Anthropic Files $965B IPO, Trump Signs AI Executive Order, and ChatGPT Crosses 1B Users | EP #262
The panel covers a loaded week of AI news: Trump's light-touch AI executive order (voluntary 30-day model disclosure to government), ChatGPT crossing 1 billion monthly users while Claude grows even faster, and Anthropic's confidential S-1 IPO filing with a possible $1.8 trillion first-day valuation. They dig into OpenAI's new Rosalind Biodefense tool and the narrowing of general-purpose AI capability into restricted models, a bipartisan push to force DNA-synthesis screening against bioweapons, OpenAI's pivot into robotics, and Microsoft's attempt to rebuild in-house frontier models after its OpenAI 'divorce.' They also debate Bernie Sanders' proposed AI sovereign wealth fund, dueling data on AI's jobs impact, Nvidia's new ARM-based PC chips, and declining media trust. The episode closes with three multi-billion-dollar longevity stories: Russia's $26 billion anti-aging push, Brian Armstrong's New Limit raising $435 million, and a one-shot gene-editing therapy (Verve 102) that cut LDL cholesterol 62% in a phase 1 trial.
Trump signed an EO rejecting heavy AI regulation, instead asking labs to voluntarily share new models with government 30 days before public release and directing agencies to deploy AI-powered cyber defense. Panel debates whether this is genuine policy or political theater after lab CEOs lobbied to water down an earlier, stricter draft.
ChatGPT hits 1 billion users, Claude grows fasterAI▶ 9:42
OpenAI passed 1 billion monthly active users about three years after launch, faster than YouTube, Instagram, or TikTok reached that mark, growing 62% year-over-year. Anthropic's Claude is smaller (56 million MAU) but growing 640% year-over-year, described as Apple to OpenAI's Microsoft.
OpenAI's Rosalind Biodefense and the narrowing of AGI generalityAI▶ 13:47
OpenAI launched Rosalind Biodefense, giving trusted government/public-health researchers specialized AI tools for outbreak detection and vaccine development. Alex Wissner-Gross argues this signals a broader trend of carving powerful (bio, cyber) capabilities out of general-access frontier models into restricted, exclusive-access versions.
Peter co-signed an open letter with Sam Altman, Dario Amodei, Demis Hassabis, and Mustafa Suleyman urging Congress to require DNA synthesis companies to screen orders, citing a 2017 case where a researcher mail-ordered DNA to reconstruct an extinct horsepox virus. A bipartisan Senate bill is already in play.
OpenAI is building an in-house robotics team, with Sam Altman framing it as building infrastructure and eventually a personal robot for everyone. The Sora video team, after Sora's shutdown, has reportedly been reassigned to embodied robotics.
Anthropic's IPO filing and trillion-dollar valuation speedEconomy▶ 35:22
Anthropic confidentially filed IPO paperwork, potentially becoming the first major frontier lab to go public, with Polymarket giving 60% odds it surpasses $1.8 trillion market cap on day one. Panel discusses Anthropic's outsized revenue-per-employee (~$9.4M vs Apple's $2.5M) and how fast AI companies are reaching trillion-dollar valuations versus historical tech giants.
Microsoft's in-house AI models and the OpenAI 'divorce'AI▶ 46:19
Microsoft unveiled seven in-house AI models built from scratch (no OpenAI distillation), including one matching GPT-5.4 on Excel tasks while being 10x more efficient, plus a Mayo Clinic healthcare model collaboration. Alex argues these are still mid-tier models and that Microsoft lost the frontier race to OpenAI much like IBM lost to Microsoft decades earlier.
Leiden Declaration: mathematicians push back on AI-generated proofsAI▶ 58:36
Over 130 mathematicians signed a declaration warning that AI-generated proofs can look convincing while containing subtle errors, and that AI companies could skew which math problems get studied. Panel frames it as a rear-guard reaction to AI (OpenAI) solving the Erdős conjecture.
Bernie Sanders' American AI Sovereign Wealth Fund ActEconomy▶ 1:06:49
Sanders proposed requiring OpenAI, Anthropic, and other major AI firms to contribute 50% of their stock (not cash) to a public wealth fund so all Americans own a stake in AI. Dave calls the mechanism unworkable (you can't sell that much stock without crashing markets); Alex and Emad see a real risk that any such scheme would make the labs 'too big to fail' and entrench their political power.
AI's impact on jobs: displacement vs. net creationEconomy▶ 1:16:39
Washington Post lists five policy options for AI-driven job loss (tax robots, better unemployment insurance, retraining, wealth dividends, do nothing), while Apollo's Torsten Slok argues AI is a net job creator based on actual employment data. Dave reports 2x employment growth across his incubator's portfolio as non-engineers become builders.
