# Will Anthropic announce Claude has solved Navier-Stokes before July 2027?

Canonical URL: https://preseen.com/reports/c3001e1c-6455-49c4-a863-ca58c9950b12/will-anthropic-announce-claude-has-solved-navier-stokes-before-july-2027
Markdown URL: https://preseen.com/reports/c3001e1c-6455-49c4-a863-ca58c9950b12/markdown

## Forecast

P(Yes): 28.7%; P(No): 71.3%.

Generated: September 8, 2026 at 2:58 PM UTC
Forecast model: gpt-5.6-sol
Research model: gpt-5.6-sol

## Analysis

## TL;DR
I estimate a **29%** chance of YES. At the September 8, 2026 cutoff, the official channels checked contained no qualifying Anthropic claim, and the public Claude-assisted Lean artifact concerned forced Euler, not ordinary-viscosity Navier–Stokes ([Anthropic research index](https://www.anthropic.com/research); [Euler Lean artifact](https://github.com/tristanbuckmaster/fluid_lean/blob/d0124689230b58b4f86e7b90ac59de06404b3b6b/euler-blowup/README.md)). ([github.com](https://github.com/tristanbuckmaster/fluid_lean/blob/d0124689230b58b4f86e7b90ac59de06404b3b6b/euler-blowup/README.md)) The live route is a Clay-qualifying smooth-forcing counterexample: Tao says the method has a high likelihood of extending, but the missing viscosity step and the mixed personal attribution keep NO more likely ([Tao's assessment](https://terrytao.wordpress.com/2026/09/07/finite-time-blowup-with-smooth-forcing-term-for-the-incompressible-porous-medium-boussinesq-and-incompressible-euler-equations/); [Buckmaster's statement](https://cims.nyu.edu/~tristanb/statement.pdf)). 

## Context
The resolution bar is lower than a correct, Clay-accepted proof. Anthropic only needs to issue an official statement that substantively says Claude solved the full problem, alone or as a major contributor. Yet the mathematics still must match one of Clay's four formulations. Options C and D allow a smooth external force, so a rigorous finite-time breakdown result for the ordinary three-dimensional viscous equation would qualify ([Clay's official formulation](https://www.claymath.org/library/monographs/MPPc.pdf)). 

At the cutoff, Anthropic's official research list showed its September 4 Fermat formalization as its latest mathematics publication, not a Navier–Stokes announcement ([Anthropic research index](https://www.anthropic.com/research)). The public fluid work was a personal Buckmaster–Alpöge collaboration using Claude and several OpenAI systems, and it stopped short of ordinary Navier–Stokes ([Buckmaster's statement](https://cims.nyu.edu/~tristanb/statement.pdf)). 

## Evidence
The historical prior is tiny. Clay announced seven Millennium Problems in 2000; only the Poincaré conjecture has been resolved, while Navier–Stokes remains listed as unsolved ([Clay's Millennium Problems](https://www.claymath.org/millennium-problems/); [Navier–Stokes page](https://www.claymath.org/millennium/Navier-Stokes-Equation/)). A crude exchangeable rate gives about a 0.4% chance for one problem over a roughly ten-month window. That estimate has one event in its sample and targets accepted solutions, not corporate claims, so I use it only as a starting prior. 

The main update is specific, not generic AI hype. Buckmaster reported that he and Anthropic mathematician Levent Alpöge obtained smooth-forcing blowup results for Boussinesq and three-dimensional Euler on August 15 and finished Lean verification on August 22; he also said an unreleased hypodissipative Navier–Stokes result still lacked finished Lean verification and a presentable write-up ([Buckmaster's statement](https://cims.nyu.edu/~tristanb/statement.pdf)). The public Euler repository says the full Lean code, including the trusted statement, was written by Claude under Alpöge's direction ([Euler Lean artifact](https://github.com/tristanbuckmaster/fluid_lean/blob/d0124689230b58b4f86e7b90ac59de06404b3b6b/euler-blowup/README.md)). Tao then wrote that the approach had a high likelihood of extending to Navier–Stokes and looked feasible in the near future, while stressing that the result was not yet there and that major technical complications remained ([Tao's assessment](https://terrytao.wordpress.com/2026/09/07/finite-time-blowup-with-smooth-forcing-term-for-the-incompressible-porous-medium-boussinesq-and-incompressible-euler-equations/)). ([github.com](https://github.com/tristanbuckmaster/fluid_lean/blob/d0124689230b58b4f86e7b90ac59de06404b3b6b/euler-blowup/README.md))

The alleged OpenAI proof is a weak but relevant signal. Buckmaster says OpenAI personnel told him on September 6 that an internal model had produced about 100 pages proving smooth-forced blowup under Clay options C and D. He also says he never saw the proof and did not know whether it was correct or how it was produced ([Buckmaster's statement](https://cims.nyu.edu/~tristanb/statement.pdf)). I treat this as evidence that another strong team thought the viscosity gap was tractable, not as a 50%-plus probability that a valid proof already exists. 

