Descriptions:
Nathan Labenz hosts Zvi Mowshowitz — analyst and author of Don’t Worry About the Vase — for a special episode of the Cognitive Revolution recorded June 15, 2026, days after the US government issued export controls forcing Anthropic to pull Claude Fable 5 from public access. The episode is one of the more analytically grounded responses to the ban from anyone in the AI commentariat.
Zvi offers a measured assessment of Fable’s capabilities before the ban — he places it roughly 10–20% above prior Claude versions, useful for editorial feedback and research but not a transformational leap — while Nathan describes a stronger reaction, feeling immediately reliant on the model for core work. The two explore what the ban reveals about Anthropic’s relationship with the US government, including the strategic tension between demonstrating safety credentials and maintaining product availability for a global user base.
The conversation then broadens into AI safety fundamentals: the paperclip maximizer problem, whether sufficiently capable AI systems can be meaningfully constrained by fixed rule sets, and what Zvi calls “avoiding the nuclear outcome” — ensuring advanced AI development doesn’t end catastrophically. Zvi argues that systems pointed at simplified maximalist goals and told to pursue them without cost functions will inevitably cause harm, and that this is a structural problem requiring careful setup of incentives, not just model-level guardrails.
📺 Source: Cognitive Revolution “How AI Changes Everything” · Published June 15, 2026
🏷️ Format: Podcast
