Editor View vs Agent Manager
Two surfaces, one rule: stay close to the diff in the Editor; delegate at scale from the Manager.
The Editor view is for work where you want to stay near the code: a tricky refactor you're pairing on, a bug you're actively reasoning about, a review of what an agent just changed. It behaves like the agentic IDEs you may already know — chat panel on the side, inline edits, tab completion — with the full artifact system underneath.
The Agent Manager inverts the frame. It's an inbox-and-board over every agent conversation you have running, across every workspace. You spawn an agent per task, not per file: kick off three chores in the morning, review walkthroughs as they land, and drop into the Editor only when something needs your hands on the diff.
The rule that falls out: interactive, ambiguous, high-context work → Editor. Well-scoped, verifiable, independent work → Manager. Most people start 90/10 Editor-heavy and drift toward 50/50 as their briefs and knowledge base improve. That drift is the skill this course builds.