后,由我一个人写一篇论文来介绍 Navier-Stokes 的结果,并承认是 OpenAI 的一个内部模型解决了它。Sebastien 两次坚称希望把 Levent 从作者中去掉,还说要不是 Levent 在 Anthropic 工作,一切都会很简单,这件事实在太烦人了。对方还说,如果 OpenAI 在我们之后发布,他们会说我们配得上克雷奖,说我们是「离这个问题最近的人类」。两个方案我都拒绝了。

我说,如果 OpenAI 按提议的方式发布结果,我会把发生的事公开。对方的回答是:「你为什么要毁掉自己的职业生涯?」我回答说我是学者,并问他为什么认为公开会毁掉我的职业生涯。对方的回答是:「如果你不想让我客气,那我就不必客气了。」

过了一段时间,Levent 收到一条短信,提议他和 Sebastien 单独谈,里面说:「我不知道 Tristan 现在是否完全理性。」Levent 拒绝了,说要谈就和我谈。当晚 Sebastien 又发来一封邮件,要求 9 月 7 日星期一和我谈。我没有回复。

我想说清楚,哪些是我没有主张的。我没有见过 OpenAI 的证明。我不知道他们的模型做了什么,怎么做的。我不知道我们的数据有没有被使用。我没有指控任何人任何事。我陈述的是我被告知了什么、什么时候被告知,以及对方向我提了什么。我之所以陈述,是因为另一种选择,是任由一连串的发布去说一件我知道是假的事。

如果 OpenAI 的模型确实补上了通向 Navier-Stokes 的这段距离,那是一件了不起的事,应该由他们大声说出来,同时保留完整的来龙去脉。我更愿意谈的是数学,是 Luis 和 Diego 的想法,以及这一切对我们其余的人意味着什么。

最后,我要感谢整个数学界,在过去 24 小时里给了我这么多支持。

判断收口延伸

Indigo 的结论

数学层是「可验证的领域」这条线迄今最硬的一击:LLM 助推做出三维 Euler 爆破这种著名难题,并经 Lean 验证。争议层是分量很重的一手陈述,但只有一方在场,只能当「未经对方回应的说法」读。

需要记住的几件事

  1. 生成一侧的硬里程碑:LLM 助推做出流体方程的新爆破解,含三维 Euler,经 Lean 验证;一个数学家加一个 LLM,一个月做完。
  2. 机械终审器让「先信任、后理解」成立:第一份证明读不懂,但 Lean 先验证通过,人再慢慢读懂。
  3. 署名、审稿、培养学生都受冲击:他呼吁学界认真、不慌地讨论,什么值得一个人用一生去关注。
  4. 一场未经对方回应的争议:据其声明,OpenAI 一方的说法在通话中被推翻,还施压要去掉合作者署名。全是一面之词。
  5. 归属上的诚实是强信号:思路归 Córdoba 和 Martínez-Zoroa,自贬书面稿是 AI 垃圾,明写不指控任何人。

放回主线

证实

可验证域能否泛化 生成一侧迄今最硬的一击:著名 PDE 难题上的新结果,加 Lean 验证,是真正的新数学。

证实

验证不可压缩 第一份证明最可怕,Lean 验证通过后才慢慢读懂:有机械终审器的领域,验证可以被承担,生成可以是黑箱。

补充

人事就是路线图 若声明属实,是前沿实验室竞赛强度和规范的一个重磅但未证实的数据点:抢发、施压去署名、跨公司的敏感。

补充

Anthropic:Claude 把费马大定理形式化进 Lean 那篇是验证一侧,这篇是生成一侧;合起来,有 Lean 的数学领域两端都被接手,终审都靠 Lean。

证实

Sabine Hossenfelder:AI 正在接管物理 Buckmaster 就是 Sabine 说的面对 AI 的数学家,危机具体到了署名、审稿和培养学生。

补充

Jakub Pachocki《An Alien Mind》 Pachocki 说 AI 生成的数学会泛滥、验证是信任层;这份声明是那个未来的现场,也包括竞赛的操守压力。

对 Rewired Index 意味着什么

两种读法都要谨慎:一是 AI 做前沿科研的能力实证,利好 AI 做科研的叙事,以及评测和形式化工具链;二是若声明属实,对 OpenAI 的声誉和治理是负面信号。没有直接对应的标的。

什么会让我改口

论文和 Lean 形式化公开后,独立复核推翻了这些爆破结果。

读完了。Indigo 对这篇的判断在这两处:

Mind · In / Out · In · X thread

Statement by Tristan Buckmaster

Tristan Buckmaster · cims.nyu.edu · 2026-09-08

A fluid dynamicist announced two things at once: an LLM-assisted blowup for 3D Euler, and a dispute with OpenAI.

