Debbuging, for example, is harder than programming. And the difficulty is related to the 'amount of code' you are debbuging.

This means it is pretty easy to verify the Copilot's output when he just spits out a couple of lines at a time. But you will have a pretty bad time when you ask ChatGPT for a complete service. Things will not work out and the time you take to fix it will be greater than what you saved in the first place.