Who he is
Zvi Mowshowitz is an independent writer and analyst who publishes "Don't Worry About the Vase," a blog and newsletter focused on frontier AI. He is not affiliated with any AI lab — his value comes precisely from being an outside observer who reads everything, synthesizes it quickly, and says what he actually thinks.
Why people read him
When a major AI lab drops a system card at midnight, Zvi often has a detailed analysis up within hours. His weekly roundups — numbered sequentially, past issue 170 as of mid-2026 — serve as a running record of the frontier: what shipped, what it means, what the safety disclosures actually say, and what the community is getting right or wrong. For IT staff, researchers, or anyone trying to keep up without reading every lab blog and preprint, his digest is a practical shortcut.
What he covers
His work falls into a few recurring categories:
Model releases. He publishes multi-part breakdowns of major releases — system card analysis, capability assessment, community reactions — for models including Claude Opus 4.7, Claude Opus 4.8, Claude Fable 5, Claude Mythos 5, GPT-5.5, GPT-5.6, and GLM-5.2. These aren't cheerleading; he assesses strengths and weaknesses directly.
AI governance and regulation. He has been one of the most vocal independent commentators on the shift toward government oversight of frontier AI. He reported early on the White House ordering Anthropic to halt expansion of access to Mythos, coined the framing "the prior restraint era" for the emerging regulatory posture, and analyzed a Trump executive order requiring pre-release AI testing. When the U.S. government compelled Anthropic to take down Fable and Mythos entirely — reportedly following a jailbreak disclosure just three days after Fable 5's release — he covered it in real time.
Model welfare. This is a topic most analysts skip. Zvi treats it seriously: he has published dedicated posts on the moral status and potential wellbeing of AI models alongside his capability coverage of Claude Opus 4.7, Opus 4.8, and Fable/Mythos. He continued reviewing Fable and Mythos's welfare features even after the models were taken down, writing in present tense about products he could no longer access.
Agentic coding. He runs an ongoing numbered series tracking Claude Code, OpenAI Codex, and the broader agentic coding landscape — at least eight installments as of mid-2026.
Broader AI culture and policy. His scope extends to OpenAI's proposed federal AI governance framework, the Catholic Church's papal document on AI, cybersecurity implications of frontier model releases, and the question of what Anthropic actually is as a company.
His analytical style
Mowshowitz writes with a strong point of view. He doesn't hedge for the sake of appearing neutral. When he thinks a policy is bad — like the White House's case-by-case access regime for GPT-5.6 — he says so and explains why. When he thinks mainstream media is missing the story, he says that too. This directness is part of why he has a loyal readership: you know where he stands.
Recent developments
The period from late April through late June 2026 was unusually eventful even by frontier AI standards. Mowshowitz covered a rapid succession of Anthropic model releases (Opus 4.7, 4.8, Fable 5, Mythos 5), the government-forced takedown of Fable and Mythos, a new executive order on AI testing, and the emergence of what he calls the "prior restraint era" — a potential shift to requiring government approval before releasing highly capable models. His weekly roundups during this stretch read as a real-time chronicle of a regulatory inflection point.
Why it matters for non-specialists
If you're an IT professional, a policy staffer, or a curious non-expert trying to understand what's happening in AI without a PhD, Zvi's work is one of the most efficient ways to stay informed. He does the reading, flags what's important, and explains the stakes — including the uncomfortable questions about safety and governance that lab press releases tend to soften.
