Imagine doing AI development in waterfall. You spend weeks writing your prompt, when you think you have it perfect, only then do you submit it to the AI. Then you wait a week or so, and see what it produced, expecting it to be exactly what you wrote.

Or, do you tell it the basic functionality you want, test it out, then add feature after feature that you want, sometimes dropping them and sometimes adding new ones that you thought of as your worked.

> Or, do you tell it the basic functionality you want, test it out, then add feature after feature that you want, sometimes dropping them and sometimes adding new ones that you thought of as your worked

the problem with this is long term maintainability. it works - and the engineer understands how it works - but a) the AI does not prioritize cleanup/organization/naming, and b) there's a blind spot/boiling frog type of phenomenon that can prevent the engineer from spotting the growing problem. the codebase becomes recognizable only to them. the engineer sees all features working, all bugs fixed, 90% test coverage, and submits it for a PR.

the engineer tasked with reviewing the PR will treat it as slop.