Course contents
Anatomy of a Work Run
Context → plan → execution → artifact: what happens between your prompt and the finished file.
Every Work run has four stages. Context gathering: the agent pulls what it needs — files you attached, apps you @-mention, connectors it auto-suggests. Planning: for bigger jobs, Plan mode presents the step-by-step approach for your approval before anything executes. Execution: the agent runs on a cloud computer — searching, reading, building — for as long as the job takes. Artifact: the finished file lands in your side panel, openable on every platform you use.
Plan mode is the stage beginners skip and experts lean on. Approving a plan costs thirty seconds; discovering after a twelve-minute run that the agent misread your intent costs the whole run. For anything with real stakes or ambiguity, read the plan, correct the misunderstanding there, then let it run. It's the same review instinct as proofreading an email — just moved to the front, where it's cheap.
Knowing the stages also tells you where to intervene. Wrong inputs? Fix context (attach the right file, @-mention the right app). Wrong approach? Fix it at the plan. Wrong details in a good structure? Let it finish and use the editing loop — which gets its own lesson.