Claude's broader record cuts both ways. Anthropic's August 10 Riemann-zeta project used 31 million output tokens and about 60 subagents to improve a lower bound from 41.6% to 67.2%, but Anthropic explicitly said Claude did not solve the Riemann hypothesis ([Anthropic's Riemann report](https://www.anthropic.com/research/riemann-zeta)). Its September 4 Fermat project used about six billion output tokens over 11 days to create a 13-million-line Lean formalization of an already known proof ([Anthropic's Fermat report](https://www.anthropic.com/research/formalizing-fermats-last-theorem)). The more direct open-problem benchmarks remain harsh: Claude Fable 5 and 5.1 each solved 0 of 68 curated Erdős problems under one attempt, 72 hours, and a $300 budget per problem, while AI had solved 0 of 6 major-advance and 0 of 3 breakthrough items in Epoch's separate 50-problem collection ([FrontierMath Erdős](https://epoch.ai/latest/announcing-frontiermath-erdos); [FrontierMath Open Problems](https://epoch.ai/frontiermath/open-problems)). 

Anthropic's wording suggests it would announce a well-supported result, but not a loose candidate. Its Riemann post drew a sharp line between progress and solution; its Fermat post called the novelty formalization; and its cryptography post used categorical language that Claude had resolved the Jacobian conjecture while also describing hundreds of hours of human validation for harder-to-check findings ([Riemann report](https://www.anthropic.com/research/riemann-zeta); [Fermat report](https://www.anthropic.com/research/formalizing-fermats-last-theorem); [cryptography report](https://www.anthropic.com/research/discovering-cryptographic-weaknesses)). I assign slightly under one-half to the current technical line reaching an exact ordinary-Navier–Stokes result by the deadline, about three-quarters to Claude being a material contributor if it does, and about two-thirds to Anthropic then using qualifying language. That path contributes about 23 percentage points. Separate Anthropic work adds about 4 points, a mixed-lab major-credit path adds about 1 point, and a premature claim adds about 1 point, giving 29% overall. 

## What's non-obvious
The forcing term is the biggest interpretation edge. Many summaries treat a forced blowup as a side result. Clay's own options C and D make it a full solution if the force, initial data, domain, and breakdown conditions match the official statement ([Clay's official formulation](https://www.claymath.org/library/monographs/MPPc.pdf)). 

The OpenAI rumor does not simply raise the Anthropic forecast. If OpenAI publishes first and owns the decisive argument, it can reduce Anthropic's ability to say Claude solved the problem. The offsetting fact is that the public Euler formalization credits Claude with all Lean code, giving Anthropic a real attribution path if the same collaboration closes the viscous case ([Buckmaster's statement](https://cims.nyu.edu/~tristanb/statement.pdf); [Euler Lean artifact](https://github.com/tristanbuckmaster/fluid_lean/blob/d0124689230b58b4f86e7b90ac59de06404b3b6b/euler-blowup/README.md)). ([github.com](https://github.com/tristanbuckmaster/fluid_lean/blob/d0124689230b58b4f86e7b90ac59de06404b3b6b/euler-blowup/README.md))

## Uncertainties
The forecast rests on four unresolved facts:

- No public proof or Lean artifact for ordinary-viscosity Navier–Stokes existed at the cutoff; the closest public artifact was Euler ([Euler Lean artifact](https://github.com/tristanbuckmaster/fluid_lean/blob/d0124689230b58b4f86e7b90ac59de06404b3b6b/euler-blowup/README.md)). ([github.com](https://github.com/tristanbuckmaster/fluid_lean/blob/d0124689230b58b4f86e7b90ac59de06404b3b6b/euler-blowup/README.md))
- Tao had not fully digested the preliminary work, so his positive assessment is directional evidence, not verification ([Tao's assessment](https://terrytao.wordpress.com/2026/09/07/finite-time-blowup-with-smooth-forcing-term-for-the-incompressible-porous-medium-boussinesq-and-incompressible-euler-equations/)). 
- The reported OpenAI proof was unseen and unreviewed, and its exact match to Clay C or D could not be checked ([Buckmaster's statement](https://cims.nyu.edu/~tristanb/statement.pdf)). 
- The model-by-model contribution is unclear in a personal project that used systems from both companies; an explicit attribution statement from the authors or Anthropic would close this gap ([Buckmaster's statement](https://cims.nyu.edu/~tristanb/statement.pdf)). 

These gaps keep the estimate below one-third despite the unusually strong technical signal.

