Field notes on agent guardrails

Short, occasional emails about the action layer: how agents run up bills, delete things, and email people they shouldn’t — and the policies, caps, and approval gates that stop it. You get The AI Agent Threat Model (PDF) on signup, a five-email walkthrough of wiring Arc around your first risky action, and after that, only when we have something worth saying. Written by the people building Arc. No growth-hacking cadence, no curated-links filler.