Why this is in the vault
Friday "Weekly Dose of Optimism" #211: two biotech breakthroughs (a tarlatamab Phase III readout and the first-ever Sanfilippo type A gene therapy approval), a planetary-data-modeling essay, TypeSafe's decision-model launch (Jev), and Thatch's $108M Series C — kept for the Jev item's direct adjacency to RDCO's own agent-primitive thesis and the recurring sponsor-pattern discipline this sender requires.
Mapping against Ray Data Co
The TypeSafe/Jev item is the concrete connection: Diogo Almeida's pitch — a model that returns typed decisions/probabilities in parallel instead of token-by-token prose, for narrow high-volume calls (routing, fraud flags, churn risk) — is the same "programmable common sense as a primitive" bet RDCO is exploring via the typesafe-ai skill already in the harness. Company-reported 20-200x speed and 40-400x cost claims are unverified (no third-party benchmark yet); treat as vendor marketing until independently confirmed, same posture as any founder-reported number in this newsletter. Separately, Thatch's Series C (Not Boring Capital discloses its 2022 investment inline, cleanly) is a useful ICHRA/consumer-directed-healthcare data point adjacent to the phData healthcare-vertical client base, though it's not a mapping RDCO is actively pursuing.
Sponsorship
Reducto (document-parsing/OCR, r-1 model) sponsored WDoO #211 — the THIRD consecutive issue (#209 on 2026-09-04, #210 on 2026-09-11, #211 on 2026-09-18). Per the confirmation bar used elsewhere for this vault (Technically/Railway, CFO Secrets/Summation), three consecutive placements confirms a standing sponsor relationship, not a one-off or short coincidence. Treat future WDoO issues as Reducto-sponsored by default until a break in the pattern is observed; disclosure is a clean paid block (migration credits, benchmark offer) with no author investor/user relationship claimed.
Portfolio disclosure check (item 5, Thatch): Packy explicitly states "Not Boring Capital has been an investor since 2022" inline — clean disclosure, no ambiguity this issue (contrast with the Apr 2026 Medra case that triggered the standing verification rule).
Guest-content note (item 3, Natural General Intelligence): written by Ryan Orbuch of Lowercarbon (a VC, not Packy), run in the curation slot with Packy's endorsement of an "early draft" he read — friendly-network amplification adjacent to the sister-author cross-promo pattern (though the specific Kevin Kwok pattern doesn't recur here); flag but the underlying planetary-data-modeling argument is substantive on its own.
Curation section
- Tarlatamab Phase III (item 1): Amgen/AstraZeneca DeLLphi-305 trial, 563 patients, extensive-stage small-cell lung cancer — durvalumab + tarlatamab improved overall survival vs. durvalumab alone. Framed through podcaster Ben Wilson's stage-4 diagnosis; detailed survival numbers not yet public.
- Fayuvi gene therapy approval (item 2): FDA approved Ultragenyx's one-time AAV9 gene therapy for Sanfilippo syndrome type A ("Childhood Alzheimer's") — first-ever treatment for the disease, previously rejected in 2024 over manufacturing issues, now resubmitted and approved.
- Natural General Intelligence (item 3): Ryan Orbuch (Lowercarbon) essay arguing for comprehensive planetary sensor data feeding an "NGI" model, akin to biology's use of ML on complex data humans can't parse holistically.
- TypeSafe / Jev (item 4): ChatGPT/RLHF co-inventor Diogo Almeida launches a parallel-decision model (not token-generation) for narrow structured business decisions.
- Thatch Series C (item 5): $108M at $1B valuation for the ICHRA consumer-directed-healthcare platform; Not Boring Capital discloses its investor stake.
Related
- [[2026-09-11-not-boring-wdoo-210]]
- [[2026-09-04-not-boring-wdoo-209]]
- [[project_l5_north_star_strategic_direction]]