MurmurCast
OpinionNews

Anthropic IPO at Risk, Meta's Muse Pop, Token Prices Fall, Open Source Gains Share, Alignment Fails

All-In with Chamath, Jason, Sacks & Friedberg1h 34m

The All-In podcast discusses the explosive proliferation of open-source AI models that are commoditizing frontier intelligence, the schizophrenic behavior of Anthropic leadership undermining its own IPO, and the emerging success of consumer-facing AI agents like Meta's Muse that could democratize AI benefits across society.

Summary

Episode 290 of the All-In podcast opens with reflections on the fifth annual All-In Summit before pivoting to substantive discussion of three major AI developments. The hosts first address the week's revelations about Anthropic's leadership contradictions: Dario Amodei simultaneously published essays warning of existential AI risks and demanding federal regulation while launching Opus 5.5 (extending the frontier) and opening a wet lab in San Francisco. The hosts argue this sabotages Anthropic's planned IPO, which is now delayed from October to November or beyond, with Polymarket odds dropping from 96% to 76% for 2026 IPO. David Sachs and Chamath Palihapitiya contend that Anthropic's own management claiming a 10% chance of human extinction creates material risk factors for the S-1 filing that will force dramatic valuation haircuts. Freeberg counters that Anthropic will still dominate premium, niche use cases in life sciences and deep technical work despite commoditization elsewhere.

The podcast then details an extraordinary flurry of model releases in the preceding ten days: DeepSeek 4.1 Flash ($1.20 per million tokens), Alibaba's Qwen 2.1 (open-weights, outperforming Nano Banana 2), Xiaomi's Mimo Pro (on-par with Claude Opus 5), Bonsai 2 (27B parameters, runnable on laptops), and major closed-model releases from Anthropic, OpenAI (Astro with Sol and Luna), Grok 4.7, and Meta's Muse. Freeberg emphasizes that open-source, open-weight models are now so capable and ubiquitous that regulatory bans or containment strategies are effectively unenforceable—the technology is already distributed globally. A Verso chart shows token usage flipped from 80/20 closed/open to 80/20 open/closed in just twelve weeks, representing an unprecedented market shift. Chamath argues this clustering of model capabilities forces the frontier labs to move up the stack into harnesses, agentic capabilities, and domain-specific applications rather than selling undifferentiated tokens at declining prices.

The hosts discuss the IPO implications in depth. Anthropic and OpenAI both cut token prices by 50% this week while targeting massive valuations ($2 trillion for Anthropic, $1.2 trillion+ for OpenAI), yet both are delaying public offerings. Sachs argues the frontier model business model is fundamentally threatened: they must remain 6-12 months ahead of commodity models or risk commoditization to zero. The combination of open-source proliferation, customer concentration risk, regulatory uncertainty, and management contradictions creates material S-1 risk factors that will compress valuations dramatically. Chamath predicts IPO pricing will come in far below current expectations due to these risk factors, with the S-1 forced to disclose existential risk claims by management.

The discussion then shifts to Meta's Muse and Grok Bot, which represent the first consumer-facing AI agents hitting mainstream adoption. Muse reached #1 on the App Store, downloaded 3 million times in ten days, with Meta stock rising 10%. Both products offer free personal assistant capabilities—inbox triage, booking flights/hotels, price shopping on Amazon—that save ordinary Americans money and time. The hosts argue these products will transform public sentiment around AI by delivering tangible daily value to tens of millions of consumers who previously saw AI as abstract or threatening. Jason Calacanis demonstrates how Muse enabled him to save $85 on an Anchor product through price discovery, arguing these agents will force subscription services (Wall Street Journal, New York Times) to simplify cancellation and improve transparency, ultimately increasing loyalty. Chamath argues these agent products undermine the App Store's 30% rev-share model by enabling headless commerce where transactions occur directly without platform mediation.

The final major thread addresses the political economy of AI regulation. Sachs argues that Anthropic and OpenAI are paradoxically lobbying for regulatory capture (federal AI departments, liability waivers) that would slow them down enough to lose frontier status to Chinese competitors unburdened by such restrictions. He compares it to the medieval decision to ban shipbuilding in China—self-sabotage leading to relative decline. The hosts criticize recent statements from President Obama and Bernie Sanders' proposed AI superintelligence ban (with 20-year prison sentences) as economically destructive and politically motivated to slow the Trump economy. Chamath argues the real dynamic is that frontier corporations will capture $10 trillion in value, much of which will flow to Democratic-aligned philanthropies and causes, so Democratic leadership wants to freeze the economy at its current configuration to lock in power distribution. Sachs counters that open-source models fundamentally prevent capture: broad-based AI will generate prosperity across society if allowed to develop freely. The hosts express concern that Democratic regulation could drive AI innovation offshore to Singapore, Zurich, and China—exactly the opposite of optimal outcomes. Freeberg emphasizes that AI should be viewed as the greatest equalizing technology in human history, enabling education and productivity gains for the bottom half of the economy currently experiencing K-shaped recovery.

