See the Difference

One prompt rarely gets you where you want to go. But each round of feedback gets you closer.

You
Write me a tagline for my bakery.
AI
"Fresh baked goodness, made with love."

Generic. Could be any bakery anywhere.

You
Too generic. We specialize in sourdough and we're in San Francisco. We want something with personality.
AI
"SF's sourdough obsession, perfected."

Better. Now it has context and place.

You
I like the SF reference but "obsession" feels a bit intense. We're more warm and welcoming than that.
AI
"San Francisco's favorite sourdough, one loaf at a time."

Now it sounds like your bakery.

Why This Works

AI doesn't know what you want. It makes its best guess based on what you give it. When information is missing, it fills the gaps with generic patterns.

Your job is to evaluate what comes back. Is it right? Is it close? What's missing? What's wrong? That evaluation is the key — it's what turns a mediocre first attempt into something good.

Each round of feedback teaches AI more about what you're actually looking for. Sometimes it takes three rounds. Sometimes it takes ten. The number doesn't matter — what matters is that you keep evaluating and guiding until you get there.

The Technique

Start simple. Look at what AI gives you. Evaluate it honestly. Tell it what's wrong or what's missing. Repeat as many times as it takes. Your evaluation is what makes it work.

Good Feedback Sounds Like

Why Evaluation Matters

Interestingly, even automated systems can iterate and improve — as long as there's a clear objective to measure against. The magic isn't in the back-and-forth itself. It's in having a target and knowing whether you're getting closer.

When you're the one iterating, you decide the objective. You know what good looks like. That's why your evaluation — your honest assessment of each attempt — is what makes iteration work.