AI & Maths watch GitHub

Retrospective research digest

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

At a glance1 entry · 2 linked sources Partial coverage

One dated update concerns the provenance of AI-assisted fluid-equation research. No new mathematical result or independently checked formalization was established for September 9 in the accessible sources. Archival and account-verification gaps limit this reconstruction.

Research updatesSelect an entry to read more

Alpöge’s account adds a provenance update to the AI-assisted fluid-equation dispute Julien Lausson, reporting statements by Levent Alpöge Published: 2 sources

Numerama’s article, dated September 9 at 11:21, reports Alpöge’s account of the discussions preceding OpenAI’s fluid-equation announcement. He describes his collaboration with Tristan Buckmaster as personal mathematical research using several AI tools, and alleges pressure concerning his participation in a proposed publication. The article also records that OpenAI representatives disputed the surrounding accusations. This updates the account of research provenance and collaboration; it supplies no new blowup theorem or proof validation.

Research relevance:

Relevant to mathematicians using commercial AI systems on unpublished work: the dispute raises concrete questions about recording contributions, communicating concurrent results and agreeing authorship across institutional boundaries.

Claim status:
Dated reporting of contested participant allegations. Neither access to unpublished research nor improper appropriation is established by the material consulted. This is not independent mathematical scrutiny or a checked formalization.
Limitations:
The original X statements were inaccessible, and their dates are inconsistent across secondary sources; published_date refers to the original Numerama article. Its mathematical framing requires correction: Clay’s alternatives C and D explicitly permit smooth forcing, contrary to the article’s suggestion that only an unforced result could address the prize problem. That correction does not establish that any submitted proof satisfies those alternatives.

Sources & further reading

Link to this entry
Coverage and limitations
  • Public web searches reconstructed September 9, 2026, in Paris; indexing and crawling dates were not treated as publication dates.
  • Read Terence Tao’s September archive, the Xena homepage, Harvard CMSA’s Daniel Litt event page, and the official websites of Epoch AI, Harmonic and Axiom. These did not establish an eligible new research announcement published that day.
  • The CMSA page dates Litt’s talk to September 9, but does not establish its announcement’s publication date; it was excluded.
  • Searches using the supplied researchers and organizations returned uneven results. Official academic pages provided some identity context, but did not authenticate every supplied social handle. No exhaustive X access is claimed; Mathstodon and Bluesky archives were not consulted.
  • Read Numerama’s September 9 reporting and its embedded excerpts from Alpöge’s statement. The linked original X posts could not be retrieved. Third-party social archives showed conflicting September 8 and September 9 dates and were not used to establish the original post date.
  • Lean’s homepage, the attempted Project Numina page, Xena’s September archive and attempted Alpöge academic pages returned retrieval errors. Elzo Brito’s September 9 formalization commentary was visible only as an indexed snippet; repeated attempts to read it failed, so it was excluded.
  • Excluded repetitions of the September 7–8 fluid-equation announcements and Lean runtime issue. Also excluded a September 9 derivative discussion of Epoch’s September 1 benchmark, and Tao posts published after the target day.
  • Read Clay’s original Navier–Stokes problem statement to check the distinction between forced and unforced alternatives. No proof was verified and no Lean code was executed.

Requested coverage window: 2026-09-09T00:00:00+02:00 — 2026-09-10T00:00:00+02:00.

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

Archives 32
Older editions (2)