Make Your First Useful Thing

Outcome: in one sitting, turn a real task you actually have into a finished, useful thing — and leave knowing how you did it, so you can do it again.

Who it’s for: anyone who’s opened an AI chat box, felt the blank-cursor freeze, and closed it again.

The big idea: you don’t learn AI by reading about it. You learn it by making one useful thing, start to finish. This sheet is the scaffold. Fill it top to bottom — about 15 minutes.

---

Step 0 — Pick one real task

Not a test. Something you actually need today. Stuck? Start with a fool-proof first win:

• ☐ Draft the awkward email you’ve been putting off.
• ☐ Summarize a long thing (a document, an article, a thread) into the 5 points that matter.
• ☐ Explain a confusing thing (a bill, a contract clause, an error message) in plain English.
• ☐ Plan something (a week of meals, a trip, a study schedule) from a few constraints.

Your task: ________________________________________

Step 1 — Write the brief (the 5 parts)

Use The Prompt Anatomy. Don’t skip Context and Example — they do the heavy lifting.

Role:    You are a ____________ who ____________.
Context: Situation: ____________. It's for ____________; it matters because ____________.
         Material: [paste only what's needed — no private details].
Task:    ____________  ← one thing.
Format:  ____________ , about ____ long, ____ tone.
Example: Good looks like: ____________.
If unsure: say so — don't guess.

Step 2 — Run it, then check it (60 seconds)

Paste the brief. Read what comes back with The “Is AI Wrong?” Checklist open:

• ☐ Any facts, numbers, names, or quotes? Check the riskiest one against a real source.
• ☐ Does it actually do your task, in your format — or is it a lecture around it?
• ☐ Anything you’ll send, publish, or can’t undo? Verify it independently before you act.

Step 3 — Make it better (one pass)

First answers are first drafts. Change one thing and re-run:

• Wrong shape? → fix Format (“make it 4 bullets, not paragraphs”).
• Missed the point? → add Context (“it’s for my landlord, who’s skeptical”).
• Wrong vibe? → show an Example of good.
• Don’t bother with “are you sure?” — that makes the AI cave, not improve.

Step 4 — Ship it + keep the recipe

• ☐ Use the result (with light edits — that’s normal and fine).
• ☐ Save the brief that worked. Paste it somewhere reusable. Next time is 30 seconds, not 15 minutes — that’s the start of your “second brain.”

What I’d reuse next time: ________________________________________

---

Worked example (in four lines)

Task: explain a scary-looking medical bill. Brief: Role: a patient billing advocate who explains plainly. Context: here’s my itemized bill [pasted, name removed]; it’s for me, and I don’t know what’s normal. Task: tell me what each charge means and which to question. Format: a short table + 3 bullets, plain English. If unsure, say so. Check: confirmed the two biggest charges against the hospital’s published price list. Better: added “flag anything that looks duplicated.” → A bill I finally understand, in one sitting.

Safety note

In Step 1, never paste private data — names, account numbers, health or legal details. Describe it generically or use placeholders. See What Not to Paste into AI.

Next

Do this once and you’ve got the whole loop. Build your own menu of go-to tasks, and keep the briefs that work — that’s how 15 minutes becomes 30 seconds.