By skyforbes I don't do much vibe-coding…at least not in the way i hear about.
I'm old – 30 years professional development – and find myself using AI as a discovery tool for new ways to do the same thing and for refactoring of small things.
I added this instruction to my saved instruction list (you can also just put it as a first instruction before you start too):
- LXL = line-by-line simulated code execution
Now, whenever I get code from AI that might be questionable as to its quality, I simply respond with: Please lxl your suggested code for quality and correctness.
Changing the text after LXL can also change your results too, so experiment with that.
The number of times LXL causes AI to come back with a "oh, i didn't do that part quite right" is very high. No surprise, but now you don't have to wait until a build and run session to find out.
Have fun out there!