Section 01

Writing

Essays and notes on AI, DevOps, cloud native systems, and the craft of technical writing.

49 posts · 2018–2026

Latest

Building AI Agents That Refuse to Guess

An agent that can take action is not automatically safe to trust. These are the patterns I use to make coding agents inspect first, ask for intent, gate side effects, and stop when they lack confirmation.

Archive

Everything, most recent first.

Command Palette

Search for a command to run...