OpenAI unforced Euler blowup (issuer side-result)
Vintage: 2026-09. Issuer side-result dated September 8, 2026. Evidence in 2026-09-08-x-afternoon-openai-navier-stokes-images-2-5 (euler.pdf linked from the NS issuer page). Unforced Euler ≠ Clay unforced Navier–Stokes. Also ≠ Buckmaster–Alpöge forced Euler and ≠ Anandkumar’s PINN candidate.
OpenAI unforced Euler blowup (issuer side-result)
One-line summary: In the same 8 Sep 2026 NS post, openai says agents resolved an unforced Euler blowup as an “easier” related problem (~100 agents, ~50 hours), with a separate euler.pdf. Issuer claim; not independently reviewed in this clip.
The insight
This is the third Euler lane in the 8 Sep window. Keep it distinct from buckmaster-alpoge-forced-blowups (smooth forcing) and from anandkumar-unforced-euler-candidate (PINN candidate / stability still lacking per Tao-via-Anima). Do not collapse into openai-forced-navier-stokes-claim (forced NS C/D) or into “Navier–Stokes solved” discourse.
OpenAI’s concurrent-work section (same clip) says proofs differ: Buckmaster–Alpöge Euler is forced; this OpenAI Euler is unforced. That is OpenAI’s framing — keep it attributed.
Evidence
All bullets are from 2026-09-08-x-afternoon-openai-navier-stokes-images-2-5.
- From 2026-09-08-x-afternoon-openai-navier-stokes-images-2-5 (On the Navier–Stokes Millennium Prize Problem): agents resolved an unforced Euler blowup as an “easier” related problem (~100 agents, ~50 hours) — separate euler.pdf; Lean/repo github.com/openai/NavierStokesAndEuler.
- From the same source (issuer Concurrent work + @OpenAI): OpenAI congratulates levent-alpoge and tristan-buckmaster for forced-Euler priority; says proofs differ (their Euler forced vs OpenAI unforced).
What this source does not establish
- Not Clay unforced Navier–Stokes.
- euler.pdf was not re-parsed page-by-page in this clip (linked from issuer; existence + title).
- Not a confirmation that Anandkumar’s PINN candidate is the same object — different authors, method, and grain (candidate vs issuer writeup).
- Not independently reviewed in this pass.
- No ticker.
Contradictions / tensions
- Three Euler lanes. OpenAI unforced (this page) ≠ Buckmaster–Alpöge forced ≠ Anandkumar PINN candidate. Not collapsed.
- Issuer “easier related problem” vs independent difficulty. Attributed process claim only.
Open questions
- Does community review treat the OpenAI euler.pdf as a finished theorem or as another candidate?
- How does it compare, equation-by-equation, to Anandkumar’s unforced R^3 writeup?