Before Module 4
You handed off one bounded job and reviewed what came back. Next week the model starts running while you are not watching, so arrive knowing which of your tasks could stand it.
Delegate one real job, with a card
Pick one piece of work this week that you would normally do yourself, and hand it off with the same three-line card from the session: the job, the model's side, your side. Add Alex's assumption disclosure and Vinnie's inline citation to the Project first, both from The Trust Boundary reading. Bring what your review caught, or what it would have cost if you had let it stand.
Required Bring to Module 4Sort a week with the consequence test
List ten things you did this week, with AI or without. Mark each one by who finds out if it is wrong and nobody catches it: only you, someone outside, or a commitment. Then take the ones marked only you and run Michele's scoping test: is the trigger easy to identify, is failure easy to detect, and is getting it wrong low risk? Bring the one that passes all three. It is your first candidate for Module 4's gate.
Required Bring to Module 4Live with your brief
Run your brief each morning this week from your brief prompt. When it gets something wrong, correct the brief prompt or your corpus once, specifically, and note whether the correction held. Bring one change you would make to the design now that you have used it.
RequiredKeep feeding the corpus
Your brief now runs on it. Add to it this week, including at least one self-meeting.
OngoingOptional: take your design brief to Claude Design
In a conversation, have Claude help you write a short design brief for your reader: who reads it, what matters most at a glance, what can be collapsed, and the feel you want. Then hand that brief to Claude Design and let it propose. Try two directions, keep one thing from each, and bring back what it did with your brief that a conversation could not. Use your own reader or an invented example, not client material.
OptionalYour reading
The Trust Boundary
Every job you hand to AI has a line in it: what the model may do on its own, and what you check before it counts. Where to draw it, and how it moves.
Time to read: about 8 minutes
Building with AI: the four Ds in practice
Read the Delegation section: assigning AI the work it does well while keeping the work that needs your judgment, shown on a real build.
Time to read: about 10 minutes