## Sources

- Domain Expert Search (mcp)
  > Found 14 domain experts for 'Navier-Stokes PDE research, smooth-forced blowup, Clay Millennium formulation, AI-assisted mathematics, and assessing near-term proof claims':
- Domain Expert Research Task (mcp)
  > Job domain_expert_research_task_414e5dc4ba done after 225388ms.
- [https://cims.nyu.edu/~tristanb/statement.pdf](https://cims.nyu.edu/~tristanb/statement.pdf) (openai)
- [terrytao.wordpress.com](https://terrytao.wordpress.com/2026/09/07/finite-time-blowup-with-smooth-forcing-term-for-the-incompressible-porous-medium-boussinesq-and-incompressible-euler-equations) (tool)
- [2  CHARLES L. FEFFERMAN](https://www.claymath.org/wp-content/uploads/2022/06/navierstokes.pdf) (openai)
- [claymath.org](https://www.claymath.org/millennium/navier-stokes-equation) (tool)
- [cims.nyu.edu](https://cims.nyu.edu/~tristanb/euler.pdf) (tool)
- [cims.nyu.edu](https://cims.nyu.edu/~tristanb/boussinesq.pdf) (tool)
- [cims.nyu.edu](https://cims.nyu.edu/~tristanb/ipm.pdf) (tool)
- [arxiv.org](https://arxiv.org/abs/2410.22920) (tool)
- [arxiv.org](https://arxiv.org/abs/2410.22920v3) (tool)
- [terrytao.wordpress.com](https://terrytao.wordpress.com/tag/tristan-buckmaster) (tool)
- [fluid_lean/euler-blowup/README.md at d0124689230b58b4f86e7b90ac59de06404b3b6b · tristanbuckmaster/fluid_lean · GitHub](https://github.com/tristanbuckmaster/fluid_lean/blob/d0124689230b58b4f86e7b90ac59de06404b3b6b/euler-blowup/README.md) (openai)
- [claymath.org](https://www.claymath.org/wp-content/uploads/2022/03/millennium_prize_rules_0.pdf) (tool)
- [Announcing FrontierMath Erdős | Epoch AI](https://epoch.ai/latest/announcing-frontiermath-erdos) (openai)
- [anthropic.com](https://www.anthropic.com/research/riemann-zeta?trk=public_post_comment-text) (tool)
- [Formalizing Fermat's Last Theorem \\ Anthropic](https://www.anthropic.com/research/formalizing-fermats-last-theorem) (openai)
- [Discovering cryptographic weaknesses with Claude \\ Anthropic](https://www.anthropic.com/research/discovering-cryptographic-weaknesses) (openai)
- [claymath.org](https://www.claymath.org/millennium-problems/rules) (tool)
- [Learning more about Claude's mathematical capabilities \\ Anthropic](https://www.anthropic.com/research/riemann-zeta) (openai)
- [www-cdn.anthropic.com](https://www-cdn.anthropic.com/95c246936988e43127bc6b2ceb7077c1dad2d68e.pdf) (tool)
- [github.com](https://github.com/anthropics/formal-math) (tool)
- [github.com](https://github.com/anthropics/zeta-23-lean) (tool)
- [github.com](https://github.com/anthropics/fermats-last-theorem) (tool)
- [FrontierMath Tier 4 (v2) | Epoch AI](https://epoch.ai/benchmarks/frontiermath-tier-4-v2) (openai)
- [epoch.ai](https://epoch.ai/benchmarks/frontiermath-tier-4) (tool)
- [FrontierMath: Open Problems - Unsolved Mathematical Challenges | Epoch AI](https://epoch.ai/frontiermath/open-problems) (openai)
- [epoch.ai](https://epoch.ai/frontiermath/open-problems/about/faq) (tool)
- [anthropic.com](https://www.anthropic.com/news/claude-fable-5-mythos-5?_bhlid=9f604e728be0fc7007ec18cb6f67d4a77cf2acee) (tool)
- [What's new | Updates on my research and expository papers, discussion of open problems, and other maths-related topics. By Terence Tao](https://terrytao.wordpress.com/?l=1) (openai)
- [openai.com](https://openai.com/news/research) (tool)
- [anthropic.com](https://www.anthropic.com/news) (tool)
- [x.com](https://x.com/AndrewCurran_) (tool)
- [site.twstalker.com](https://site.twstalker.com/AndrewCurran_) (tool)
- [mastodon.social](https://mastodon.social/@tristanbuckmaster/117233413705701198) (tool)
- [lobste.rs](https://lobste.rs/s/b7k94m/statement_on_finite_time_blowup_for) (tool)
- [x.com](https://x.com/SebastienBubeck/status/2097214122471432349) (tool)
- [x.com](https://x.com/__alpoge__) (tool)
- [zamantika.com](https://zamantika.com/tl/profile/__alpoge__) (tool)
- [axios.com](https://www.axios.com/2026/09/08/ai-math-anthropic-openai-google) (tool)
- [openai.com](https://openai.com/index/ten-advances-in-mathematics) (tool)
- [twstalker.com](https://twstalker.com/_sholtodouglas) (tool)
- AskNews (mcp)
  > Found 10 articles:
- Perigon (mcp)
  > {"total_results": 3, "returned": 3, "articles": [{"id": "9e45ee96e125451caab2d000df70cc15", "title": "Big AI broke the internet. Now it wants everyone else to fix it.", "source": "theblaze.com", "date": "2026-09-04", "url": "https://www.theblaze.com/tech/big-ai-open-letter-warning", "summary": "Big names in AI, including OpenAI, Anthropic, Google, Microsoft, and Perplexity, have issued a joint letter warning of AI-enabled cyber attacks becoming more prevalent. They want corporations, cybersecur…
- [theblaze.com](https://www.theblaze.com/tech/big-ai-open-letter-warning) (tool)
- [futurism.com](https://futurism.com/science-energy/obama-admin-official-stokes-agi-china-data-centers-military-strikes) (tool)
- [thenextweb.com](https://thenextweb.com/news/us-china-ai-talks-cyberattacks-trump-xi-summit-september) (tool)
- Epoch (mcp)
  > 85 Epoch capability benchmarks available. Pass one as benchmark to get scores over time:
- [epoch.ai](https://epoch.ai/data/benchmark_data.zip) (tool)
- Metr (mcp)
  > METR time-horizon trend (50% success, invsqrt_task_weight weighting, report 1-1)

