Iterating and Refining Results

Iteration is one of the most important concepts when working with AI. Instead of treating Claude like a vending machine where a perfect prompt produces a perfect answer, think of it as a conversation partner.

In many cases, you don't know exactly what the ideal result looks like at the start. The first response gives you something to react to. From there, you can refine, adjust, and improve the output based on what you learn.

The goal isn't to get the perfect answer on the first try. The goal is to work toward a better answer through multiple rounds of feedback and refinement.

Back to the Wedding Speech

Let's return to Helen's wedding speech. You've got Claude's suggested cuts and improvements. Now Helen says:

It gives us something to react to. Conceptually it looks good - but Claude has tried to render it as a Word document, and personally I find that a bit awkward to work with. So:

Now we can edit alongside the AI. Reading through the draft, Helen notices something. She doesn't like the segment that describes Ruth as serious. It sets the wrong tone. She tells Claude exactly that.

Notice what Claude does next, instead of just redrafting, it asks her why the idea of Ruth being serious bothers her. That gives Helen a chance to open up and explain. Two important things come out of this: