Ch. 01 — the system
Policy, not script
A chatbot answers; an agent acts. What separates them is a policy — a way of choosing actions under uncertainty, scored by consequences, not applause. Every node here is a decision; every edge, a consequence.
Ch. 02 — the harness
Skills with edges
Capabilities as portable instruments — explicit inputs, honest outputs, permissions you can enumerate. Like the five solids, the good forms are few and provable: everything else is decoration.
Ch. 03 — the interface
Trust through legibility
Agentic UI is not a chat window — it is a surface that shows the work: plans you can inspect, tool calls you can veto, evaluations you can read.