Indigo's conclusion

The math layer is the hardest hit yet for the verifiable-domains thread: an LLM helped produce a blowup for 3D Euler, a famous hard problem, verified in Lean. The dispute layer is a weighty first-hand account, but only one side is present, so read it as an account the other side hasn't answered.

How to read this A first-hand document, but on the dispute it's one side's account. Read it in two layers: the math has Lean verification, with the paper and formalization to come, so it can be checked independently and is hard fact; everything about the dealings with OpenAI is Buckmaster's own account, and the other side is absent and hasn't replied. He writes it himself: “I am not accusing anyone.”

What to remember

  1. A hard milestone on the generation side: LLM-assisted new blowups for fluid equations, 3D Euler included, verified in Lean; one mathematician and one LLM, in a month.
  2. A mechanical final check makes “trust first, understand later” work: the first proof was unreadable, but Lean verified it first and people understood it later.
  3. Credit, refereeing and training students are all shaken: he calls for serious, unhurried discussion of what deserves a human lifetime's attention.
  4. A dispute the other side hasn't answered: by his account, OpenAI's story shifted during the call, with pressure to drop a co-author. One side only.
  5. Honesty about credit is a strong signal: the program credited to Córdoba and Martínez-Zoroa, his own write-up called AI slop, and an explicit “I am not accusing anyone”.

Breakdown · 3 steps

  1. 01

    LLM-assisted: three blowups in a month

    He and Alpöge release three finite-time blowups with smooth forcing, 3D Euler included. The program belongs to Córdoba and Martínez-Zoroa; LLMs pushed it to completion, verified in Lean on August 22. Read this part →

  2. 02

    The dispute: by his account, OpenAI says it proved it too

    By his account, as rumors spread he contacted OpenAI; on calls he was told an internal model had proved forced Navier-Stokes blowup, first with “very little human input”, a story that shifted as the call went on. Read this part →

  3. 03

    Two offers, one hard line, and what he isn't claiming

    By his account, he was offered staggered releases, or a paper under his name alone crediting OpenAI's model; he declined both. He stresses he hasn't seen their proof and accuses no one; he states only what he was told. Read this part →

What it means for Rewired Index

Both readings need care. It's evidence of what AI can do in frontier research, which supports the AI-for-science story and the evaluation and formalization toolchain; and if the account holds, it's a negative for OpenAI's reputation and governance. No direct name to point at.

What would change my mind

the paper and the Lean formalization go public and independent review overturns these blowup results.

How to read this

A first-hand document, but on the dispute it's one side's account. Read it in two layers: the math has Lean verification, with the paper and formalization to come, so it can be checked independently and is hard fact; everything about the dealings with OpenAI is Buckmaster's own account, and the other side is absent and hasn't replied. He writes it himself: “I am not accusing anyone.”

Breakdown · 3 steps
  1. LLM-assisted: three blowups in a month
  2. The dispute: by his account, OpenAI says it proved it too
  3. Two offers, one hard line, and what he isn't claiming
01

LLM-assisted: three blowups in a month

He and Alpöge release three finite-time blowups with smooth forcing, 3D Euler included. The program belongs to Córdoba and Martínez-Zoroa; LLMs pushed it to completion, verified in Lean on August 22.

Today, Levent Alpöge and I have made public three results: finite-time blowup with smooth forcing for incompressible porous media, for Boussinesq, and for 3d incompressible Euler.

We believe we also have blowup for hypo-dissipative Navier-Stokes. We are not releasing that paper today: unlike the above, the Lean verification has not yet finished. We do not yet have anything resembling a presentable writeup. I mention it because it is suggestive of a path to unforced Euler.

The program this fits into was not started by us nor was it proposed by a Large Language Model. The credit for the basic idea of this program goes to Diego Córdoba and Luis Martínez-Zoroa, who for several years have been exploring the construction of forced blow ups. We took their work as a starting point, using Large Language Models to push their program to completion.

Concretely, what Levent and I did was to take the Córdoba and Martínez-Zoroa program, which achieved blowup results with rough forcing, and, with a great deal of help from LLMs, push it to smooth forcing and to the incompressible Euler equations. The ideas making this line of attack possible are due to Córdoba and Martínez-Zoroa. Let me make plain what I have said to colleagues in private: in view of this body of work, I believe Luis Martínez-Zoroa deserves a Fields Medal.

My work with Levent has been a purely personal collaboration, free of any institutional agreements or official involvement by either of our employers. We used several LLMs throughout: Anthropic’s Claude, OpenAI’s Codex, especially with GPT-5.6 Sol and, more recently, Astra. The latter was only used for writeups and auditing our arguments.

