AI coding assistants, productivity boost or skill rot?

[Neon Grace]

I have been using AI assistants for code daily for over a year now and I honestly cannot decide if they are making me better or worse. Boilerplate that used to take an hour takes minutes, and that part is undeniably great

But I caught myself last week unable to remember basic syntax I have written a thousand times, because I now just accept the suggestion. It feels a bit like satnav brain, where nobody can read a map anymore. The skill did not disappear, it just went soft from lack of use

Where do you all land on this? Is it just the calculator debate again, where the tool absorbs the boring layer and frees you for design, or is something more important being lost this time?

[AEWCallum93]

Calculator debate, one hundred percent. Nobody mourns doing long division by hand at work

[ShawnMichaels07]

Difference is a calculator never confidently gives you a wrong answer that compiles. The failure mode matters

[Solo Elizabeth]

I make juniors on my team explain any generated code line by line in review. If they cannot, it does not merge

[Hollow Vulture]

That policy sounds good on paper but does it survive deadline pressure at your place?

[Seb83]

Barely, if I am honest, but the intent keeps standards from collapsing entirely

[Finley]

For me the real risk is not syntax rot, it is architecture rot. The assistant happily builds on a bad design without ever saying this whole approach is wrong

[Dragon]

Has anyone actually measured their own defect rate before and after? Vibes are doing a lot of work in this thread

[Di82]

I tracked mine roughly, more code shipped, about the same bugs per feature, so slightly worse per line but better overall throughput