By skyforbes I am loving 3.0 pro with the best of them. I have noticed that the model has been pulling from other conversations and referencing previous things that I have searched and done deep research with. It's not exactly bad, though it's just particularly cheesy at times since it somewhat seems to be trying very hard to to appeal to things I asked it to do in other conversations. Like for example, I have been asking it to "think like a forensic accountant" in previous conversations and now it's giving me "forensic accounting insights" every other prompt, even when I don't ask for that. Has anyone experienced any hallucinations or any problems from the personalization feature? This seems to be much more prevalent than 2.5 pro. I figure you can turn this off? Not sure if I'd want to.