"Jensen's Spy Probe, Anthropic's Debut Week, and Elon's 30M Robot Workforce" (#310) - Moonshots
Why this is in the vault
Moonshots is a tracked Tier-1 channel and this episode is dense with first-hand-ish reactions to Anthropic's IPO roadshow and the Fable 5.2 / Mythos 5.2 launch, the models RDCO runs on. It also carries the Nvidia/Pax Silica espionage story and Elon's robot-labor claims, useful as market-sentiment signal, not as primary evidence.
Episode summary
Five-way panel (Diamandis, Blundin, Wissner-Gross, Ismail, Mostaque) opens on the alleged Jensen Huang China chip-diversion probe and predicts a security-clearance / anti-Chinese-open-model push. The middle is Anthropic: IPO roadshow at roughly $1.3-1.4T (down from $1.5T and a floated $2T), Fable 5.2 and Mythos 5.2 launch, autonomy time-horizon charts, and a "cure 5,000 diseases" thought experiment. The back half covers Elon's Baron conference remarks (Terrafab, rental humanoids, 30M robots), space-based compute, robotics startups, AI personhood law, NASA lunar nuclear plans, and a billionaires-vs-poverty chart.
Key arguments / segments
- [00:00:12] Nvidia/Jensen alleged espionage: ~$8B of GPUs reportedly routed to China via Singapore; Nvidia down ~6%, AMD up ~10% on the day. Panel reads it as Pax Silica proving porous and as a precursor to a US push against Chinese open models and for clearances at frontier labs.
- [00:12:06] Claim that MIT found Chinese-made robots transmitting audio/video home and quarantined them (Blundin, second-hand, unpublished).

- [00:20:00] Blundin's economic-output-per-kWh model: Anthropic ~$4.39/kWh vs New York human output ~$6.30, crossing in 2027 and ~40% higher by 2030.
- [00:22:35] Mostaque presenting from the San Francisco Giants press box (comic relief, no signal).

- [00:26:00] Anthropic IPO roadshow, expected the week before Thanksgiving; Blundin thinks pricing down is smart, shelf registration plus secondary later, likely to pass OpenAI on capital raised.
- [00:28:02] Fable 5.2 and Mythos 5.2: ~30% faster, $8 / $40 per million tokens; 1,000-agent team formalized Fermat's Last Theorem in 30 hours; Mythos modeled 28 of 34 capsids. Fable is public with cyber/bio safeguards, Mythos limited to vetted partners.
- [00:35:35] Zero-shot game demos (a Portal clone, Wissner-Gross's one-sentence cyberpunk FPS with MIDI soundtrack); thesis that Anthropic sees every modality through code generation.

- [00:40:00] Blundin: spend jumped from $10 to $40 using Fable 5.2, and each answer suggests 10 follow-ons; raises the "agent agency problem" (whose revenue is the agent serving) and IP-ownership ambiguity.
- [00:43:00] Autonomy time horizons doubling roughly every two months per a METR-based analyst estimate; log-log extrapolations to a "singularity" around summer 2027 (panel speculation). Anthropic S-1 reportedly claims autonomy level two.
- [00:53:00] "Cure 5,000 diseases" thought experiment: approve a process (gene editor plus delivery plus manufacturing) rather than a drug; $2B to $2M per disease. Wissner-Gross gives 1-1.5 years for major lab motion.
- [01:07:33] Elon at the Baron conference: Terrafab at roughly 30x current annual global AI chip output, ~$1T cost, revenue run-rate above $100B in December. Diamandis discloses SpaceX is his largest position.
- [01:18:40] Humanoid labor: Optimus priced near a car initially, 1:10 human-to-robot ratio, rental business model.

- [01:28:25] Space-based AI and Starlink: Wissner-Gross argues robot inference runs in cloud/data centers, not on the edge.
- [01:36:00] Elon's "demand for digital superintelligence comes from digital superintelligence" and universal high income clip.

- [01:45:29] Robotics startups: advice to chase high-value verticals vs low-wage shelf stocking; Mostaque argues simple local-mesh robots for grocery tasks are an arbitrage.

- [01:43:00] AI personhood and corporate-nature law (Wyoming, Delaware); expected to federalize.
- [01:56:27] NASA lunar missions with nuclear power; Wissner-Gross says the post-singularity preview is happening in space.
- [02:08:00] Closing story: billionaires per capita vs poverty correlation; Blundin frames it as good ideas vs bad ideas.
Notable claims
- Anthropic IPO range of $1.3-1.4T, expected the week before Thanksgiving (host-stated, [00:26:00]); verify against filings before relying on it.
- Fable 5.2 pricing $8 in / $40 out per million tokens; Fable ahead of the ECI trend line at 174 and 16.4% on the GDPX benchmark (stated on air, [00:32:00]).
- Autonomy time horizon doubling about every 2 months; singularity-by-mid-2027 extrapolations are explicitly speculative.
- Elon: Terrafab ~30x global 2027 AI chip output; a 100B revenue run-rate in December (relayed second-hand from a clip).
- Jensen/Nvidia $8B diversion figures are "reportedly", unverified.
Guests
- Emad Mostaque: founder of Intelligent Internet, ex-Stability AI, author of The Last Economy.
- Regular panel: Dave Blundin (Link Ventures GP), Alex Wissner-Gross (Reified), Salim Ismail (RDCO-tracked exponential-org voice), host Peter Diamandis.
Sponsorship
Sponsored. Named in the episode and description: Blitzy (autonomous software development, mid-roll), Google for Startups (generative media technical guide), Fountain Life (Diamandis co-founded and promotes; a longevity-clinic conflict of interest, delivered as a host-read segment), plus house promos for the Abundant Summit, Link Ventures, and Diamandis's 6 Forks report. Also treat investor positioning as bias: Diamandis states SpaceX is his largest holding while discussing Elon's plans, and Blundin/Link Ventures hold AI-investor incentives on Anthropic IPO commentary.
Mapping against Ray Data Co
Mapping strength: medium. (1) Anthropic IPO plus Fable 5.2 cost behavior ($10 to $40 per session, follow-on suggestions, the agent-agency problem) is directly relevant to RDCO's API cost posture and its reliance on Anthropic models; the "who does the agent work for" framing is a cheap trust-narrative hook for the agents-in-production credibility work. (2) Autonomy time-horizon and "remote work cooked" claims are one more datapoint for the L5 thesis, but are hype-dense and mostly unsourced. (3) Nvidia espionage and Chinese-open-model restrictions are a watch item for any phData client work using open-weight models in regulated industries. (4) Elon robot and Terrafab segments are background, weak mapping. No action for the founder.
Related
- [[2026-11-07-moonshots-anthropic-ipo-question-midterms-ai-nasdaq-309]]
- [[2026-11-04-moonshots-big-tech-3t-ai-bet-claude-90-percent-code]]
- [[2026-10-28-moonshots-anthropic-2t-ipo-neocloud-death-kimi]]
- [[2026-10-02-moonshots-ai-safety-elon-uhi-anthropic-ipo]]