Tell AI what “good” looks like — before it does the work
The single highest-leverage prompting move: get AI to define success criteria before it writes a word.
What you’ll do
- Open Claude or ChatGPT in two tabs.
- In one, ask for the task straight. In the other, ask it to define what good looks like first, then do it.
- Compare the two outputs side by side.
Common questions
Do I need any technical skills to do this?
None at all. You're just typing two different prompts and comparing the results. If you can send an email, you can do this lesson.
Do I need to pay for anything?
No. The free versions of ChatGPT, Claude or Gemini work fine for this lesson — open one in your browser and you're ready.
What does 'defining success criteria' actually mean in plain English?
It means asking the AI to tell you what a great version of your task looks like before it tries to write it. Instead of jumping straight to the answer, it pauses to set the bar — and that one step noticeably improves the first draft.
Why does this produce better results than just asking normally?
When AI defines what 'good' looks like first, it sets its own quality target before writing. The same way a builder checks the plans before picking up a tool — it just produces better output.
Is eight minutes really enough to see a difference?
Yes — the lesson is a side-by-side comparison, so you'll see the difference in the same session. Most people notice it immediately on the first try.