Judgement: trusting the agent vs. driving yourself.
Late evening, kitchen table, the same table as Module 0, a very different Sam. BookNook has real users, a test gate, a tidy git history, and an agent she works with daily. Tonight's list has three jobs on it. She could hand any of them to the agent, pair through them in chat, or write them herself.
The last skill this course can teach is the one that never automates: choosing. Try it cold.
Check your intuition
Three jobs tonight. (1) Rename getBooks to fetchBooks across the project, tests green, everything committed. (2) Migrate every member's saved wishlist to a new format, one way, no undo. (3) Fix a crash in async code Sam can't confidently read yet. Which one is safe to hand entirely to the agent?
Three settings, one dial.
Delegate, hand the agent the job and review the diff at the end. Pair, work it through in chat, deciding together. Drive, write it yourself, asking questions as you go. What moves the needle is stakes × reversibility × understanding: cheap, undoable, readable → delegate freely. Costly, permanent, or past your reading level → slide towards drive. Watch it move.
This lesson is part of the full course
The free lessons are just the opening chapter. Unlock the rest of Code with AI to keep going.
AcademyLv 1 · 0