[JeremyHerrman]

it's tough for people here to admit, but AI is good at summarizing and that includes commit messages and PR fields

but that does NOT mean you get to skip understanding the code & whipping the AI in line during implementation

my preferred workflow:

- think hard about the problem before touching the keyboard

- back and forth with the AI, swatting down bad implementation ideas and poking holes

- settle on a detailed implementation plan

- let the AI go for however long it needs to (usually minutes, sometimes hours)

- review, iterate, test

- "this looks good, commit in focused chunks and create a PR"

- review commits & PR summary, hand edit for clarity

[meerita]

This is what I do mainly. I spend more time designing the solution, iterating by hand when it matters and anything that should be automated, let it be.

[blooalien]

Yeah, one developer I recently had a chat with on Chatto (recently featured here on Hacker News) actually wrote the core code and project layout by hand before invoking the LLM and found that he had to hand-hold it a lot less because he had given it solid foundation to work from. That design stage you speak of there and the "iterating by hand when it matters" is a big part of what really helps the models do their part of the job a lot better. That's what the "vibe coders" (and often their bosses) don't seem to understand (yet), but they'll figure it out eventually I'm sure. Or maybe not. I guess we'll have to wait and see how it all turns out. In the meantime, we can have fun playing with the new tools they've given us. ;)