brain/
← all entities
entitygenericartificial-intelligence

Tristan Buckmaster

Notes

Vintage: 2026-09. Primary evidence is Buckmaster’s fetched statement PDF (2026-09-08) as synthesized in 2026-09-08-x-morning-buckmaster-alpoge-ai-fluid-proofs-openai-credit. X amplifiers are not substitutes for the PDF. Not a diarized source — claims are source-attributed.

Tristan Buckmaster

One-line summary: NYU mathematician; with levent-alpoge, announced Lean-checked finite-time blowups with smooth forcing for incompressible porous media, Boussinesq, and 3D Euler (8 Sep 2026). Same statement alleges authorship pressure from sebastien-bubeck — contested; see openai-fluid-proofs-credit-dispute.

What it is

A named mathematician (CIMS / NYU; statement hosted at cims.nyu.edu/~tristanb/statement.pdf). This page exists because the PDF is the load-bearing primary for both the math announcement and the OpenAI-credit allegations. Full math treatment: buckmaster-alpoge-forced-blowups. Full dispute treatment: openai-fluid-proofs-credit-dispute.

Why it matters to this thread

AI-assisted formal math and lab-credit conflict are in-scope. Buckmaster’s writeup names Claude, OpenAI Codex (esp. GPT-5.6 Sol), and Astra as tools, and claims a personal collaboration “free of any institutional agreements” with either employer.

Key facts (from 2026-09-08-x-morning-buckmaster-alpoge-ai-fluid-proofs-openai-credit)

  • From 2026-09-08-x-morning-buckmaster-alpoge-ai-fluid-proofs-openai-credit (statement PDF, fetched 2026-09-08): with levent-alpoge, public today are three results — finite-time blowup with smooth forcing for incompressible porous media, Boussinesq, and 3D incompressible Euler. Companion PDFs: euler.pdf, ipm.pdf, boussinesq.pdf. Lean repo: github.com/tristanbuckmaster/fluid_lean.
  • From the same source (same PDF): they believe they also have hypo-dissipative Navier–Stokes blowup but are not releasing that paper today (Lean unfinished; no presentable writeup).
  • From the same source (same PDF): program’s basic idea credited to Diego Córdoba and Luis Martínez-Zoroa (rough forcing → they + LLMs pushed to smooth forcing / Euler). Personal collaboration only.
  • From the same source (same PDF): LLMs used — Anthropic Claude, OpenAI Codex (esp. GPT-5.6 Sol), Astra “only used for writeups and auditing.” Timeline claimed: Aug 15 blowup results for Boussinesq and Euler; Lean verify of a first LLM proof Aug 22.
  • From the same source (@deaneyang, @AndrewCurran_): X amplifiers of the PDF/GitHub set — not substitutes for the PDF.

What this source does not establish

  • Forced smooth blowup ≠ Clay Millennium unforced Navier–Stokes. Keep “Millennium adjacent / forced variant” language. See buckmaster-alpoge-forced-blowups.
  • No independent re-parse of euler.pdf / ipm.pdf / boussinesq.pdf page-by-page in this clip.
  • No person pages for Córdoba, Martínez-Zoroa, Dean Yang, or Andrew Curran.
  • Not speaker-aware. No ## Said quotes table.
  • Allegations vs Bubeck live on openai-fluid-proofs-credit-dispute — do not flatten here.

Afternoon OpenAI concurrent-work (from 2026-09-08-x-afternoon-openai-navier-stokes-images-2-5)

Sources

Related

Referenced by