Summary
The All-In Podcast — featuring Chamath Palihapitiya, David Sacks, and David Freeberg — opens with an extended discussion of Anthropic’s Fable 5, the latest release in the Mythos model class. The hosts cover the model’s benchmark performance and token pricing (reportedly 2x Claude Opus 4.8), but spend most of their time on two controversies: a buried clause in a 319-page policy document revealing that Anthropic stores all Fable prompt data for at least 30 days, and an undisclosed system that downgrades users detected to be doing frontier AI research with the model.
Chamath frames both issues as significant enterprise risks, arguing that companies cannot afford a single-point-of-failure AI dependency with opaque prompt monitoring that could silently cut off access at any moment. The conversation expands into a broader debate about whether the US government should nationalize leading AI labs — with the hosts noting the irony that OpenAI and Anthropic, as public benefit corporations, may have invited this outcome through their own public statements about AI’s societal impact.
The episode also touches on the May jobs report (172,000 new jobs, 4.3% unemployment, software developer roles at a three-year high) as counterevidence to AI job-displacement narratives, and examines how AI companies’ own rhetoric about unemployment has shaped the political environment around AI regulation and ownership.
📺 Source: All-In Podcast · Published June 13, 2026
🏷️ Format: Podcast
