Core insight: risk tolerance is contextual and should be set deliberately for each decision, not guessed in the moment. The Decision Thermostat is a tiny habit you run in under ten minutes that converts vague appetite into a simple, machine‑readable band (0–100) plus three anchoring answers: (A) Worst‑Credible Outcome and its maximum acceptable impact; (B) Likely Upside and minimal acceptable delta; (C) Recovery Window—how long you tolerate adverse drift before you act. In this episode I give the exact one‑line Thermostat token, three enforcement patterns (pre‑auth gating, automatic throttles, and a one‑click rollback leash), and two constrained AI prompts that 1) propose a starting band from past outcomes and 2) monitor live drift against the thermostat and surface the single line of evidence to trigger review. You’ll get a 7‑day pilot to apply a Thermostat to one recurring decision stream, measure how often thresholds drive action, and one immediate action: publish a Thermostat for your next pricing or feature experiment. CTA: subscribe. Stay agentic.