Definition
Plain language
The part of an AI helper that decides what to do next and, crucially, when to stop.
As stated in the literature
The outer control process in an agent that chooses successive actions (e.g. which search to run) and applies a stopping policy that determines when the task is considered complete.
Why it matters: It determines whether an AI keeps working usefully or either quits too early or spins forever without finishing the task.
For example, an AI research assistant keeps running new searches on a question and only halts once it decides it has gathered enough to answer.
Heard on the show
“And the iteration trajectory for this scenario is the kind of thing that makes the agentic loop feel real.”Episode 027 — When AI Agents Build the Serving Stack: A Bet on Bespoke Infrastructure