GPT4 has been outright wrong for me only a handful of times. I know this because I typically have an academic paper for the topic I'm working through next to me and knowledgable friends who I discuss things with as well. I'm rarely using GPT4 for things like syntax or APIs which can be easily looked up elsewhere. I also rarely ask it to write code.

I know mileage may vary, but I just haven't had this often quoted (but I suspect actually less first hand experienced) situation when it comes to development.