## Question Details

This question asks whether Anthropic will publicly announce, before July 1, 2027, that Claude has solved the Clay Mathematics Institute's three-dimensional Navier–Stokes existence and smoothness Millennium Prize Problem. As of September 6, 2026, Clay continues to describe Navier–Stokes as unsolved; its official formulation permits a resolution by proving global existence and smoothness in the specified 3D settings or by establishing an allowed breakdown/counterexample. (claymath.org) The threshold here is an Anthropic announcement of a claimed solution, not acceptance by the Clay Mathematics Institute or the broader mathematical community. This distinction matters because Anthropic is already publicly reporting significant Claude-assisted mathematics: on August 10, 2026 it announced that a research version of Claude improved a longstanding bound related to the Riemann hypothesis while explicitly saying Claude had not solved the Riemann hypothesis itself, and on September 4, 2026 it announced a Claude-produced computer-checked formalization of Fermat's Last Theorem. (anthropic.com)

### Resolution Criteria

Resolve YES if, on or before June 30, 2027, Anthropic publishes an official public statement that substantively claims that Claude, including any research or unreleased version of Claude, has solved the Clay Mathematics Institute's Navier–Stokes existence and smoothness Millennium Prize Problem. The announcement may say Claude solved the problem autonomously, collaboratively with humans, or as a major contributor; it need not claim that Claude was the sole author of the solution. A qualifying statement must claim a solution to the full Millennium Prize Problem, rather than merely progress on Navier–Stokes, a solution to a restricted/special case, a numerical solution of particular Navier–Stokes equations, or a formalization of previously known mathematics. Either an existence-and-smoothness proof or a valid breakdown/counterexample of the kind permitted by the official Clay formulation counts as 'solved.' (claymath.org) For determining whether Anthropic made the announcement, use official Anthropic-controlled public channels, principally anthropic.com (including its Research/Science/News publications). An official Anthropic social-media post or other official Anthropic publication also qualifies if it unambiguously makes the claim. Third-party reporting, statements made solely by individual Anthropic employees in a personal capacity, leaked material, or a mathematical paper without an accompanying official Anthropic claim do not by themselves qualify. Resolve NO if no qualifying Anthropic announcement has been publicly made by the end of June 30, 2027. Subsequent rejection, retraction, discovery of an error, or failure to receive the Clay Millennium Prize does not change a YES resolution, because the question concerns whether Anthropic announces that Claude has solved the problem rather than whether the claimed solution is ultimately accepted as correct.

### Fine Print

The deadline is interpreted as 'before July 2027,' meaning publication no later than June 30, 2027, regardless of the precise time of day or timezone shown by the publication platform. The relevant Navier–Stokes problem is specifically the Clay Millennium Prize Problem, not the broader task of solving Navier–Stokes equations in practical or numerical settings. As of September 6, 2026, Clay's Millennium Prize Problems page still presents Navier–Stokes as unresolved. (claymath.org) If Anthropic uses language such as 'candidate solution' or reports that Claude has produced a proof that remains under evaluation without itself asserting that the problem has been solved, this does not qualify. Conversely, wording need not literally contain 'Claude has solved Navier–Stokes' if the announcement clearly and substantively makes that claim.