Loose Control

The talk examines design trade‑offs between free‑form agent decision‑making and deterministic workflows, covering state memory, reinforcement learning, and supporting frameworks.