Code? not much, other than small functions/classes/prototype libraries to get started, but I've often used it to figure out where code was that I was concerned with in huge project code bases and analyze where some of the edges of interfaces are without digging for a few hours. Copilot can give a decent summary of where to look in a couple of seconds instead of a half hour of marking what I think are important sections and jumping around/grepping
If someone is using AI effectively, there's often no way to tell that they're using AI at all. Toupée fallacy etc.
Code? not much, other than small functions/classes/prototype libraries to get started, but I've often used it to figure out where code was that I was concerned with in huge project code bases and analyze where some of the edges of interfaces are without digging for a few hours. Copilot can give a decent summary of where to look in a couple of seconds instead of a half hour of marking what I think are important sections and jumping around/grepping
It is best used for yack shaving in my opinion. Anything other than that and I feel like I cannot trust its output.
Its Code Generators all the way down...