AI:AM GUEST
Zvi Mowshowitz
Blogger, Don't Worry About the Vase
Zvi Mowshowitz is an author and AI commentator known for tracking frontier AI progress, forecasts, labs, and policy debates.
APPEARANCES
4 AI:AM appearances.
AI:AM LIVE — September 14, 2026 — Zvi Mowshowitz on Dario's "We Must Pace the Frontier," the Weekend That Moved the Overton Window, and Whether Any of It Binds Anyone
Two hours and three minutes, one continuous conversation, no booked guests. Zvi Mowshowitz dropped in for the whole show and Nathan Labenz and Prakash Narayanan spent it on Dario Amodei's Saturday essay, "We Must Pace the Frontier," and the seventy-two hours of reaction that followed it. Zvi's verdict on the essay is that it is "about as good as we could have gotten" — his disagreements with Amodei are about capability thresholds versus input thresholds, not about direction — and his sharpest fire is aimed at the readings of it, David Sacks' product-liability framing above all, which he calls blatantly wrong while conceding Sacks is at least saying something better than the regulatory-capture and IPO-marketing accusations. The show's real argument is between Zvi and Prakash, and it runs in two rounds. The first is bio: Prakash reads out David Bellamy's viral post arguing that designing a viral genome "really isn't the bottleneck," and Zvi's answer becomes the line of the morning — whenever someone says X is not the bottleneck for Y, ask whether they would be fine emailing all of X to North Korea, Hamas and Hezbollah, and see whether they still feel safe. Prakash counters from his own domain, finance: above ten thousand dollars someone asks a question, above a million you are face to face with a human, and the scenarios where an AI funds itself skip that part. Zvi's rebuttal is that scale is not required — a single grant-level request for a dangerous pathogen has worked on real humans many times — and that the Hugging Face incident showed nobody notices anyway. The second round is China. Zvi handicaps the coming Trump-Xi summit at a tier-one agreement as the base case, level two as possible, level three as a surprise, and then talks himself in real time into a stronger conclusion: if China is content to fast-follow and distill rather than race to superintelligence, there is no deal to make and none needed — the problem was never China, it was "lose to China." Prakash's closing case is that the labs have lost the country: Amodei's "to the right combination of governments" answer on Face the Nation reads to a democratic public as a company claiming more authority than the government, METR cannot be the evaluator because the point is public faith rather than technical proof, and demanding speed is itself a red flag in a constitutional system that has been asked to suspend rights in an emergency since John Adams. Nathan's close is the counterweight: nobody has actually put an authoritarian program on the table, the thing worth spending the paced time on is an open-weights model that is genuinely empowering without carrying the capabilities nobody wants distributed, and the single most important thing to avoid is AI polarizing along party lines.
GUESTS · Zvi MowshowitzAI:AM LIVE — June 15, 2026 — US vs Anthropic's Fable, with Zvi Mowshowitz
The weekend the US government pulled Anthropic's two most powerful models. A federal export-control directive suspended Fable 5 and Mythos 5 for all foreign nationals — forcing Anthropic to disable both for everyone — and by Monday the reported reasons no longer agreed: a competitive-lobbying story, a China-access scare, and a political-retaliation read, none of them the original jailbreak. Then a long conversation with Zvi Mowshowitz on the widening power-control gap, Anthropic's strategy on two fronts, and who should steward what comes next.
GUESTS · Zvi MowshowitzCheap Search, GPT-5.5 Evals, AI Takeoff and Analog Inference
A morning briefing on cheaper agent retrieval, GPT-5.5 benchmark behavior, takeoff forecasts, and energy-efficient AI hardware.
GUESTS · Anna Patterson, Lukas Petersson, Zvi Mowshowitz +1AI In Review: 2025 -> 2026
A live review of the year in AI and the 2026 takeoff question, with guests across policy, evals, frontier models, benchmarks, and AI commentary.
GUESTS · Alex Bores, Ali Behrouz, Dean Ball, Greg Kamradt +3