I've had a good experience with it in Python and typescript but not so much with Go although I think that's because the models themselves aren't very well trained on some languages. That said, what I really liked was that I was able to attach other files in the context and ask out to write tests using the same frameworks and style from the existing codebase. The tab completion guesses are the most magical. For example converting single quotes to double. Do it once in a spot and it highlights the relevant next spot to apply.