Are we chatting directly with the model? Maybe the interface has changed. With long term use the likelihood of hitting edge cases is higher and maybe that is a

Are we chatting directly with the model? Maybe the interface has changed. With long term use the likelihood of hitting edge cases is higher and maybe that is a cause as well for what users are seeing. People probably ask more vague questions over time. I might have done that. I have never experienced the amnesia problem in v3.5 though, that v4 clearly has. Just repeating incorrect answers that you ask it not to give. I did not have access to v4 in march so I can't do that comparison.