Closer, but still not there. In side by side tests on subjects I knew something about (or could do a quick search and confirm) ChatGPT did a much better job at retaining the thread of the conversation, adjusting stylistically to my requirements, and not giving outright incorrect information.

Maybe I just tested in Gemini’s blind spots.