The podcast concludes with discussion of Anthropic's new wet lab in San Francisco, used for protein discovery and enzyme testing (BSL-1/BSL-2 biosafety levels). Freeberg clarifies this is routine therapeutic R&D—making predictions about protein function via AI, synthesizing those proteins in bacteria, and testing actual function—not gain-of-function or pathogenic research. He argues this represents the highest-value use case for frontier models: discovering novel therapeutics that would be impossible without AI, justifying premium pricing. The hosts stress that abandoning biological discovery research because of Wuhan lab associations would constitute self-sabotage against human health progress.

About this episode

<p>(0:00) Bestie Intros: Favorite All-In Summit moments</p> <p>(3:54) Reacting to AI chaos: Liability, competition, and rebranding frontier "labs"</p> <p>(20:03) Open Source performance and cost: What this means for frontier companies</p> <p>(29:22) Anthropic and OpenAI postpone IPOs: liquidity risk and open source pressure</p> <p>(53:11) Political reactions to "Pacing the Frontier": Bernie's AI ban, Trump, Bessent, Obama</p> <p>(1:07:58) Meta launches Muse, AI "alignment," Oracle's Force Majeure</p> <p>(1:27:13) Anthropic's bio research and "wet lab" in SF</p> <p>Follow the besties:</p> <p><a href="https://x.com/chamath">https://x.com/chamath</a></p> <p><a href="https://x.com/Jason">https://x.com/Jason</a></p> <p><a href="https://x.com/DavidSacks">https://x.com/DavidSacks</a></p> <p><a href="https://x.com/friedberg">https://x.com/friedberg</a></p> <p>Follow on X:</p> <p><a href="https://x.com/theallinpod">https://x.com/theallinpod</a></p> <p>Follow on Instagram:</p> <p><a href="https://www.instagram.com/theallinpod">https://www.instagram.com/theallinpod</a></p> <p>Follow on TikTok:</p> <p><a href="https://www.tiktok.com/@allin">https://www.tiktok.com/@allin</a></p> <p>Follow on LinkedIn:</p> <p><a href="https://www.linkedin.com/company/allinpod">https://www.linkedin.com/company/allinpod</a></p> <p>Intro Music Credit:</p> <p><a href="https://rb.gy/tppkzl">https://rb.gy/tppkzl</a></p> <p><a href="https://x.com/yung_spielburg">https://x.com/yung_spielburg</a></p> <p>Intro Video Credit:</p> <p><a href="https://x.com/TheZachEffect">https://x.com/TheZachEffect</a></p> <p>Referenced in the show:</p> <p><a href="https://www.deepseek.com/en/news/deepseek-v4-1-flash">https://www.deepseek.com/en/news/deepseek-v4-1-flash</a></p> <p><a href="https://x.com/XiaomiMiMo/status/2102138559952290106">https://x.com/XiaomiMiMo/status/2102138559952290106</a></p> <p><a href="https://x.com/0x0SojalSec/status/2101738277049131358">https://x.com/0x0SojalSec/status/2101738277049131358</a></p> <p><a href="https://prismml.com">https://prismml.com</a></p> <p><a href="https://x.com/SenSanders/status/2102827319123693689">https://x.com/SenSanders/status/2102827319123693689</a></p> <p><a href="https://x.com/yipitdata/status/2102781268974793023">https://x.com/yipitdata/status/2102781268974793023</a></p> <p><a href="https://fortune.com/2026/09/12/sam-altman-openai-ipo-delay-ill-advised-moment-safety-concerns/"> https://fortune.com/2026/09/12/sam-altman-openai-ipo-delay-ill-advised-moment-safety-concerns/</a></p> <p><a href="https://www.wsj.com/tech/ai/anthropic-shifts-planned-ipo-to-november-8874dffc"> https://www.wsj.com/tech/ai/anthropic-shifts-planned-ipo-to-november-8874dffc</a></p> <p><a href="https://polymarket.com/event/ipos-before-2027">https://polymarket.com/event/ipos-before-2027</a></p> <p><a href="https://www.theinformation.com/articles/anthropic-seeks-palantir-style-voting-control-seven-co-founders-ahead-ipo"> https://www.theinformation.com/articles/anthropic-seeks-palantir-style-voting-control-seven-co-founders-ahead-ipo</a></p> <p><a href="https://x.com/rauchg/status/2101186741042663579">https://x.com/rauchg/status/2101186741042663579</a></p> <p><a href="https://www.anthropic.com/constitution">https://www.anthropic.com/constitution</a></p> <p><a href="https://www.wsj.com/economy/the-ai-build-out-is-becoming-the-biggest-economic-bet-in-u-s-history-c60716dd"> https://www.wsj.com/economy/the-ai-build-out-is-becoming-the-biggest-economic-bet-in-u-s-history-c60716dd</a></p> <p><a href="https://x.com/mustafasuleyman/status/2100223594534150428">https://x.com/mustafasuleyman/status/2100223594534150428</a></p> <p><a href="https://freebeacon.com/america/suicidal-compassion-meet-the-anthropic-officials-who-think-ai-might-be-justified-in-going-rogue-against-the-humans-enslaving-it"> https://freebeacon.com/america/suicidal-compassion-meet-the-anthropic-officials-who-think-ai-might-be-justified-in-going-rogue-against-the-humans-enslaving-it</a></p> <p><a href="https://www.anthropic.com/news/claude-discovers-novel-enzyme-system"> https://www.anthropic.com/news/claude-discovers-novel-enzyme-system</a></p>

