Everything in this series can be done by hand — the chapters show you how. Prompt Optimizer does the writing for you: paste your rough request, pick a mode, copy back the finished artifact. Here's the whole flow, animated.
You don't prepare anything. Type the wish exactly as it exists in your head — vague, unpunctuated, half-formed. The entire premise of the tool is that turning wishes into specs is a skill (Chapter 01), and the tool has it so you don't have to warm it up. Then pick your mode: Optimize when you want one strong attempt, Loops when the work deserves passes and checks (Chapter 03 is the ten-second test).
✗ "make my candle instagram better"── becomes ──────────────────────────ROLE: marketing strategist for handmade brandsDELIVERABLE: 30-day Instagram plan, candle businessCONSTRAINTS: gift shoppers 24–45 · 5 hrs/weekFORMAT: week-by-week table
Optimize reads your wish, makes the four decisions — role, deliverable, constraints, format — and hands back a complete spec. Anything it couldn't infer it makes a sensible call on, visibly, so you can correct it in a word rather than write it from scratch. Copy the result into whatever AI you already use; the spec is plain text and works everywhere.
YOUR MISSION 30-day plan · every week has 3 posts + 1 goalEACH PASS draft — or fix only what failed — then re-gradeCHECK YOURSELF 4 yes/no checks · quote the proofLOOP LOG passes · failures · lessonsSTOP WHEN all pass OR 4 passes → honest reportRULES stuck twice → ask me
Loops mode writes the entire six-label file from Chapter 02 — mission with a number, binary checks, the log, both exits — tuned to your task. Paste it into your AI and watch the run: draft, self-grade with quoted proof, targeted fixes, honest stop. The animation above is the shape of every run — red checks becoming green ones, without you in the middle.
Nothing. The free tier gives you optimized prompts every day, no card required — enough to feel the difference on real work this week. If you're running it constantly, paid tiers raise the ceiling and add the stronger model. That's the entire pitch; the chapters you just read are the longer one.
Bring the roughest request you have — the one you've been re-prompting all week. That's the fair test.
Open Prompt Optimizer