The times it's wrong has become vanishingly small. At least for the things I use it for (mostly technical). Chatgpt with extended thinking and feeding it the docs url or a pdf or 3 to start you'll very rarely get an error. Especially when compared to google / stack exchange.
No comments yet.