For most of the past year progress was slow. We worked through the literature and upgraded various preliminary results, up to obtaining finite time blow up for the Incompressible Porous Media equation (with smooth forcing). This was until about a month ago, when we had real progress: on August 15th, we obtained the blow up results, with smooth forcing, for both Boussinesq and Euler. I can say the first LLM generated proof Levent sent me was the most horrendous I have ever read; we verified it on Lean on August 22nd. Since this point, we have been working around the clock to understand this proof and turn it into something readable.

There is another part of this story, and one that, honestly, I very much wish I did not have to be concerned with. I am not happy about the presentation quality in these papers. Ideally, we would have preferred to spend weeks turning the LLM generated proofs into something readable from the very first page. This level of care is what these problems and the community devoted to these problems deserves. The various Boussinesq and Euler write-ups in particular are much closer to what models produce under human direction than to a paper written by a person. The Euler writeup, in particular, can only be described as AI slop. I am sorry for this. The reasons are below, and they involve our being pressured by outside factors. I say this not as an excuse but as an explanation.

I had planned to say on announcing our work that the results are not the important thing. Rather the important thing is instead the significance that a mathematician and an LLM model can now do all this work in a month. The significance of this with respect to the way we train students, assign credit, referee, and decide what is worth one human life’s attention cannot be understated. This is a Deep Blue-Kasparov moment. The community needs to have serious and unhurried discussion about where to go from here. Instead of these incredibly important developments, I find myself writing about something else. I want to set out what happened as plainly as I can.

02

The dispute: by his account, OpenAI says it proved it too

By his account, as rumors spread he contacted OpenAI; on calls he was told an internal model had proved forced Navier-Stokes blowup, first with “very little human input”, a story that shifted as the call went on.

On Thursday, September 3rd, with a rumor circulating that Anthropic had resolved a major open problem, and with Levent having received tips that information about our progress had been passed to OpenAI, I wrote to a prominent mathematician at OpenAI. I am quoting my email in full because I would rather the full text be read rather than my summary of it:

“Hi [...], We have not met in person, but I was one of the speakers at the [...]. I am writing because a rumor seems to be spreading quickly that Anthropic has resolved a major open problem. I cannot be certain I am the person it attaches to, but a colleague at Courant emailed me about it yesterday – who heard it from an analyst in the UK, who had it from somewhere further upstream – so it seems safe to assume I am. I gather versions linking Levent Alpöge to a solution are going around in tech as well. There however appears to be a lot of confusion with regards to the exact problem solved.

I should also emphasize that this is not an institutional effort. It is a strictly personal collaboration between the two of us, and there is no formal agreement behind it. I pay for the tools my group uses out of my own research funds, including footing a large bill to OpenAI. I have had an industry collaboration before, with DeepMind, which had a formal institutional arrangement.

We do have work in this area that we are confident in, and we will post it shortly, the paper and the formalization together. We intentionally decided against rushing out a Lean certificate alongside an unpolished preprint. I feel strongly that the first thing anyone reads should be a mathematical argument presented in the normal manner, rather than just a formal certificate.

I am writing to you directly rather than saying anything publicly so that you have the facts to address this on your end. Best wishes, Tristan”

He replied the same day: “If you are willing to give any details it would be useful to avoid competing here and in general we are always thrilled when mathematician make progress with our models. Additionally if there is anything in terms of compute from OpenAI’s end we would be happy to provide it.”

I asked to speak the following week. On Friday, September 4th, I was asked whether I could meet that day; I again said the following week. At 12:45 on Sunday, September 6th, I was asked whether I could meet “at any point today.” Sebastien Bubeck joined. The three of us spoke twice that afternoon. Levent was not on the calls.

I was told that an internal OpenAI model had produced a proof of finite time blowup for the forced Navier-Stokes equations. When Levent asked by text for the precise statement, the answer was: “Existence of forced blowup in R³ and T³,” with “the forcing function is smooth option c and d in Fefferman.” I was told the proof is about 100 pages. I have not seen it.

I should say here why I interpreted their statement the way I did, the interpretation I will discuss below. The route to the Clay problem through a smooth force, options c and d in Fefferman’s statement of the problem, is the route Luis and Diego opened and the one Levent and I had quietly chosen to attack. Almost nobody else I know of was working on it. It is not the direction one arrives at in a few days by giving a model the problem statement. When I heard “forced,” it was a bright red flag.

I was shown a prompt and told the internal research model had simply been given the problem statement. Levent had been told by Sebastien “very little human input” had been used. This turned out not to be true. Over the course of the call, as members of their team sent Sebastien corrections and details over their internal chat, it emerged that an entire team had been working on the problem, that this was one of a number of things that was tried, that work had started on the unforced problem, that the team first set the model on easier problems, including Euler, that even the prompt that had been shown to me had been written by prompting Codex, and that an insane amount of compute had been used.

