"Anthropic Files $965B IPO, Trump Signs AI Executive Order, and ChatGPT Crosses 1B Users | EP #262" — Moonshots

Why this is in the vault

Three of this episode's lead stories sit directly on top of RDCO's active investing and operating theses: Anthropic's confidential IPO filing (RDCO runs its COO agent on Claude and tracks the Anthropic IPO closely), the Trump AI executive order (regulatory landscape for any AI deployer), and ChatGPT crossing 1B MAU with Claude growing 640% YoY (frontier-lab distribution dynamics). The panel also restates the "innermost loop" AI-infra capital thesis and offers a clean, founder-relevant data point on the solopreneur / non-engineer-as-builder shift — which is literally RDCO's own operating model. Captured as a primary-source pulse on where the Diamandis/Blundin/Wissner-Gross/Mostaque circle sees capital and capability flowing the week of the filing.

Episode summary

Weekly "WTF Just Happened in Tech" roundup (~18 stories) hosted by Peter Diamandis with regulars Dave Blundin (Link Ventures) and Alex Wissner-Gross, plus guest Emad Mostaque (Intelligent Internet) dialing in from the UK; Salim Ismail absent. Anchored on the Anthropic IPO, the Trump AI EO, and the ChatGPT 1B-user milestone, the show fans out across biodefense, robotics, Microsoft's in-house models, an AI sovereign-wealth-fund proposal, AI-and-jobs data, Nvidia's first PC chip, data-center water FUD, collapsing media trust, and a closing block of billion-dollar longevity stories. Tone is uniformly accelerationist and pro-private-sector; framing is "no doom, no gloom." Two paid sponsor reads (Fountain Life, Blitzy) plus heavy house promotion of Diamandis's own funnels.

Key arguments / segments

• [00:03] Trump AI Executive Order — Order rejects heavy-handed, permission-based regulation; instead asks labs to voluntarily give government access to new models 30 days before public release, and directs agencies to deploy AI-powered cyber defense. Altman and Anthropic publicly on board. Panel reads it as the US "planting its flag" — compete, don't constrain — downstream of the "Mythos moment" (privatized capabilities that used to be NSA-exclusive, e.g. zero-day discovery at scale). Mostaque: AI is now "full-spectrum dominance," so even 30 days is a tactical edge; Europe can't move this fast. Blundin: it's effectively letting labs self-regulate for one short window — fine short-term, not a long-run governance model.

• [00:09] ChatGPT crosses 1B MAU; Claude growing 10x faster — ChatGPT hit 1B monthly active users ~3 years after launch (vs YouTube ~10yr, Instagram 8yr, TikTok 5yr); OpenAI YoY growth ~62%. Claude at ~56M MAU growing ~640% YoY. Mostaque frames Claude as the "Apple" to OpenAI's "Microsoft," with ~7B people still untouched and Altman's claimed 100x drop in cost-of-equivalent-intelligence over 18 months. Recurring strategic thread: distribution beats model weights (Altman reportedly would pick 1B users over the strongest model) — but post-"code red," GPT-5.5 reportedly gives OpenAI both. The next race: "who is your Jarvis" — the one coordinating agent next to you is the only game in town.

• [00:13] OpenAI "Rosalind" biodefense + the shrinking "G" in AGI — OpenAI launched Rosalind Biodefense (outbreak detection, disease surveillance, vaccine acceleration) for government/public-health/trusted researchers. Wissner-Gross's thesis: dangerous-capability models (bio via Rosalind, cyber via the GPT "Mythos" series) are being carved out of general public models into fine-tuned, access-gated versions — the beginning of "closing off the generality of intelligence" for security reasons. Blundin asks whether it's guardrailed/distilled/retrained; consensus is scaffolding "unshackling" plus targeted post-training and tool access.

• [00:19] DNA-synthesis screening letter to Congress — Diamandis co-signed (with Altman, Amodei, Hassabis, Suleyman) an open letter pushing laws to force DNA-synthesis firms to screen orders. Debate on where to put the AI-safety choke point (conception vs. action vs. after-the-fact), Twist Biosciences as the responsible-player example, and Mostaque's bigger point: public sector uses ~0.01% of world tokens but is ~20% of GDP — governments should subsidize compute for public goods (e.g. screening every sequenced sample).

