deniz.in

Markets

Weather

Loading weather

· via Hacker News – Front Page (native)

Claimed Navier–Stokes blow-up proof arrives with 600,000-line Lean 4 formalization

A 79-page paper by L. Alpöge and T. Buckmaster claims finite-time blow-up for the Navier–Stokes equations, and ships with a Lean 4 formalization of roughly 600,000 lines.

Claimed Navier–Stokes blow-up proof arrives with 600,000-line Lean 4 formalization

A claimed Millennium-scale result, with a machine-checked skeleton

Mathematicians L. Alpöge and T. Buckmaster have released a 79-page paper, Finite-time blow-up for the Navier–Stokes equations, together with a Lean 4 formalization of its central results. The repository reached the front page of Hacker News on 20 November 2026. If the claim holds, it settles the Clay Mathematics Institute's Millennium Prize problem on Navier–Stokes existence and smoothness in the blow-up direction: solutions of the three-dimensional equations, started from smooth initial data, need not remain smooth for all time.

What the release contains

The GitHub repository holds the paper, dated 20 November 2026, and a Lean package built with Lean v4.33.1 against a pinned Mathlib revision. According to the README, the formalized statements correspond to Theorem 1.1, Theorem 7.1, Corollary 7.2 and Appendix B of the paper, and the package documents where the formal proofs diverge from the written ones. A separate notice file describes in prose what each statement file contains and what is actually proved.

How the formalization is structured

The package separates claims from evidence. Files at the top of the lean directory define the objects and state the theorems with sorry placeholders; a single Solution.lean then discharges those statements using a library under lean/Breakdown. Machine-readable comparator files pair each statement file with the solution, name the theorems whose statements must agree, and allow only three axioms: propext, Quot.sound and Classical.choice, the standard axioms of Lean's core library. The design is meant to let an outsider recompile everything and confirm that what was proved matches what was claimed, with no smuggled assumptions. The cost is scale. The library spans roughly 2,270 files and 600,000 lines, and the README warns that a full build demands substantial memory and considerable time.

What is missing from the repository

The formalization alone does not carry the result from end to end. The README states that the formalization contains no data: a numerical certificate, identified by a hash in the paper, is absent from the repository, and the paper promises a later release of it. Verifying that the certificate satisfies the conditions of Appendix B falls to a rigorous computation using interval arithmetic at 128 bits of precision, with the Arb library accessed through FLINT. That computation is neither part of the formalization nor included in the release. The paper's own source files are also not in the repository.

A patch that has not been rebuilt

One further caveat sits in the fine print. The Lean package is the authors' working version with a patch of 20 November 2026 applied to the cell-layer statements and to the library proofs that depend on them. The patch changes the weight of one bound from m² + 1/4 to (|m| + 1/2)², allows one constant of a cell to be zero, and restricts the harmonics of the model in a cell to those that do not vanish there. Appendix B of the paper states the conditions in this patched form, but the README acknowledges that the patched package was not rebuilt when the repository was assembled, so the released artifact may not yet correspond exactly to the published statements.

Why it matters

The Navier–Stokes existence and smoothness question is one of the seven Millennium Prize Problems, each carrying a $1 million reward, and a confirmed construction of finite-time blow-up would close one of the most storied open problems in mathematics. Beyond the prize, the release is a milestone for computer-verified proof: a proposed resolution of a flagship problem now arrives with a machine-checkable development attached, extending proof assistants' reach from textbook theorems toward the research frontier. The caveats matter just as much. The numerical certificate and the interval-arithmetic computation that anchors it are not yet public, the final patched build has not been run, and Clay's rules require journal publication and roughly two years of community acceptance before any award. The repository makes independent verification possible; it does not yet complete it.

  • #lean
  • #formal-verification
  • #navier-stokes
  • #mathlib
  • #millennium-prize

Related posts