Bernie Demands the Labs Stop, Wall Street Turns GPUs Into Bonds, Grok 4.7 Takes #1 ft. Emad Mostaque
Peter Diamandis and the panel cover nine stories spanning longevity, AI filmmaking, frontier model races, GPU financialization, AI safety politics, and flying cars. Diamandis reports live from the $101M Healthspan X Prize finals, and the group debates whether GLP-1 drugs are already delivering major longevity gains. AI-generated filmmaking takes center stage as Higgsfield's $2M feature film and China-dominated video models signal Hollywood's cost structure collapsing, while xAI's rapid Grok 4.5-4.7 cadence draws scrutiny over its Cursor-acquisition strategy. Nvidia's $500B compute-financing deal with Wall Street giants sparks a mortgage-backed-securities comparison, and Bernie Sanders' AI pause letter, following Stanford's AI-designed bacteriophage, triggers sharp panel pushback in favor of 'defensive co-scaling' over pausing. The episode closes with Zuckerberg's personal-superintelligence vision, a wave of eVTOL industry consolidation, and an AMA touching alignment, career advice, and abundance mindset.
Diamandis reports from the $101M Healthspan X Prize finals in Utah: 800 teams competed to reverse 20 years of functional aging in cognition, muscle, and immune function; $1M awarded to 10 finalist teams toward an eventual $80M grand prize, to be won by 2030.
GLP-1s and longevity escape velocityLongevity▶ 16:22
The panel discusses GLP-1 drugs (Salim is trying retatrutide) as potentially the first mass-market longevity drug, with Alex citing informal analysis suggesting third-gen GLP-1s could already deliver ~70% of longevity escape velocity in some people.
AI collapses Hollywood filmmaking economicsAI▶ 19:59
Higgsfield's $2M full-length AI movie with licensed celebrity likenesses (vs. $20-100M normal cost), China's dominance of text-to-video leaderboards, and LTX 2.5 running on a MacBook all signal AI filmmaking is now cheap and fast enough to disrupt Hollywood.
Moonshots' own Future Vision X Prize has 5,000 entrants creating 3-minute AI trailers and treatments; top 5 finalists will compete live at Moonshots Live on September 25 for a $10M prize package.
Grok 4.6 launch and xAI's catch-up strategyAI▶ 40:26
xAI's Grok 4.6 ties GPT-5.6 on the Artificial Analysis Intelligence Index at $2-6/M tokens; the panel debates whether xAI's Cursor acquisition (for its reasoning-trace data) plus Nvidia compute lets Grok 4.7 leapfrog the frontier.
Nvidia and Wall Street financialize GPU computeEconomy▶ 51:56
Nvidia partners with Apollo, BlackRock, Blackstone, Brookfield, and KKR to mobilize $500B+ in third-party capital for AI infrastructure, creating a new compute-backed securities asset class the panel compares (with caveats) to mortgage-backed securities.
A new non-transformer architecture (Dragon Hatchling / BDH) is climbing the ARC-AGI benchmark at a fraction of transformer compute cost; Alex Wissner-Gross reads the paper live on-air and dismisses it as an inelegant 'hot mess' unlikely to generalize.
Bernie Sanders' AI pause letter and Stanford's AI-designed virusGeopolitics▶ 1:11:57
Sanders sends a formal letter demanding Anthropic, Meta, and OpenAI pause AI development, citing Stanford's Evo 2 model designing novel bacteriophages; the panel argues pausing is futile and pushes 'defensive co-scaling' plus regulating DNA/RNA synthesis hardware instead.
AI content watermarking and EU labeling backlashAI▶ 1:33:49
Anthropic begins embedding invisible statistical watermarks in Claude's text output and the EU mandates AI-content labeling icons; within 24 hours tools emerged to strip the watermarks, and Alex calls both moves regressive cookie-banner-style overreach.
