“Weekly Dose of Optimism #191” — @Packy McCormick

Why this is in the vault

Friday curation slot, filed for cadence completeness. The issue is heavily bio/health-science (cancer detection x2, virtual-cell, GLP-1 successor) plus one space-energy story and one fatherhood essay. Mapping to RDCO is weak — Friday WDO issues often are; the Tuesday Packy essay (e.g. Scarce Assets, filed 2026-04-30) is where the load-bearing thinking lives. One item (Overview Energy / space-based power beam) extends the AI-compute-power-demand thesis Packy has been building across multiple issues. Filing as standing-thesis tracking, not as primary research input.

Sponsorship

Same recurring “brought to you by Create” block as #189 — Creatine + Electrolytes packets, co-run by Packy’s brother Dan, 30%-off code attached. New disclosure detail this issue: the product is “run-rating $20 million” a month after launch. Treat as recurring house ad, not rotating advertiser. No surprise sponsor.

Issue contents

Dognosis — beagles + EEG helmets + Bayesian fusion for multi-cancer breath detection. Phase 2 study (3,275 participants, six Karnataka hospitals) published in Journal of Clinical Oncology. 90.8% sensitivity / 91.3% specificity across seven cancer types; sensitivity holds at 90.6% for Stage I-II vs 16.3% for GRAIL Galleri at Stage I. Cheaper test meant for India + emerging markets first.
(1b) Mayo Clinic REDMOD — AI on routine abdominal CTs detects pancreatic cancer ~16 months before clinical diagnosis. 73% sensitivity vs 39% for radiologists on the same scans. Moving to prospective AI-PACED trial.
Zuck/CZI Biohub commits $500M to Virtual Biology Initiative. $400M internal (multi-modal proteomic/genomic/cellular datasets, cryo-ET, advanced microscopy); $100M external grants (Allen, Arc, Broad, Sanger, Human Cell Atlas, Human Protein Atlas). NVIDIA technology partner. Goal: dataset 10x current ~1B cell baseline.
(2b) Overview Energy + Meta — first 1 GW capacity reservation for space-based solar. GEO satellites capture sunlight, convert to laser, beam to existing terrestrial solar farms. Different model than Reflect Orbital (mirrors). Packy spoke with founder Marc Berte; says it feels nearer-term than expected.
Rainmaker — first verified weather modification at scale. 143MM gallons of “unambiguously man-made precipitation” last quarter, validated via radar + satellite. Packy has covered them more than any non-portfolio company in the Dose.
(3b) Skydio — $3.5B over five years on US drone manufacturing. $1B+ to US suppliers; SkyForge facility 5x existing footprint with co-located suppliers. Framed as the China-decoupling drone bet.
Colossal — Bluebuck antelope (Hippotragus leucophaeus) added to de-extinction roadmap. After Wooly Mammoth and Direwolf. No technical detail in the blurb.
Derek Thompson + Aziz Sunderji — millennial dad time-with-kids data. 1965 married fathers ~30 min/day active engagement; today’s millennial dads >80 min. ~3x Boomer baseline, ~4x Silent Generation. Maternal time also up — not a substitution. Four explanations canvassed (women in workforce, dads enjoying it, Rug Rat Race status competition, Dunkelman weak-ties decline pushing leisure indoors); Packy adds remote-work / phones-on-playground as a fifth.

Extra Doses (truncated in email): Science Breakthroughs including a 5-in-1 GLP-1 successor, Earth AI, Paul Tudor Jones, Nicholas Thompson.

Curation section — notes

Per-link disclosures and self-cross-promo:

Dognosis (item 1). Packy met founder Akash Kulgod in person at the Progress Conference. Deep coverage handed off to Rahul Sanghi at Tigerfeathers — Tigerfeathers is a friend-of-Packy publication, flagged as a recurring third-party amplifier rather than a sponsor relationship. No disclosed Packy investment.
Mayo REDMOD (1b). Bridge from prior week’s BioNTech/Genentech mRNA pancreatic cancer vaccine coverage — Packy is intentionally building a multi-issue early-detection thread.
CZI Biohub Virtual Cell (2). Linked external explainer is Elliot Hershberg’s “What Are Virtual Cells?” — Packy’s standing recommended-reading author for bio/AI intersection.
Overview Energy (2b). Packy spoke directly with founder Marc Berte the prior week; deep-coverage handed off to Rob L’Heureux’s “Making Space Lasers Boring” in Arena Magazine. Rob is a Not Boring orbit writer. No disclosed investment.
Rainmaker (3). Explicitly flagged “non-portfolio” — Packy has covered them recurrently without economic interest, which is itself a strong signal of his conviction.
Skydio (3b). Coverage handoff to “my friend Molly O’Shea” — friend-of-Packy disclosure, not sponsor.
Colossal (4). No deep-fetch handoff. Packy is on the Colossal hype train across the vault (cf. 2026-04-07-moonshots-ep-synbio-ben-lamm-colossal, 2026-04-18-moonshots-colossal-cloning-de-extinction); this is incremental flag, not new substance.
Millennial Dads (5). Derek Thompson + Aziz Sunderji piece. Packy shares first-person remote-work parenting anecdote — this is Packy-as-dad, not Packy-as-analyst.

Deep-fetch budget unused. Skipped both available slots. The two strongest candidates were the Tigerfeathers Dognosis interview and the Arena space-lasers piece. Both are bio/space-frontier coverage with no direct line into data engineering, agent architecture, prediction markets, content-as-product, or founder-COO operating rhythm. The in-newsletter blurbs already carry the sensitivity/specificity numbers, the GW figure, and the bayesian-fusion mechanism — fetching wouldn’t add load-bearing facts for the vault.

Mapping against Ray Data Co

Weak overall. This is a Friday WDO issue — bio-heavy, parenting essay, no agent-architecture or operating-rhythm content. Filed for cadence, not for active use.
AI compute power-demand thesis (medium). Overview Energy’s space-based solar is a new candidate to add to the “what powers the inference build-out” list alongside the geothermal / wave / nuclear / natural gas options tracked in 2026-04-17-not-boring-weekly-dose-189 and 2026-02-18-not-boring-power-age-of-intelligence. The Meta GW reservation is the load-bearing data point — Meta is committing real procurement dollars to space-based power, which moves the technology from sci-fi to procurement-stack.
Virtual cell / bio foundation models (weak-but-tracking). CZI’s $500M is the bio analog of frontier-lab compute spend. Worth tracking as a parallel to the LLM compute-arms-race story but no direct RDCO line.
Self-knowledge for the founder, not the business (parenting item). Filed because the founder is a millennial dad and the data validates the lived experience — irrelevant to RDCO operating model. No work product implication.
Cross-China industrial policy thread (weak). Skydio’s $3.5B announcement extends the broader US-manufacturing-reshoring story but isn’t directly RDCO-adjacent.