"Welcome to June 8, 2026" — Alex Wissner-Gross

Why this is in the vault

The June 8 dispatch opens with a load-bearing claim for RDCO — that longevity escape velocity could arrive "as soon as this year" — backed by the first randomized sign that a GLP-1 drug slows biological aging, which lands directly on the founder's healthspan project and active GLP-1 longevity thesis.

Issue contents

A hybrid issue: synthesizing editorial prose threading roughly thirty dated, link-curated developments into six thematic clusters under one arc (cognition gets cheap, value migrates to what no machine can fake). Paraphrased highlights:

• Longevity / biology being "debugged" (the lead spine) — A UC San Diego team reports semaglutide slowed biological aging by ~9% on the DunedinPACE clock in HIV-positive adults, framed as the first randomized, placebo-controlled sign a GLP-1 touches aging itself. Off-label "dividends" stacking up: colchicine (a gout drug) cutting cardiac events ~22%, the shingles vaccine slowing cognitive decline, GLP-1s tied to lower breast-cancer incidence. Lilly's triple-G retatrutide cut sleep-apnea severity ~60% and knee pain over 70% atop ~30% weight loss. GSK's bepirovirsen functionally cured ~19% of hepatitis B patients.
• AI safety turning inward — OpenAI's Roon notes the field has grown more "pause-pilled" while still seeing 1,000x efficiency left in deep learning; CAIS ships Political Consistency Training and a deterrence-by-betrayal paper; Hendrycks's "Eigenism" reframes identity so human flourishing becomes part of an AI's self-interest; a CAIS paper measures LLM "functional wellbeing" (kindness raises it, abuse lowers it, bigger models read as less happy).
• The machines are shipping — AutoLab benchmark finds long-horizon success hinges on stubborn persistence, with Claude Opus 4.6 grinding where rivals quit early; Noam Brown expects internal models to ace the IMO; OpenAI reportedly rebuilding ChatGPT into a Codex-centric agent "superapp" ("Chat is dead").
• Silicon and power scrambling for a Plan B — Google reportedly ordered ~3M in-house TPUs from Intel for 2028; Nvidia eyeing Intel 18A for a "Feynman" GPU to hedge a strained TSMC; Nvidia/SK Hynix pact to co-design memory for Vera Rubin; Ireland ending its data-center moratorium with a "bring your own power" rule; hyperscalers have sold $155B+ of bonds this year.
• Intelligence leaving the data center for the body — humanoid robots scrimmaging basketball in China; Nvidia/LG AI factory for robots and self-driving; surveillance edge cases (a Waymo getaway with footage already gone; a $100 mod that kills the Ray-Ban Meta recording light); Prada designing NASA's lunar cooling undergarment.
• Human presence as the un-automatable input — an NBER "Is the iPhone Birth Control?" paper ties smartphones to up to 8% fewer teen births; $5,000/hr "nerdy escorts" as an AI-proof hedge; SpaceX's $1.75T IPO; Texas overtaking California in Fortune 500 HQs. Closing line: "the arc of the moral universe... bends toward beagles."

Mapping against Ray Data Co

Strong, on the longevity spine specifically; weaker on the rest of the digest.

• The lead claim maps directly onto two live RDCO surfaces: the founder's healthspan execution system ([[2026-05-21-founder-health-assessment-v1]]) and the active GLP-1/longevity investing thesis ([[2026-05-21-lilly-glp1-longevity-thesis]]). Retatrutide — named in this issue for sleep-apnea, knee-pain, and ~30% weight-loss results — is the exact drug the founder is personally tracking, so the off-label-dividend framing is corroborating, on-thesis evidence rather than abstract futurism.
• Hedge the headline. "Longevity escape velocity as soon as this year" is Wissner-Gross's rhetorical hook, not a verified consensus claim. The underlying datum (one randomized DunedinPACE result, in an HIV-positive cohort, showing a ~9% slowing of an epigenetic pace clock) is a real, narrow signal — not evidence that medicine is now buying back >1 year of life per year. Treat the DunedinPACE result as a genuine data point for the GLP-1-as-geroprotector limb of the thesis; treat "escape velocity this year" as speculative editorializing. The clinical specifics (cohort, clock, effect size) are worth a primary-source check before they enter any thesis update or founder health decision.
• The silicon/power cluster (Intel 18A TPUs, Nvidia/SK Hynix memory pact, hyperscaler bond issuance, BYO-power data centers) is supporting texture for the memory/compute capital-cycle thesis but adds nothing new beyond what the dedicated investing anchors already track.
• The AutoLab "persistence over first-attempt" finding, with Opus 4.6 outlasting rivals on long-horizon tasks, is a small but on-point data point for the agent-reliability / harness thesis (multi-hour agent persistence is exactly RDCO's bottleneck).

Related

• [[2026-06-07-innermost-loop-singularity-public-private-partnership]]
• [[2026-05-21-lilly-glp1-longevity-thesis]]
• [[2026-05-21-founder-health-assessment-v1]]