Notes on AI colleagues that do real work.

Clear answers about persistent context, local agent architecture, human oversight, and the operating systems required to make AI work accountable.

Jul 21, 20266 min

Human-in-the-Loop AI: Put Review at the Point of Consequence

Effective human oversight is risk-based: routine reversible work continues, while consequential actions cross an explicit approval boundary.

Read field note

Jul 21, 20266 min

Local AI Agent vs. Cloud AI Agent: What Actually Changes?

Local and cloud AI agents differ less by model location than by where context, credentials, tools, and operational control live.

Read field note

Jul 21, 20265 min

What Is an AI Colleague?

An AI colleague is a persistent, context-aware agent that can carry work across tools while staying inside explicit authority and escalation boundaries.

Read field note