06-reference

not boring wdoo 207

2026-08-21·reference·source: Not Boring (Weekly Dose of Optimism)·by Packy McCormick
ai-routingreturn-on-tokensnot-boring-capital-portfoliophysical-aibiotech

Why this is in the vault

Issue #207 of the standing Friday optimism series — kept for item 4 ("Big Week for Routers"), which extends the vault's already-tracked Return on Tokens thesis with a concrete, disclosed data point: Stripe's $7.5B acquisition of OpenRouter plus router launches from two Not Boring Capital portfolio companies in the same week.

Curation section

  1. Merck/Moderna mRNA melanoma vaccine hits Phase III. Given with Keytruda post-surgery in high-risk melanoma patients, the personalized mRNA vaccine significantly cut recurrence/spread vs. Keytruda alone — the first successful late-stage trial for a personalized mRNA cancer therapy, with a possible ~2027 approval/launch window. One cancer type, still expensive/complex, but a real efficacy signal.
  2. "Voluntary attention regulates acute immune responses" (Mizrachi, Rottem & Rozenkrantz, Nature Human Behavior). Packy re-quotes his own prior essay "Means and Meaning" (attention as identity-constitutive) to frame the finding — self-cross-promo, not new analysis layered on the study.
  3. Generalist's GEN-1.5 robot claims one-shot generalization — trains from a few seconds of demonstration data (sim, human video, or chained sub-tasks) and transfers to unseen tasks. Framed as a step toward the "train on X, ask for not-X" holy grail in robotics; genuinely early-stage per Packy's own hedge.
  4. Big Week for Routers. Stripe confirmed acquiring OpenRouter for a reported $7.5B (NYT); OpenRouter was founded by OpenSea's Alex Atallah. In the same week, Merge (Merge for Workforce, 75x token reduction claim) and Ramp (bought router.com, launched its own Router product) — both explicitly disclosed by Packy as Not Boring Capital portfolio companies — shipped competing router products. Packy ties this to Markie Wagner's June "Return on Tokens" (ROT) framing already tracked in the vault. Portfolio-disclosure note: this is a case where the disclosure is clean and self-initiated (no verification gap), unlike prior issues' ambiguous cases — worth noting for calibration on how the series usually handles this.
  5. Big Week for Drones — Zipline/Uber partnership (drone delivery investment) and Amazon Prime Air expansion reporting. Content reads as reused/dated boilerplate (references "1 million deliveries a day by 2019" and DoorDash Air) rather than fresh reporting this week — flagging as a possible newsletter production artifact, not verified further since it doesn't cross the RDCO relevance bar either way.

One deep-fetch considered (item 4, the Stripe/OpenRouter acquisition) but not taken: the email itself already carries the load-bearing facts (acquirer, target, price, NYT sourcing) and the portfolio angle is self-disclosed, so a public follow-up wouldn't add a verification finding beyond what's already stated.

Cross-promo/affiliate flag: the "in lieu of a sponsor" favorites block up top includes Ramp — the same company that appears substantively in item 4 as a Not Boring Capital portfolio company shipping a competing router. Whether or not the top-block placement is paid, Ramp gets both an affiliate-style plug and a full curated item in the same issue — a repeat-brand pattern consistent with prior WDoO issues (e.g. #206's Matic double-coverage).

⚠️ Sponsorship

No paid sponsor this issue (explicitly stated: "in lieu of a sponsor"). The favorites block (Matic, Quince, Notion, Ramp, Create) reads as affiliate-style placement rather than disclosed sponsorship — sponsored: false is accurate for the paid-ad question, but Ramp's presence in both the favorites block and item 4's substantive coverage (as a disclosed portfolio company) is a structural repeat-amplification pattern worth flagging, not a hidden conflict.

Mapping against Ray Data Co

The concrete connection is item 4: Stripe's $7.5B OpenRouter acquisition is a market-validating data point for the Return on Tokens thesis already anchored in the vault ([[2026-06-10-not-boring-return-on-tokens]]) — token-routing-for-cost-and-capability is now big enough to draw a $7.5B strategic acquisition, and two portfolio companies (Merge, Ramp) are racing to build the same layer. This directly touches RDCO's harness-engineering thesis: routing intelligence to the cheapest sufficient model is the same "efficiency compounds" argument as CLAUDE.md's context-management hard rule, just at the model-selection layer instead of the token-budget-within-a-session layer. Items 1, 2, 3, and 5 are pattern-continuation of the series' recurring optimism throughline already well-covered by prior WDoO notes and don't add new RDCO-specific signal.

Related