A question I keep going back and forth on, and I'm curious where everyone lands.

When you're building with Claude, what's your actual method?

Two camps, roughly:

Camp 1: Iterate. Small prompt, see what comes back, adjust, go again. You build in conversation. It feels fast and you discover what you want as you go, but you can end up halfway down a path before you realise it's the wrong one.

Camp 2: Plan first. You write the full spec up front. Requirements, structure, edge cases, the lot. Then you hand Claude one well-defined brief and let it run. Slower to start, but the output tends to hold together better.

I flip between them depending on the task, and honestly I'm not sure I've found the right balance. Quick experiments pull me toward iterating. Anything touching the database or auth and I suddenly want a plan written down before I let it near my code.

So, two questions:

Which camp are you mostly in? And has that changed the longer you've worked with these tools?

4 comments