I put up a new Ask Good Questions pair today around something I think AI-assisted development makes especially important.
Guide:
https://askgoodquestions.dev/guides/the-moving-target
Field note:
https://askgoodquestions.dev/field-notes/when-done-wasnt-ready-to-release
This pair is about what happens when a project gets close enough to release that the real question stops being “does it work?” and becomes “should this actually ship?”
That sounds like a small distinction, but it really isn’t.
AI is very good at helping us get to functional. It can help build features, fill in repetitive work, and move things along fast enough that visible progress starts to feel like proof that we’re nearly done. The danger is that once that momentum builds, it gets easier to accept slowdowns, rough interactions, or architectural friction that didn’t seem too serious when each one was viewed in isolation.
Then you finally step back and walk through the product as a whole, and suddenly the target has moved.
The issue is no longer feature count. It is release readiness.
The guide talks through that shift and why “done is better than perfect” is still useful advice, but incomplete on its own. Done is better than perfect, but ready to release is what counts.
The field note is the story side of it. It came out of a recent stretch of work where the feature set was in good shape and the finish line looked close, but taking a broader view made it clear that some core pieces were stepping on each other in ways that needed to be understood before anything should ship.
The main takeaway:
AI can help you get to done faster. It still takes a programmer to decide whether done is actually ready.
I’m curious as to whether others here have had that same moment with AI tools, where everything looked fine until you stopped looking at the parts and started judging the whole.
Charles