Notes from the engineering, not a blog.
Three pieces at launch. Each one states a control we can implement. This replaces the old agentic-notes link.
November 2026
How we design approval gates for AI agents
An approval gate is a named person and a specific action. Silence is not a yes.
November 2026
Read-only by default: the permission model behind our agents
The agent starts with the reads a new analyst would get, and no writes.
November 2026
What the IT incident demo does, and what is simulated
The sequence and the approval are real mechanics. The outage is not.