Archived edition

Updates on artificial intelligence for mathematical research, proof discovery and formal verification, with links to papers, code and tools.

Research updatesSelect an entry to read more

1 entry · 5 linked sources News found
Armstrong and Bou-Rabee publish a claimed full-range Lean formalization of Deng-Hani's long-time wave-turbence theorem, with a documented repair below gamma = 1/2 Scott Armstrong and Ahmed Bou-Rabee Published: 5 sources

A public Lean 4 repository accompanies Deng and Hani's existing result on the long-time justification of wave turbulence for cubic NLS on a large square torus. Its README claims that the formalized main theorem covers 0 < gamma < 1, that the statement was frozen before proof construction, and that the sealed proof was replayed in Lean by two separate project audit sessions. Of mathematical interest, the repository says the printed argument's counting proposition fails for gamma <= 1/2; it supplies a different 'vine' family and a corrected counting step as a repair. This is a formalization and repair claim for an existing result, not a newly announced theorem.

Research relevance:

This is a concrete research workflow artifact for long analytic/PDE proofs: frozen target statements, source-linked errata, explicit deviation records, no-sorry/axiom checks, and a public dependency-heavy Lean codebase. The claimed lower-gamma repair makes it more consequential than a transcription alone.

Claim status:
Public source code and project documentation are available. The original Deng-Hani paper is available. The repository claims Lean kernel checking, no added axioms beyond Lean's standard axioms, and two project-run audit replays; those claims have not been independently reproduced here. No independent human scrutiny of the asserted repair was located.
Limitations:
The announced date comes from the supplied direct X API post, not a dated release note. The underlying paper is from 2023/2024. I did not run Lean or inspect the approximately 1.8 million claimed lines of code. The repository itself notes a scope deviation at gamma = 1: its formal result is on the square torus, whereas the paper treats generic rectangular tori in that endpoint case. Its audits were conducted by AI sessions selected by the project, so they are not independent peer review.

Sources & further reading

Coverage and limitations
  • Time window used: 2026-11-09 18:15 Europe/Paris to 2026-11-11 06:15 Europe/Paris. The item is dated 2026-11-10 from Scott Armstrong's supplied direct X API post; its underlying mathematical paper was first posted in 2023 and revised in 2024.
  • Direct X collection: API results were available for all 14 supplied X handles, with bounded pagination; replies and reposts were excluded and cached posts may be included. Only Scott Armstrong's 2026-11-10 posts led to a substantive public mathematical artifact. The accounts themselves were not treated as identity proof; Armstrong's academic identity was checked against NYU's official faculty page. The remaining collected posts were non-substantive for this digest or linked only to unavailable/insufficiently verified material.
  • Direct social collection: Mathstodon collection was unavailable because of an HTTPError; the stated cap was 300 posts, with replies and reposts excluded. The Kevin Buzzard/Xena Bluesky collection succeeded but returned no posts, also capped at 300 and excluding replies and reposts. Empty results do not establish inactivity.
  • Primary sources read: the public GitHub repository and its README, the Deng-Hani arXiv record, and the repository's linked errata/repair documentation. I did not execute Lean, build the repository, inspect every source file, or independently validate correspondence between all Lean statements and the paper.
  • Related sources consulted: Terence Tao's November archive showed no new relevant post in the requested interval; Simon Frieder's blog showed no 2026 post in the interval. A search also surfaced several papers posted on 2026-11-07/08, but they predate the requested window and are therefore excluded. The repository's GitHub commits page could not be fetched from the public web cache.
  • No independent human peer review or third-party replay of the new WaveTurbulence development was found in the sources consulted. The repository reports audits by separate AI sessions; this is useful project documentation, not independent scrutiny.

Requested coverage window: 2026-11-09T18:15:01.780306+01:00 — 2026-11-11T06:15:01.244783+01:00.

Public web research; coverage is not exhaustive. A source link is not a certification of a claim.

Archives 79
Older editions (49)