Zuckerberg publishes a 6,500-word vision statement and open-sources on-device models Muse Glimmer and Muse Spark 1.2, plus a community investment fund for data-center host towns; the panel debates whether this reflects genuine consumer-AI ambition or protects Meta's ad business.
eVTOL industry consolidationTransportation▶ 1:56:37
Archer Aviation acquires Boeing's Wisk Aero, Insitu, and SkyGrid AI; the panel reviews the top five flying-car players (Joby, Archer, EHang, Beta, Eve), their pricing, safety case versus helicopters, and the real-estate-unlocking implications of cheap short-hop flight.
AMA: alignment, careers, and abundance mindsetOther▶ 2:06:48
Listener questions cover whether smarter models get more or less aligned, whether any lab can reach permanent competitive escape velocity, advice for skilled trades workers, and whether the singularity has a resource cost given talk of scarcity.
Predictions made
openEmad Mostaque: Grok 4.7 will surpass Opus and become the #1 ranked frontier model.
“we could put our avatars live in the next month or two at a level that you won't be able to tell.”
Your call:
openDave Blundin: A major compute or physics breakthrough will make today's chip-fab investments, like Tesla's Terafab, largely obsolete before they finish being built.
“it's almost inconceivable that we get to 2030 without some major breakthrough that makes everything that we've built so far kind of moot.”
Your call:
Numbers that matter
$101M Healthspan X Prize, $80M grand prize, $20M already awarded800 teams entered; $1M split among 10 finalist teams at the Utah finals
Global cost of age-related disease: $20 trillion/yearout of a $120-130 trillion global economy
85% of healthcare costs occur in the last 5 years of lifecited from Daniel Kraft
US life expectancy rose from 47 (1900) to 79 todaybut healthy only to average age 63, 16 years spent in poor health
Third-gen GLP-1 studies suggest ~70% longevity escape velocity in some subpopulationAlex Wissner-Gross's back-of-envelope estimate, explicitly not medical advice
Higgsfield's AI feature film cost $2M, $1M of it compute, with 28 people over 4 weeksvs. $20-100M and 12-18 months for a traditional celebrity feature film
Grok 4.6 ties GPT-5.6 at 61 on the Artificial Analysis Intelligence Indexpriced at $2-6 per million tokens input/output
Grok parameter scaling: 1.5T (4.5/4.6) to 2T (4.7), roadmap toward 10Tparameter roadmap discussed by Emad Mostaque
Nvidia mobilizing $500B+ in third-party capital with Apollo, BlackRock, Blackstone, Brookfield, KKRfinancing framework for AI infrastructure build-out
71% of Americans don't want a data center in their backyardmore opposition than to a nuclear plant nearby
Stanford's Evo 2 model synthesized ~300 phage designs, producing 16 viable phages against E. colifirst AI-designed bacteriophages that don't exist in nature
45% of dementia cases are entirely preventable; brain age improved 26% with healthy-living interventionFountain Life data cited by Dr. Don Musalem
Joby targets $3/seat-mile; EHang EH216-S costs $330,000 and flies autonomously at 40 sites in ChinaeVTOL industry pricing and operational status overview
Alex Wissner-Gross calls it inelegant and unlikely to generalize, but Pathway AI's Zuzanna flagged it beating transformer cost-performance on ARC-AGI, worth reading the primary paper to judge independently.
🕳️ Stanford Evo 2 bacteriophage design
The first AI-designed viable bacteriophages not found in nature triggered Bernie Sanders' pause letter and a biosecurity debate over whether to regulate models or the DNA/RNA synthesis hardware.
Nvidia, Apollo, BlackRock, Blackstone, Brookfield, and KKR are creating a new $500B+ asset class tied to GPU cash flows; the panel explicitly compared the risk to mortgage-backed securities.
🕳️ GLP-1 drugs as a longevity escape velocity driver
Alex's claim of ~70% LEV from third-gen GLP-1 studies in some subpopulations is a striking, informally-sourced number that could reshape how the panel thinks about near-term life extension.
🕳️ Zuckerberg's 'personal superintelligence' strategy vs. Meta's ad-driven incentives
The panel split on whether Meta's open on-device models and community investment fund reflect genuine consumer-AI ambition or just protect the ad-and-distribution business, worth tracking follow-through.