deniz.in

Markets

Weather

Loading weather

· via Hacker News – Front Page (native)

AI-assisted Lean proofs claim 3D Euler and Navier-Stokes blowup; GRH claims emerge

Terence Tao reports AI-assisted, Lean-formalized proofs of finite-time blowup for 3D Euler and Navier-Stokes, plus a machine-generated GRH proof — and a separate human GRH draft surfaced the same day.

AI-assisted Lean proofs claim 3D Euler and Navier-Stokes blowup; GRH claims emerge

On November 20, 2026, two announcements landed on the same day that may mark a turning point in how mathematics is produced. A post on Terence Tao's blog relayed three results obtained with extensive AI assistance: proofs of finite-time blowup for the three-dimensional Euler and Navier–Stokes equations, both formalized in the Lean proof assistant, and an enormous machine-generated Lean proof of the Generalized Riemann Hypothesis. Separately, a hurried announcement on the math216 blog released a human-written draft proof of GRH whose author says it involved no AI at all.

Two blowup results, formalized in Lean

According to Tao's post, which ordered the three results by how settled they are, the most mature is a human-readable proof by Levent Alpöge, Tristan Buckmaster and Matei Coiculescu that the unforced, incompressible Euler equations in three dimensions admit solutions that blow up in finite time. The proof comes with a Lean formalization, and Tao pointed readers to a Mastodon post by Buckmaster covering the two fluid-dynamics results.

Close behind is a companion claim by Alpöge and Buckmaster: finite-time blowup for the unforced, incompressible Navier–Stokes equations in three dimensions, likewise produced with heavy AI usage and likewise formalized in Lean. Tao described the paper as technically human-readable, with a broadly clear strategy — numerically construct an approximate discretely self-similar solution, then apply rigorous perturbation theory to convert it into an exact one — a route specialists had anticipated. The write-up still needs work before it becomes a properly polished professional proof, he noted.

A machine-generated GRH proof

The third item is stranger. Alpöge, working with an internal AI model, generated a vast Lean-only proof of the Generalized Riemann Hypothesis: there is no conventional paper, only the formalization. Verification efforts are concentrated on a Lean Zulip thread, where participants are also trying to extract the underlying strategy. Tao wrote that very initial analysis suggests the proof draws substantially on modern developments in the Langlands program and arithmetic geometry, and that digesting it will take time — though at least there is a human author available to answer questions.

A second GRH claim, this one human

Also dated November 20 is a hastily written announcement on the math216 blog. Its author, posting between duties at an American Mathematical Society meeting, released a draft GRH proof by Alessandro Masullo. The post states the work is entirely human, with no AI involved. According to the announcement, Brian Conrad has been examining the manuscript closely for months, and his extensive feedback is being incorporated; so far the work has held up under that scrutiny. The original plan was to bring Masullo to Stanford for a year to polish the writing before announcing, but the author decided to publish the current draft immediately and asked the community to check it, formalize it and fill any small gaps.

What has actually been verified

The verification status of the four claims varies considerably. The Euler and Navier–Stokes results are described as formalized in Lean, meaning machine-checked artifacts exist alongside the papers, even though the Navier–Stokes exposition still needs polishing. The two GRH claims are both unverified: Alpöge's Lean proof is under active checking, and Masullo's draft has only just been released. The two GRH efforts are also methodologically unrelated — one generated inside a proof assistant by an AI model, the other written by hand and vetted by a leading expert — so their appearance on the same day should not be read as corroboration of either.

Why it matters

The Navier–Stokes existence-and-smoothness question is one of the Clay Mathematics Institute's Millennium Prize Problems; a confirmed finite-time blowup for the unforced three-dimensional equations would resolve it in the blowup direction, and the Euler result would settle a major open problem in its own right.

The larger story is the method. If these results hold, landmark theorems are now being produced by small teams relying heavily on AI, with correctness anchored not in human reading alone but in proof-assistant formalization. That shifts the bottleneck from discovery to verification: an AI-generated GRH proof can apparently be produced faster than the community can digest it. The same-day arrival of a fully human GRH draft is a reminder that traditional workflows continue — but both now funnel into the same slow, careful checking that mathematics has always required.

  • #mathematics
  • #lean
  • #theorem-proving
  • #navier-stokes
  • #riemann-hypothesis

Related posts