The agentic loop — trigger, goal, verify, stop

The conceptual core of loop engineering. A loop hands an agent a verifiable goal and
a trigger, then lets it run until the goal is met — replacing the one-shot prompt
that needs a human to keep nudging it. Students learn the anatomy every solid loop
shares: a trigger (a schedule, an issue label, a CI failure, or a human command), a
clearly defined "done", an isolated workspace, a verification signal, and explicit
stopping rules. The lineage is taught so students see this is not new magic but the
maturation of a long line — ReAct (the in-loop reasoning of EB-1) through single-agent
"ralph"-style runners to today's supervised orchestration loops.