Notes on AI agents in production
What we learn putting AI agents into production: where they go wrong, what the regulation asks of them, and what it takes for Legal and Risk to let an agent act on its own.
Why auto mode is not enough for enterprise AI agents
Every agentic harness now ships an auto mode. Why that helps developers, and what it leaves open for an enterprise.
Carlos Lebre Ribeiro ·
What human oversight means for AI agents
Human oversight means a competent person can intervene on the decisions that matter. Why approving everything fails.
Carlos Lebre Ribeiro ·
What is a supervision harness?
A supervision harness is the layer above an AI agent that decides, independently, which of its actions may run.
Carlos Lebre Ribeiro ·
What is straight-through rate for AI agents?
Straight-through rate is the share of agent decisions that execute with no person involved and stay correct. How to measure it.
Carlos Lebre Ribeiro ·