I find co-pilot to be useful. It forces me to write a lot of comments and the comments drive the code. For new personal projects, chatGPT can get some ideas going fast, but I end up changing a lot of it. ChatGPT does show me a lot of new libraries and parts of languages I didn't know about before. I wouldn't use LLMs on code review because I need to use my human brain to check the work of co-pilot that my coworkers are using. Code review and QA are the last line of defense against LLM shenanigans.