The best AI prompts don’t describe what I wanted, they describe what success looks like.

Most people prompt like they're giving instructions to a contractor. Specifying every step, every constraint, or every assumption. But the moment you anchor on your own framing, you've already biased the model.

I stopped leading with specific ideas, and instead described the areas I wanted to explore, contrast, or challenge.

Outcome-oriented prompting forces reasoning from first principles rather than validating yours. You get back thinking you wouldn't have reached yourself. Try it!