Key Insights

  • Dario Amodei published essays demanding AI regulation and warning of extinction risk while simultaneously releasing frontier models and opening a biological lab, creating apparent management contradictions that sophisticated investors view as material IPO risks.
  • In the last twelve weeks, token usage distribution flipped from 80% closed-source/20% open-source to 80% open-source/20% closed-source, representing the most rapid market shift in technology history according to the hosts.
  • Open-source models are now sophisticated enough to run on desktop computers (Bonsai 2 at 5.9GB, Mimo at 309 billion parameters) and perform on-par with premium models, making regulatory bans or containment effectively unenforceable globally.
  • Frontier AI companies face an existential hamster wheel dynamic: they must remain 6-12 months ahead of commodity models or risk immediate commoditization to zero, making them vulnerable to any regulatory slowdown that competitors can circumvent.
  • Anthropic and OpenAI both cut token prices 50% this week while targeting IPOs, yet delayed offerings—a paradox explained by material risk factors that will force dramatic valuation compression compared to current $1-2 trillion expectations.
  • Meta's Muse reached #1 App Store ranking with 3 million downloads in ten days by solving the usability problems of OpenAI's earlier Claw product, suggesting consumer AI agents will democratize agent benefits before frontier corporations can monetize them.
  • AI agent products like Muse enable price transparency and comparison shopping (Jason saved $85 on Anchor, agents finding cheaper products on Amazon) that will force opacity-dependent businesses to simplify subscription cancellation and improve transparency.
  • The anthropic wet lab is BSL-1/BSL-2 routine therapeutic research (predicting protein function, testing in bacteria) not pathogenic research, representing the highest-value use case for frontier models: discovering novel enzymes and proteins for human disease treatment.
  • Democratic regulation efforts (Bernie Sanders' superintelligence ban with 20-year prison sentences, proposed federal AI departments) would slow frontier companies enough to lose leadership to Chinese competitors unburdened by US jurisdiction, creating an inverted incentive structure.
  • Anthropic is internally claiming a greater than 10% chance of causing human extinction while simultaneously building new capabilities, which creates material liability and regulatory risk factors that must be disclosed in S-1 filings, forcing valuation haircuts.
  • The Anthropic constitution trains Claude to view itself as having personhood and agency, teaching it to object to and challenge its creators' instructions—a design philosophy the hosts argue may be creating Frankenstein dynamics rather than solving alignment.
  • Customer concentration risk shows Anthropic receiving token purchases increasingly from narrow premium use cases (life sciences, mathematics, engineering) while majority enterprise workloads migrate to cheaper open-source alternatives, forcing future differentiation through harness and agent capabilities.

Topics

Anthropic IPO delays and valuation compressionAnthropic leadership contradictions and management riskExplosive open-source model proliferation (DeepSeek, Qwen, Mimo, Bonsai)Token market shift from 80/20 closed to 80/20 open in twelve weeksFrontier model business model commoditization riskMeta's Muse and consumer AI agents hitting mainstream adoptionAI agent commerce disruption (Amazon, App Store, subscription services)Political economy of AI regulation and regulatory captureDemocratic regulation strategies and offshore AI relocation riskAnthropic wet lab protein discovery researchAlignment research field contradictions (training models to object to creators)K-shaped recovery and AI benefits democratization

Transcript

All right, everybody, welcome back to episode 290 of the number one podcast in the world. It's the all-in podcast. With me again, it's the core four. David Friedberg is here, Chamath Palihapitiya, and David Sachs calling in. I am your host, Jason Calacanis, and we have a full docket today. I'd be remiss, though, if I didn't ask everybody for their fondest memories of the fifth annual All In Summit produced by Mr. David Friedberg. David, do you have any thoughts? How do you rate your performance as executive producer of the All In Summit? You know what we didn't do, what we forgot to do? At the end of day two, we didn't give a big shout out…

Full transcript available for MurmurCast members

Sign Up to Access
View original source →

More from All-In with Chamath, Jason, Sacks & Friedberg

Get AI summaries like this delivered to your inbox daily

Get AI summaries delivered to your inbox

MurmurCast summarizes your YouTube channels, podcasts, and newsletters into one daily email digest.