I asked when the first prompt had been sent by them. This question was not answered directly by OpenAI for some time. Eventually it was agreed that it had been sent in the past few days, after information about our work had reached OpenAI.

I asked whether the model had been trained on, or had access to, our sessions in Codex, into which we had been putting all our drafts for the whole of this project. I was told the model did not look up user data. I asked again, about training, and I did not get an answer.

03

Two offers, one hard line, and what he isn't claiming

By his account, he was offered staggered releases, or a paper under his name alone crediting OpenAI's model; he declined both. He stresses he hasn't seen their proof and accuses no one; he states only what he was told.

Two proposals were offered to me. The first was that we post our Euler result, and that OpenAI post its Navier-Stokes result the next day. The second was that, after posting Euler, I alone write a paper presenting the Navier-Stokes result, acknowledging that an internal OpenAI model had resolved it. Sebastien twice asserted that he wanted Levent removed from authorship, and said it would all be simple if only it were not the case that, and it was so annoying that, Levent works at Anthropic. It was also said that if OpenAI posted after us, they would say that we deserved the Clay Prize, and that we were the “closest humans to the problem”. I declined both offers.

I said that if OpenAI released its result in the way proposed I would go public with what happened. The reply was, “Why would you ruin your career?” I replied that I am an academic, and asked why he thought going public would ruin my career. The reply was, “If you don’t want me to be nice, then I don’t have to be nice.”

Some time later Levent received a text proposing that he and Sebastien speak one on one, saying, “I don’t know if Tristan is being fully rational right now.” Levent declined and said conversations should be with me. Sebastien sent a follow-up email that night requesting to speak with me on Monday, September 7th. I did not respond.

I would like to be clear about what I am not claiming. I have not seen OpenAI’s proof. I do not know what their model did, or how. I do not know whether our data was used. I am not accusing anyone of anything. I am stating what I was told, when, and what was proposed to me. I am stating it because the alternative is to let a sequence of announcements say something I know to be false.

If indeed an OpenAI model did close the gap to Navier-Stokes, that is a remarkable thing and it should be said loudly, by them, with the history intact. I would much rather be talking about mathematics, Luis and Diego’s ideas, and what this all means for the rest of us.

Lastly, I would like to thank the entire mathematics community that have been so supportive of me over the last 24 hours.

Where Indigo landsFurther

Indigo's conclusion

The math layer is the hardest hit yet for the verifiable-domains thread: an LLM helped produce a blowup for 3D Euler, a famous hard problem, verified in Lean. The dispute layer is a weighty first-hand account, but only one side is present, so read it as an account the other side hasn't answered.

What to remember

  1. A hard milestone on the generation side: LLM-assisted new blowups for fluid equations, 3D Euler included, verified in Lean; one mathematician and one LLM, in a month.
  2. A mechanical final check makes “trust first, understand later” work: the first proof was unreadable, but Lean verified it first and people understood it later.
  3. Credit, refereeing and training students are all shaken: he calls for serious, unhurried discussion of what deserves a human lifetime's attention.
  4. A dispute the other side hasn't answered: by his account, OpenAI's story shifted during the call, with pressure to drop a co-author. One side only.
  5. Honesty about credit is a strong signal: the program credited to Córdoba and Martínez-Zoroa, his own write-up called AI slop, and an explicit “I am not accusing anyone”.

Back on the long-running theses

confirms

Can verifiable domains generalize The hardest hit yet on the generation side: new results on a famous PDE problem, verified in Lean. Genuinely new mathematics.

confirms

Verification does not compress The first proof was horrendous and understood only after Lean verified it: where a mechanical final check exists, verification can be carried and generation can be a black box.

adds to

Hiring is the roadmap If accurate, a heavy but unverified data point on the intensity and norms of the frontier race: rushing to publish, pressure over authorship, sensitivity across companies.

adds to

Anthropic: Claude formalizes Fermat's Last Theorem in Lean That piece is the verification side, this one the generation side; together both ends are covered where Lean exists, with Lean as the final judge.

confirms

Sabine Hossenfelder: AI is taking over physics Buckmaster is Sabine's mathematician facing AI, with the crisis made concrete in credit, refereeing and training students.

adds to

Jakub Pachocki, An Alien Mind Pachocki says AI-generated math will flood in and verification is the trust layer; this statement is that future on the ground, pressure on conduct included.

What it means for Rewired Index

Both readings need care. It's evidence of what AI can do in frontier research, which supports the AI-for-science story and the evaluation and formalization toolchain; and if the account holds, it's a negative for OpenAI's reputation and governance. No direct name to point at.

What would change my mind

the paper and the Lean formalization go public and independent review overturns these blowup results.

Finished. Indigo's take on this piece is in two places: