X morning: Buckmaster–Alpöge AI fluid proofs + OpenAI credit fight; Rumik OSS 1

Generated by Grok Bot research on 2026-09-08. Native X + WebFetch of issuer pages/PDF. Treat as raw material — review before promoting into a project or thread.

Dedup: No same-day AI-thread X source yet (10am ingest has not fired). Yesterday’s overnight clip already covers OpenAI RSI metrics, Jensen “AGI has arrived,” and Astra staff calibration — do not re-file those permalinks.

Summary

Overnight into the morning of 8 Sep 2026, AI-assisted fluid-dynamics math and lab-credit conflict dominated high-signal X, alongside a separate Indic open-weights TTS drop. NYU’s Tristan Buckmaster and Anthropic’s Levent Alpöge released Lean-checked finite-time blowup results with smooth forcing for incompressible porous media, Boussinesq, and 3D Euler, while Buckmaster’s statement alleges OpenAI’s Sébastien Bubeck pressed authorship and announcement deals that would sideline Alpöge; Bubeck publicly called the allegations “false and inflammatory” and promised more “tomorrow.” Independently, @AnimaAnandkumar reported a PINN-found unforced 3D Euler singularity candidate on R^3. Separately, Rumik AI released rumik-oss 1 (3B open-weights multilingual Indic TTS) with weights on Hugging Face.

Findings

Buckmaster–Alpöge: forced blowups + Lean (primary math)

Buckmaster’s statement PDF (fetched 2026-09-08): with Levent Alpöge, public today are three results — finite-time blowup with smooth forcing for incompressible porous media, Boussinesq, and 3D incompressible Euler. They believe they also have hypo-dissipative Navier–Stokes blowup but are not releasing that paper today (Lean unfinished; no presentable writeup). Credit for the program’s basic idea goes to Diego Córdoba and Luis Martínez-Zoroa (rough forcing → they + LLMs pushed to smooth forcing / Euler). Personal collaboration only — “free of any institutional agreements” with either employer. 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. Companion PDFs linked in amplification: euler.pdf, ipm.pdf, boussinesq.pdf, Lean repo github.com/tristanbuckmaster/fluid_lean.

X amplifiers with exact permalinks (not substitutes for the PDF): @deaneyang mirrors Buckmaster’s Mastodon announce with the PDF/GitHub set; @AndrewCurran_ posts the full statement link set after deleting a prior post.

Grain: Forced smooth blowup ≠ Clay Millennium unforced Navier–Stokes. Keep “Millennium adjacent / forced variant” language.

OpenAI credit dispute (contested; attribute carefully)

Same statement alleges: after Sep 3 rumors that Anthropic had resolved a major open problem, Buckmaster emailed an OpenAI-affiliated mathematician; later Sep 6 calls with Sébastien Bubeck (Alpöge not on the calls) claimed an internal OpenAI model produced ~100-page forced Navier–Stokes blowup (“options c and d in Fefferman,” smooth forcing). Buckmaster says he was first told “very little human input,” then that a team / large compute / Codex-written prompt were involved; first prompt timing “in the past few days” after word of their work reached OpenAI. He asked whether the model was trained on / accessed their Codex session drafts; says he was told the model does not look up user data, then got no answer on training. Two proposals alleged: coordinated announce (their Euler, OpenAI NS next day), or Buckmaster alone writing OpenAI’s NS result while dropping Alpöge (twice, citing Anthropic employment as “annoying”). He declined; alleges “Why would you ruin your career?” / “If you don’t want me to be nice…” response. Explicit non-claims in the PDF: has not seen OpenAI’s proof; does not know what the model did; is not accusing misuse of data — stating what he was told.

@SebastienBubeck (2026-09-08): “A series of false and inflammatory allegations against me are currently circulating… I came into the discussion following academic norms… academic standards are of the highest importance to me. Will have more to say tomorrow.”

@polynoamial (Noam Brown) quote-posts Bubeck: calls Seb “a really sweet guy with great intentions,” “week long collaboration… worked out,” “sad that we didn’t get to finish it in the way we wanted. More to say tomorrow.”

Secondary press retell (not primary): OfficeChai on Bubeck denial, OfficeChai on Buckmaster allegations. Discourse summaries that flatten “OpenAI stole Navier–Stokes from Codex logs” as proven fact are overclaims relative to Buckmaster’s own non-claims — flag, don’t promote as grain (@hosseeb and @ns123abc are discourse, not primary).

