06-reference

not boring wdoo 210

2026-09-11·reference·source: Not Boring (Weekly Dose of Optimism)·by Packy McCormick
multi-agent-systemsai-researchmath-proof-automationworld-modelsnot-boring-capital-portfolio

Why this is in the vault

Issue #210's item 1 (OpenAI's Navier-Stokes proof) is the same story already receiving heavy same-week vault coverage from four other tracked senders — kept here specifically for Packy's distinct framing angle (the OpenAI/Anthropic credit dispute as a preview of "AIs doing all of our work while we argue for credit") rather than for new facts.

Curation section

  1. OpenAI publishes its Navier–Stokes proof. ~10,000 AI agents worked ~88 hours on an unreleased model (more capable than GPT-6 Astra) to produce a proposed solution to the forced Navier–Stokes existence/smoothness problem (one of the seven Millennium Prize Problems), then spent 17 more hours formalizing and verifying it in Lean. The construction starts fluid at rest, applies a smooth external force, and produces unbounded velocity while keeping total energy finite — addressing the forced version of the problem, not the full unforced Millennium Prize statement. Controversy: Anthropic was reportedly pursuing its own Navier-Stokes solution with a mathematician and accused OpenAI of using leaked Codex conversation logs; Sam Altman denied it but said rumors of Anthropic's progress spurred OpenAI's own push. This is the same underlying story already filed in depth via [[2026-09-09-alphasignal-navier-stokes-agent-swarm-mistral-funding]], [[2026-09-09-stratechery-ben-thompson-openai-math-muse-agent]], [[2026-09-06-innermost-loop-astra-agi-agent-collusion-fermat-proof]], and the Moonshots transcript note — this WDoO entry adds no new facts but frames the episode as evidence that "AIs [will do] all of our work for us while we spend time doing more human things, like arguing for credit," a distinct closing-line take worth logging alongside the more technical treatments elsewhere in the vault.
  2. Google DeepMind releases AlphaGenome Atlas — a one-petabyte database predicting molecular effects of all ~9 billion possible single-letter DNA changes across the human genome (including the 98% that's non-coding). Already helped a Broad Institute team identify a splicing-disrupting variant in an unresolved rare-disease case. Framed as a "where to look" tool, not a replacement for wet-lab experiments. Science-optimism item, no direct RDCO mapping.
  3. Hermeus unveils Ramjet-X, an expendable ramjet test vehicle launched from its reusable Quarterhorse aircraft, sold as "Flight Test as a Service" so customers can test sensors/materials/autonomy/propulsion without building their own test program. Series-recurring subject — Hermeus previously covered at [[2026-05-29-not-boring-weekly-dose-195]]. No RDCO mapping; aerospace-testing infrastructure item.
  4. Michael Levin publishes the peer-reviewed "Ingressing Minds" paper, proposing minds as patterns in a structured nonphysical ("Platonic") space that biological or engineered bodies serve as interfaces for — extended to argue computers could in principle be such an interface, framed as the most compelling argument Packy's encountered for AI "consciousness." Philosophy-of-mind item; no RDCO mapping, not independently verified beyond Packy's framing.
  5. Neko Health opens a New York City location. Preventative-health scanning company (Hjalmar Nilsonne + Spotify's Daniel Ek) offering a $499 full-body scan (bloodwork, cuffs, imaging) in a high-design retail format. Packy explicitly discloses "They're not a sponsor and I'm not an investor" while sharing a referral link to skip the waitlist — clean non-investor disclosure, not portfolio-disclosure ambiguity for once.

No verification research was run this issue: the one item with any Not Boring Capital portfolio-disclosure risk (Neko Health) came with Packy's own explicit non-investor disclaimer, so the usual public-investor-list check wasn't needed. Zero deep-fetches were performed — every item's RDCO relevance (or lack of it) was resolvable from the blurb text plus existing vault coverage of the Navier-Stokes story; none of the remaining four items (AlphaGenome, Ramjet-X, Ingressing Minds, Neko) had a specific enough RDCO hook to justify following a link out.

⚠️ Sponsorship

Explicit paid sponsor block, top of issue: Reducto (document-ingestion/parsing API), promoting its new "r-1" frontier parsing model — claimed 20% reduction in parsing errors, improved latency at volume, 1¢/page all-in, plus up to $5,000 in migration credits. This is the same sponsor as the prior issue ([[2026-09-04-not-boring-wdoo-209]]), confirming Reducto is running a sustained, not one-off, campaign across Not Boring's Friday slot — consistent with its parallel appearances in AlphaSignal issues from July 2026. No editorial item this issue touches document parsing, so the sponsorship doesn't appear to bias the curated content.

Mapping against Ray Data Co

The concrete connection is item 1: the OpenAI/Anthropic Navier-Stokes dispute is the same live agentic-math-verification story the vault is tracking in depth this week from four other angles (AlphaSignal's swarm-mechanics framing, Stratechery's "math muse agent" competitive-positioning read, Innermost Loop's agent-collusion angle, Moonshots' AGI-timeline framing) — this note's value-add is narrow but real: Packy's closing line ("AIs doing all of our work for us while we spend time doing more human things, like arguing for credit") is a sharper, more cynical one-liner on lab-credit dynamics than the more technical treatments carry, and is worth citing directly if Sanity Check ever runs a piece on AI-lab competitive behavior around headline math/science results. The Reducto sponsor recurrence is a secondary, minor data point reinforcing that Reducto is treating the newsletter-sponsorship channel (Not Boring + AlphaSignal) as a standing acquisition lane for document-ingestion buyers — mildly relevant to RDCO's own data-engineering roots as a vendor to watch, not act on.

Related