• [00:30] OpenAI Robotics hiring; the "innermost loop" — OpenAI building an in-house robotics team; Aditya Ramesh (ex-Sora lead) heading it after Sora's shutdown — the video researchers moved to embodied robotics. "Innermost loop": robots build the data centers and fabs that build the chips that host the models. Mostaque argues the physical-robot market is far bigger than the GPU market and robots won't depreciate like GPUs. Blundin calls robotics a "very good 10-year investment theme."

• [00:35] Anthropic confidential IPO filing (lead story) — Anthropic confidentially filed with the SEC, potentially the first major frontier lab to go public, beating OpenAI. Polymarket gives ~60% odds Anthropic surpasses ~$1.8T market cap on day one (≈ SpaceX's anticipated IPO). Blundin: only ~5,000 employees splitting enormous value; this cohort can do "thousands of billion-dollar acquisitions" — unprecedented dry powder. Mostaque pegs Anthropic revenue ~$50-60B at ~20x revenue (rich but not Palantir-rich), and says $100-150B revenue next year wouldn't shock anyone. Note: episode title cites a "$965B IPO" valuation; the on-air discussion centers on the ~$1.8T first-day-cap Polymarket line.

• [00:40] Speed-to-$1T and revenue-per-employee — Time to $1T: Apple 42yr, Google 21yr, SpaceX 24yr, OpenAI ~10yr, Anthropic ~5yr. Anthropic ~$9.4M revenue/employee vs Apple ~$2.5M, Google ~$2.1M (≈4x) — the exponential-organization thesis live. Blundin's sharpest claim: the AI economy is now "funded to live within itself" — IPO capital can flow straight into the agent economy without touching banks/Main Street. Open question raised: do we start measuring revenue-per-agent?

• [00:46] Microsoft Build 2026 — 7 in-house models — Microsoft shipped 7 from-scratch models (reasoning, coding, image, video, transcription), no OpenAI distillation; Suleyman claims training compute up 1-trillion-fold with another 1000x in 3yr; a tuned Excel model matches GPT-5.4 at ~10x efficiency; Mayo Clinic healthcare-model collab. Wissner-Gross verdict: "they're not in the game" — mid-tier, compute- and talent-starved; Microsoft is now playing IBM to OpenAI's Microsoft. Mostaque: Microsoft is at the level of "a good Chinese lab," pursuing hyper-specialized office intelligence, not general superintelligence. Blundin: it's a battle of people, not brands (Zuckerberg's reported $1B offer to Mark Chen, declined; could Microsoft even make it?).

• [00:52] Fountain Life (sponsor read) — Diamandis with Fountain Life's CMO: 3.3% of "healthy" members are found to have an undetected cancer via full-body MRI + early-cancer-detection screening. CTA fountainlife.com/peter.

• [00:54] NYT Elon "19% of goals" piece — Diamandis rejects an NYT analysis claiming Musk met only 19% of 602 goals stated 15 years ago (he notes ~75% on-time for 2015); panel frames it as an ideological hit piece and invokes VC power-law logic (10% of companies = 90% of returns).

• [00:58] Leiden Declaration (mathematicians vs AI) — 130+ mathematicians / International Mathematical Union warn AI proofs can be convincingly wrong and that AI firms could steer which problems get funded. Wissner-Gross: rear-guard action after OpenAI solved the Erdős conjecture; "AI is cooking math," wrong side of history. Mostaque: it's AI "confabulation," not psychosis; over-specialization left mathematicians unable to do daring cross-domain work.

• [01:02] American Federation of Teachers AI plan — 1.8M-educator union's 10-point plan: no screens preK-2, K-12 safety/privacy, limits on AI use, and a "tech tax" on big tech. Panel hostile (Blundin's anti-new-tax rant; Gandhi "first they ignore you…" framing); plugs Math Academy as the scientifically-grounded counter-model.

• [01:06] Bernie Sanders AI Sovereign Wealth Fund Act — Proposes a one-time 50% tax on stock (not profits) from OpenAI, Anthropic, etc. into a public wealth fund. Blundin: structurally broken — you'd have to dump ~$5T of stock to realize value, tanking the market. Wissner-Gross: a sovereign wealth fund isn't intrinsically bad (cf. the 10% Intel stake, golden shares) and "universal basic equity" is coming, but not via forced divestiture. Mostaque: it would entrench the labs as too-big-to-fail; better to seed Anthropic/OpenAI shares into Invest-America accounts for every US child.