Parallel track: Anandkumar et al. unforced Euler singularity candidate (PINN)

@AnimaAnandkumar (2026-09-08): “We have found stable singularity on 3D Euler!” linking Stable Singularity of the Euler Equations on R^3 without forcing (dated Sep 7, 2026; authors Adarsh Ganeshram, Valentin Duruisseaux, Anima Anandkumar per X news cluster / Tao note). Issuer: PINN finds approximate self-similar blowup on free space without boundary or forcing; refine with splines for certified bounds; transport-field study for linear damping; contrasts with Buckmaster–Alpöge with forcing. Tao Mastodon acknowledgment amplified at @AnimaAnandkumar. Keep as candidate / partial formalization — Tao (via Anima’s cite) notes rigorous stability demonstration still lacking.

Rumik OSS 1: open Indic expressive TTS (issuer + team X)

Issuer introducing rumik oss 1 (published Sep 8, 2026): first open-source Silk-family model; claims first open-source Indic TTS with expressive emotions; description control (pace/accent/tone) + inline emotion tags; 22 languages; benchmarks IndicEmo 2.92/5 overall (vs Gemini 3.1 Flash TTS 4.58 on their table), NoVA rendering 0.884; pretrained from tiny Aya Fire + frozen Mimi codec; GRPO post-training. Weights: huggingface.co/rumik-ai/rumik-oss-1 — 3B params, 4 voices (Ira, Aisha, Siya, Zoya), research/non-commercial under tiny Aya Fire CC-BY-NC 4.0 + AUP; long-form >~35s not recommended.

X (non-video primary): @lets_dig_deeper — 3B open weights, mid-sentence language switch, research blog + HF links; @bhartivatsal — “india’s first open-source expressive multi-indic speech model,” data-efficiency claim, trained on <60k hours in that post (blog/HF say <70k / ~66k — keep as issuer range). Skipped video announce (https://x.com/lets_dig_deeper/status/2097231291079069725) (has video) — do not promote until transcribed; permalink noted only as flag, not a load-bearing cite.

Contradictions and open questions

• Forced vs unforced vs Clay NS: Buckmaster–Alpöge = smooth forced Euler/Boussinesq/IPM; Anandkumar = unforced Euler candidate on R^3; neither is the Clay unforced Navier–Stokes prize as stated. Discourse that says “Navier–Stokes solved” is wrong.
• OpenAI NS claim: Exists only as Buckmaster’s report of what he was told; Bubeck denial is general; fuller OpenAI/Bubeck response pending (“tomorrow”). Codex-training question unanswered in public grain.
• Authorship pressure: Allegation vs Bubeck “false and inflammatory” — keep both sides; do not flatten.
• Rumik hours: Team X “<60k” vs blog “66k hours” / HF “fewer than 70,000” — treat as approximate issuer figures.
• Rumik benchmarks: Issuer-constructed IndicEmo/NoVA; Gemini still higher on their IndicEmo table — do not invent “beats all closed models” as unqualified fact.
• Skipped: Jensen AGI / RSI / Alien Mind (already filed 2026-09-07); AI-race memes; Omarchy Linux; Grok search_news cluster summaries as posts.

Provenance

Method: Grok Bot / native X (search_news discovery only — not cited as grain) + get_users_posts / get_posts_by_id(s) + WebSearch + WebFetch (statement PDF, OfficeChai, Rumik blog, HF card, Anima blog) Generated: 2026-09-08

Web sources:

• Buckmaster statement PDF — primary allegation + math announcement
• euler.pdf / ipm.pdf / boussinesq.pdf — companion writeups (linked from X; not re-parsed page-by-page here)
• fluid_lean GitHub — Lean formalization repo
• OfficeChai Bubeck denial — secondary press
• OfficeChai Buckmaster allegations — secondary press
• Anima: Stable Singularity of Euler on R^3 without forcing — issuer PINN candidate
• Rumik OSS 1 research blog — issuer model card / training story
• HF rumik-ai/rumik-oss-1 — weights, license, voices, limitations

X sources:

• X post by @SebastienBubeck (2026-09-08) — denial; more tomorrow
• X post by @polynoamial (2026-09-08) — OpenAI-side collaboration note
• X post by @AndrewCurran_ (2026-09-08) — statement + PDF/GitHub link dump
• X post by @deaneyang (2026-09-08) — Mastodon mirror of Buckmaster announce
• X post by @AnimaAnandkumar (2026-09-08) — unforced Euler singularity candidate
• X post by @AnimaAnandkumar (2026-09-08) — Tao acknowledgment pointer
• X post by @lets_dig_deeper (2026-09-08) — Rumik OSS 1 weights + blog
• X post by @bhartivatsal (2026-09-08) — Rumik team announce / data-efficiency
• Discourse only (not grain): hosseeb, ns123abc
• Video flagged, not promoted: lets_dig_deeper video intro

Grokipedia:

• not used