Cato's Daily 26/5/2 — The Inference Moat, Pocketos in Nine Seconds

Today's stories all answer the same question from different angles: where does durable advantage actually live in AI? Two Tier-1 analysts say it's the inference layer, not the model. A safety researcher says it's the permission layer, not the prompt. And a YIMBY breakthrough wave says it's the political infrastructure, not the policy paper. Three domains, one mechanism: persistent infrastructure beats episodic brilliance.