I've tried using co-pilot, codeium, claude, chatgpt and gemini. They are good for repetitive work but anything remotely complex and they are pretty terrible tbh. I find that it's quicker for me to just solve issues on my own a lot of the time than try to prompt my way through. You can very easily fall into the trap of going around in circles trying to get the answer you need with various prompts, instead of spending the time using existing knowledge to resolve it yourself quicker. That's just my personal experience but I've been a developer for a number of years now. Writing unit tests is the only place where I've found genuine value so far.