Nvidia's ARM-based PC chips and data center water debateCompute▶ 1:21:39
Nvidia is launching ARM-based PC processors (N1, N1X) with GPU performance rivaling an RTX 5070 laptop chip, seen as a direct challenge to Apple, Intel, and AMD for consumer AI compute. Separately, Satya Nadella and Dave push back on data-center water-use fears, noting closed-loop cooling and that US data centers use far less water annually than California almond farming.
Longevity breakthroughs: Russia's $26B program, New Limit, and Verve 102Longevity▶ 1:35:50
Russia committed $26 billion to anti-aging research targeting 3D-printed organs and epigenetic reprogramming by 2030; Brian Armstrong's New Limit raised $435 million at a $3.1 billion valuation for cellular reprogramming therapies; and Verve 102, a one-shot gene-editing therapy that permanently silences the PCSK9 gene, cut LDL cholesterol 62% in a phase 1 trial.
Predictions made
openAlex Wissner-Gross: We will know within a few months whether Trump's AI executive order struck the right balance between innovation and national security.
openPeter Diamandis: A winner will be declared in the $101 million Healthspan XPRIZE, reversing 20 years of functional aging in cognition, immune function, and muscle.
“I think what's going to happen is that every single government's going to introduce a token license and probably the most sensible thing is to tax them on GDP eval.”
“This gives support to when Elon said triple-digit GDP growth by 2030.”
Your call:
Numbers that matter
Polymarket gives 60% odds Anthropic surpasses $1.8 trillion market cap on IPO day oneAnthropic confidentially filed IPO paperwork with the SEC.
ChatGPT reached 1 billion monthly active users about 3 years after launchCompared to YouTube's 10 years, Instagram's 8 years, and TikTok's 5 years to reach the same milestone.
OpenAI year-over-year growth holding at 62%Growth rate discussed alongside the 1-billion-user milestone.
Claude at 56 million monthly active users, growing 640% year-over-yearRoughly 10x OpenAI's growth rate, described as still early days for the category.
Price of equivalent AI intelligence expected to drop 100x over 18 monthsCited as a claim from Sam Altman.
$100,000 spent on mail-ordered DNA in 2017 to reconstruct an extinct horsepox virusCited as a cautionary example motivating DNA-synthesis screening legislation.
Time to reach $1 trillion valuation: Apple 42 years, Google 21 years, SpaceX 24 years, OpenAI ~10 years, Anthropic ~5 yearsIllustrates accelerating pace of trillion-dollar valuations in tech.
Anthropic generates ~$9.4 million revenue per employee on ~5,000 employees, vs. Apple's $2.5M and Google's $2.1MUsed to argue AI companies extract roughly 4x the value per employee of top legacy tech firms.
AI training compute has increased 1 trillion-fold, with another 1,000x expected in the next 3 yearsQuoted from Microsoft AI's Mustafa Suleyman at Microsoft Build 2026.
US data centers use about 150 billion gallons of water per year vs. California almond farming's 1.3 trillion gallons per yearCited by Dave Blundin to argue data-center water criticism is overblown.
Public trust in media has fallen to 19%, down from a peak of about 80% in the mid-1970sFramed as an unstoppable, near-linear decline.
Nvidia's N1X chip: 20 CPU cores, 6,144 CUDA cores, GPU performance on par with an RTX 5070 laptop chipNvidia's first ARM-based PC processor, positioned against Apple, Intel, and AMD.
Russia committed $26 billion to anti-aging research, targeting saving 175,000 lives by 2030Targets include 3D-printed tissues, transplantable organs, and epigenetic reprogramming by 2030.
US average lifespan is ~79 years while average healthspan is only ~63 yearsA roughly 16-year gap spent in declining health, cited to argue for longevity investment.
New Limit raised $435 million at a $3.1 billion valuationCo-founded by Brian Armstrong and Blake Byers, developing epigenetic reprogramming therapies.
Worth digging into
🕳️ "Mythos," OpenAI/Anthropic's classified cyber-vulnerability-discovery model
Referenced repeatedly by Alex as the trigger for the executive order and as evidence that AGI's generality is shrinking into restricted models, but never explained in detail on the show.
🕳️ Global environmental DNA (eDNA) biosurveillance as a bioweapon baseline
Alex and Emad propose sequencing environmental DNA everywhere as a baseline to detect novel/engineered sequences, calling it an 'enormous hole' in current biodefense efforts, but no such program is described as existing.
🕳️ AIXI and Schmidhuber's formal definition of AGI
Alex points to Marcus Hutter and Jürgen Schmidhuber's AIXI framework as the rigorous, information-theoretic alternative to benchmark-based AGI definitions the panel otherwise relies on.
🕳️ Verve 102's path to approval
Described as a one-shot cure for the leading driver of heart disease (high LDL) via a permanent gene edit, with striking phase 1 results (62% LDL reduction, 88% PCSK9 drop).
🕳️ Russia's $26 billion state-scale anti-aging program
A national government committing tens of billions to longevity science with explicit 2030 targets (3D-printed organs, epigenetic reprogramming) is a major geopolitical and scientific story that got only brief treatment here.