Working with Co-Pilot, I learned early on that in order to maximize its effectiveness you have to be consistent since it is just prediction. Lately, I've been playing with structuring my projects differently, building the requirements accordingly, and then I use a LLM to generate the initial code and next the automated testing. Then it performs the initial code review and testing compliance and compares that against to the the written requirements. And of course I use the LLM's to help me assemble the code/scripts/project. It's a great way to enforce discipline in your process since you get this "Peer" that is working along side you. It can't do it on its own, but you can certainly do 10x the work with it by your side. You can take this whole concept pretty far when you consider next integrating this into your CI/CD pipelines. And if you are working in a more powerful development environment, it's unreal what you can do.

Developers won't be replaced by AI; they'll be replaced by other developers using AI to work faster and more effectively.