i keep wondering if copilots are making devs feel more confident for the wrong reasons. not in a doom-y way, just that it’s way easier to get to “looks good” without spending as much time sitting with the code.
has anyone else noticed that? sometimes i trust the output less even when it’s technically fine, which is a weird place to be.
I’ve seen that with quick fixes at work. The suggestion looks tidy, you merge it, and then later you realise you never actually thought through the weird cases.
That’s the bit that bothers me. It skips the annoying pause where you ask, “what if this comes back null?” or “what if the API is just having a bad day?” Without that, the code can be fine and still feel a bit dodgy.
I had one last month where the change passed tests, but I still didn’t trust it until I’d walked through the timeout path by hand. Do you trust copilot output any more once you’ve got tests around it, or does it still feel off?
:: Copyright KIRUPA 2024 //--