Summary
Peter Diamandis, Salim Ismail, Dave Blundin, and Alex Wezner-Gross open with the Musk v. Altman $100B lawsuit (trial begins April 27) and the parallel reorg at xAI — eight founding engineers gone, SpaceX leadership parachuted in, seven new models in training including a 10T-parameter frontier model. The episode turns into a state-of-the-AI-economy piece: OpenAI raised $122B at an $852B valuation while Anthropic now trades 3x heavier on secondaries; $3B/day flowing into AI; 64% of Q1 2026 VC ($242B record) concentrated in OpenAI/Anthropic/xAI/Waymo. Long arc on AI’s labor displacement (Andreessen vs. reality), Altman’s “new social contract” framing, and Anthropic’s $400M acqui-hire of 10-person Coefficient Bio as the template for AI-into-biotech. Closes with quantum-vs-Bitcoin, vertical farming hitting commercial scale, iron-air batteries, and the “boy band outro.”
Key Segments
- [00:04] xAI being “rebuilt from the foundations” — 7 models in training, 10T-parameter frontier teacher model
- [00:15] Musk v. Altman trial April 27 — Musk asking Altman/Brockman step down + revert to nonprofit
- [00:33] OpenAI cap table dissected: CEO has no shares, Microsoft owns ~25%, employees only 15%
- [00:35] Anthropic secondaries at $600B (up from $380B); 3x demand vs. OpenAI on secondary market
- [00:38] $242B Q1 2026 AI VC — 64% in 4 companies; Alex’s “recursively self-improving” bar
- [00:42] Diamandis recounts Peter Dannenberg (DeepMind) saying ~99% of white-collar jobs replaceable in 2 years
- [00:52] Andreessen “AI job loss is fake” framing vs. zero-junior-hire reality; “don’t get a job, build a company”
- [00:56] Altman’s “new social contract” — UBI, portable benefits; Andrew Yang quote: “all we can do is write checks”
- [01:02] Alex’s hot take: redistributive social contract = “extreme lack of imagination”
- [01:12] OpenAI Foundation’s $130B; Coefficient Bio acquired by Anthropic for $400M (10 people, 8 months old)
- [01:18] Intelligence explosion “metastasizing” into biotech; FDA collapsing 2 trials → 1; virtual cell by 2030
- [01:28] China robotics overwhelm — Agibot 10K shipped, Unitree IPO ($610M), Xiaomi Cyber 1
- [01:34] US has the VLA foundation models, China has manufacturing — race to converge
- [01:37] Quantum-vs-Bitcoin: Google moves Q-day to 2029; Saylor adds 88K BTC ($7.25B); Alex: AI agents will reinvent crypto stack
- [01:55] Vertical farming hits $8B (target $40B by 2030); Holland is #2 food exporter on tiny footprint
- [01:57] Iron-air 100-hour batteries commercial — 1/10 cost of lithium-ion for grid storage
- [02:00] AI tutors deliver 2x learning gains; “cruel” to teach kids without AI
Notable Claims
- xAI training 7 models simultaneously: 2x 1T-param, 2x 1.5T-param, 6T-param frontier, 10T-param teacher
- Colossus 2 running ~700K GB200/GB300s; ~$18B in hardware
- OpenAI: $122B raised at $852B (Amazon $50B contingent on AGI, Nvidia $30B, SoftBank $30B, retail $3B)
- Anthropic secondaries: $2B demand vs. $600M for OpenAI (3x delta)
- VC into AI Q1 2026: $242B (record); 64% in OpenAI/Anthropic/xAI/Waymo; ~$3B/day pace
- Software dev openings rebounding +30% YoY to 67K; 80K layoffs in Q1 2026 in marketing/sales/CX
- 88% of companies report revenue gains from AI per Nvidia survey; 30% claim 10%+ lift
- AI-developed drugs: 85% phase 1 success (vs 52%); 70% phase 2 (vs 38%)
- Anthropic acquires Coefficient Bio for $400M (10 people, 8 months, no revenue)
- Eli Lilly + Insilico: $2.75B AI drug deal ($115M upfront, rest milestones)
- iron-air batteries: 1/10 cost of Li-ion for grid; 100-hour storage now commercial
- EVs sold globally 2025: 12.7M; China = 1 in 2 new cars; IEA’s 2015 prediction was off by 25 years
Guests / Speakers
- Peter Diamandis — Host, XPRIZE founder, early SpaceX investor, Insilico investor
- Salim Ismail — Co-host, Exponential Organizations author, working on “organizational singularity” book
- Dave Blundin — Co-host, EverQuote founder, MIT entrepreneur
- Alex Wezner-Gross — Co-host, physicist; running Innermost Loop daily AI newsletter; Eon Systems (uploads)
Mapping against Ray Data Co
- Anthropic-OpenAI ARR gap is structural, not cyclical: Anthropic surpassing OpenAI in secondary demand validates the enterprise/agent-first thesis. RDCO’s positioning of Claude as the operational backbone for the “AI COO” pattern lines up with where capital is actually flowing — see ai-landscape, anthropic.
- “Recursively self-improving startup” as the new bar: Alex’s framing — Frontier Labs already qualify; the bar for series-A AI companies is now “are you using your own product to improve your own product?” RDCO’s vault-as-product pattern (Claude editing its own skills, scripts, graph) hits this directly. Worth a Sanity Check piece.
- The labor displacement arc is non-linear: junior devs at -20% YoY hires while senior devs grow; “no mass layoffs, just no hiring.” This matches the dynamic the founder has flagged for the newsletter — AI removes the bottom rung, not the top, and the political response will be checks-not-thoughtfulness (Andrew Yang).
- Coefficient Bio acquisition is the template: $400M for 10 people, no revenue, 8 months old. Same pattern as DeepMind in 2014. RDCO’s research-brief and curiosity skills should be flagging acqui-hire targets in the longevity/bio space — this is where compute capital is metastasizing next.
- xAI rebuild signals the benchmaxing penalty: Alex’s read that earlier Grok models “smell benchmaxed” is now consensus inside xAI itself; new leadership re-architecting from scratch. Validates RDCO skepticism of vendor benchmark claims.
Related
- ai-landscape
- anthropic
- openai
- xai
- 2026-04-11-moonshots-ep246-spacex-ipo-claude-mythos
- longevity-stack
- abundance-thinking
- ubi-and-social-contract