• [01:16] AI and jobs — the optimistic data — WaPo's 5 policy options (tax robots / unemployment insurance / upskill / dividends / do nothing); Forbes counterpoint via Apollo's Torsten Slok: AI is a net job creator and firms blame AI to justify cuts. Cognizant hiring 20,000 grads. Blundin: employment up 2x across his portfolio — the real shift is non-engineers (e.g. healthcare.com's domain experts) becoming builders, which outweighs automation losses. Mostaque: no job losses yet but a hiring pause; the real disruption is next year as agents/robots get competent. Wissner-Gross: don't get bogged in the "moral panic" — point superintelligence at hard problems.

• [01:21] Nvidia N1/N1X — first ARM PC chip — N1X: 20 CPU cores, 6,144 CUDA cores, ~RTX-5070-class laptop GPU; a shot at Apple/Intel/AMD. Theories: laptop as the test-bed for an AI-native OS and Nvidia's first direct consumer/data channel; leveraging its dominant TSMC frontier-node pipe; "Jarvis at home" edge intelligence; blocking AMD's Strix Halo.

• [01:27] Data-center water FUD — Satya Nadella clip: closed cooling loop filled once, ~zero ongoing water, annual use ≈ one restaurant. Diamandis: US data centers ~150B gal/yr vs California almonds ~1.3T gal/yr. Mostaque blames a viral ~1000x math error (attributed to Karen Hao's "Empire of AI"). Advice: own the narrative, co-locate with water/clean-energy production.

• [01:31] Media trust at 19% — Down from ~80% (mid-70s) on a near-straight line toward zero by ~2030. Wissner-Gross's structural read: ad-revenue collapse forces fluff/outrage, a death spiral; call for an AI-built, open-model, open-reasoning-trace trustworthy news source.

• [01:34] Blitzy (sponsor read) — Autonomous software development with "infinite code context": thousands of specialized agents reason for hours over enterprise codebases, deliver ~80% of dev work autonomously, claimed 5x engineering velocity. CTA blitzy.com.

• [01:35] Longevity block — Russia commits $26B to anti-aging (national priority: 3D-printed tissue, transplantable organs, epigenetic reprogramming by 2030; goal ~175k lives). New Limit (Brian Armstrong + Blake Byers) raised $435M at $3.1B for epigenetic reprogramming, first indication alcohol-related liver disease, human trials next year. Context: Altos (Bezos/Milner), Retro (Altman), Diamandis's $101M (later $157M raised) Healthspan XPRIZE; Kurzweil's longevity-escape-velocity by 2033; US healthspan ~63 vs lifespan ~79.

• [01:44] Verve-102 gene-editing therapy — Single-infusion CRISPR base-editing + mRNA/LNP delivery that permanently switches off the liver PCSK9 gene; Phase 1 (NEJM) cut LDL ~62% and PCSK9 protein ~88%, sustained up to 18 months. Framed as a near "one-and-done cure for heart disease," derived from humans with a natural low-LDL mutation — opening the question of how many other diseases have analogous single-shot edits.

• [01:48] Audience Q&A speed-run — Perfect algorithm still needs massive compute (Jevons paradox); Blundin running ~170 agents (parallel particle-sim and neural-net-research competitions) as practice for the billion-agent future; catching OpenAI/Anthropic is possible but distribution is the moat; non-material abundance ≈ happiness/agency/optionality; AGI definition defers to AIXI/Schmidhuber; ~70% of unicorns now come through incubators; the backlash is genuinely anti-AI (not just anti-corporate); average person's playbook = buy AI equities/ETFs, use AI to raise earning power, or build.

Notable claims

• Anthropic confidentially filed for IPO with the SEC; Polymarket gives ~60% odds of a >$1.8T day-one market cap (~SpaceX's level). Anthropic revenue cited ~$50-60B at ~20x; ~$9.4M revenue/employee on ~5,000 staff. (Episode title separately cites a "$965B IPO" valuation.)
• ChatGPT hit 1B MAU in ~3 years; OpenAI ~62% YoY growth; Claude ~56M MAU growing ~640% YoY.
• Trump AI EO: voluntary 30-day pre-release model access to government instead of mandated regulation; Altman and Anthropic supportive.
• Microsoft shipped 7 from-scratch models at Build 2026; tuned Excel model claimed to match GPT-5.4 at ~10x efficiency; Suleyman: training compute up 1-trillion-fold, another 1000x in 3 years.
• Verve-102 Phase 1 (NEJM): ~62% LDL reduction, ~88% PCSK9 protein reduction, single infusion, durable ≥18 months.
• Russia committing $26B to longevity as a national priority; New Limit raised $435M at $3.1B valuation.
• US data centers ~150B gal water/yr vs California almonds ~1.3T gal/yr; Nadella claims near-zero ongoing data-center water use.
• Media trust at ~19%, down from ~80% in the mid-1970s.

