Try the prompts.

You type a lazy one-liner at the AI, get back something generic, and blame the model. The problem is almost never the model. A real prompt has five parts: role, task, constraints, examples, and an output spec. In this exercise you take a vague request you would actually type and rebuild it into a structured prompt, then watch the quality jump. You keep judgment over what "good" looks like; the AI just helps you see the parts.

Here is a weak prompt I might lazily type on a busy day:

"[PASTE A VAGUE ONE-LINER YOU WOULD ACTUALLY TYPE]"

Do not answer it yet. First, tell me in plain language what is missing that would force you to guess, and where you would most likely guess wrong.

Most of your AI work is a handful of moves you repeat: summarize, sharpen, draft a reply, audit a document. Instead of retyping them from scratch every time, you build a small drawer of reusable prompts wired to your own context. In this exercise you pick the tasks you actually do each week and turn the four cores into prompts you keep. You stay the editor; the drawer just stops you from starting from blank.

I want to build a small drawer of reusable AI prompts for the work I actually repeat.

Here is my AI Context Document:
[PASTE YOUR CONTEXT DOCUMENT]

Here are tasks I do at least weekly:
[LIST 3-4 TASKS]

Sort them into the four cores: SUMMARIZE, SHARPEN, DRAFT A REPLY, AUDIT A DOC. If one does not fit a core, tell me which core it is closest to and why.

"It feels better" is not proof. When you change a prompt, you need a way to tell real improvement from luck. This exercise builds the cheapest evaluation that works: a few reference inputs, an answer key for what good looks like, and an A/B scorecard in a plain Google Sheet. You change one thing at a time and let the evidence decide. You stay the judge of what "good" means; the harness just keeps you honest.

Help me build a cheap evaluation for a prompt I use a lot.

The task this prompt does:
[DESCRIBE THE TASK]

Propose exactly three scoring criteria that actually matter for this task (for example: correct facts, right tone, followed the format). For each, give me a one-line definition and a simple way to score it (pass or fail, or 1 to 3). Do not give me a single vague "quality" score.

This is the point of the whole track. Not a certificate, but a one-page playbook you would pin above your desk: what you hand to AI, what you keep fully human, which tool you reach for, where your context lives, and the checkpoints that never move. In this capstone you turn everything from Modules 1 through 11 into a durable artifact in your own voice. The AI facilitates; you decide every line, because this is your playbook.

Be my facilitator for writing a one-page AI playbook I will actually reuse. Ask one question at a time and wait for my answer.

Here is my AI Context Document:
[PASTE YOUR CONTEXT DOCUMENT]

Tasks I do most:
[LIST 5-8 TASKS]

For each task, help me place it in one of three buckets and say why:
- DELEGATE: hand to AI, I review
- COLLABORATE: AI and I work it together
- KEEP HUMAN: stays fully mine, AI does not decide

If a task could hurt a person, is binding, or is something I must sign, push me to justify before it leaves the KEEP HUMAN bucket.