(All figures are panelist on-air assertions, paraphrased; treat as claims to verify, not confirmed facts.)

Guests

• Peter Diamandis — host (XPRIZE, Abundance360, Fountain Life, Singularity).
• Dave Blundin ("DB2") — co-host, Link Ventures founder/MD; the investing voice.
• Alex Wissner-Gross ("AWG") — co-host, in-house "polymath"; physicist/entrepreneur.
• Emad Mostaque — guest, founder/CEO Intelligent Internet (ex-Stability AI), dialing in from the UK.
• Dr. Don Mucalem — Fountain Life Chief Medical Officer (appears only in the sponsor segment).
• Salim Ismail — regular co-host, absent this episode.

Sponsorship

Sponsored — two paid third-party reads plus extensive house promotion. Fountain Life [00:52] (full-body MRI / early cancer detection; CTA fountainlife.com/peter) — note Diamandis is a Fountain Life founder, so this is also a self-interested plug. Blitzy [01:34] (autonomous enterprise software dev; CTA blitzy.com). House promos throughout: Diamandis's MetaTrends newsletter / moonshot Substack summary, the $2M Gemini XPRIZE "build with Gemini" hackathon, Abundance360 longevity trip, and the $101M/$157M XPRIZE Healthspan. Bias: uniformly accelerationist, pro-private-sector, anti-regulation/anti-tax; multiple panelists have direct financial stakes in the companies and IPOs discussed (acknowledged on-air re: the DNA-letter signers).

Mapping against Ray Data Co

Strength: STRONG. This episode is unusually load-bearing for RDCO across three fronts.

• AI-infra investing thesis. The Anthropic IPO is a primary catalyst for RDCO's memory/chip-fab capital-cycle and AI-infra positioning (founder places RDCO at Phase 2). The "innermost loop" framing (robots → data centers → fabs → chips → models) is the same capital-cycle logic RDCO already tracks via [[2026-05-12-diamandis-innermost-loop-ai-infrastructure-thesis]] and hyperscaler-capex anchors. Blundin's "AI economy funded to live within itself" claim — IPO capital recirculating into the agent economy without touching legacy finance — is a thesis worth stress-testing against RDCO's smart-money and capex anchor data. Robotics-as-bigger-than-GPUs and "robots don't depreciate" are concrete, testable forward claims to log.
• Agent-deployer positioning. RDCO's COO agent runs on Claude; Anthropic going public (financial disclosure, capital, durability) de-risks RDCO's core vendor dependency. The "who is your Jarvis / the one coordinating agent next to you" thread is exactly RDCO's product posture — the single orchestrating agent is the prize, and the panel agrees CAC for that agent will go vertical.
• Operating-model validation. Blundin's healthcare.com anecdote (domain experts who can't code now shipping products) and the solopreneur-doubling data are the RDCO operating model: a single domain-fluent founder + AI building real products. This is direct external corroboration of the L4→L5 "unhobble the COO agent" thesis ([[2026-05-30-moonshots-ep259-pope-leo-gpt55-altman-jobs]] continues the jobs thread).
• Regulatory landscape. The Trump AI EO (voluntary 30-day pre-release sharing, no heavy regulation) sets a permissive near-term US deployment environment for any AI-built product surface — relevant to RDCO's investing and shipping posture. The Bernie Sanders SWF / "universal basic equity" thread is lower-relevance but worth a watch-flag.

Action-worthy: nothing requires founder decision. Candidate follow-ups for idle cycles — (1) cross-check the Anthropic ~$50-60B revenue / 20x and the $965B-vs-$1.8T figure gap against [[2026-06-03-alphasignal-anthropic-ipo-filing]]; (2) note the robotics-depreciation and "AI economy lives within itself" claims as thesis inputs for the next investing-build cycle.

Related

• [[2026-06-03-alphasignal-anthropic-ipo-filing]]
• [[2026-06-03-ark-invest-spacex-ipo-wave]]
• [[2026-05-12-diamandis-innermost-loop-ai-infrastructure-thesis]]
• [[2026-06-03-moonshots-kurzweil-agi-close-not-here]]
• [[2026-05-30-moonshots-ep259-pope-leo-gpt55-altman-jobs]]