The Trade-Off Between Context and Quality

GPT takes too long, so I always choose GPT for the drop down even though it's annoying. There comes some point where the amount of context that it's built up starts to cause problems and I start to see a degradation in the quality of output. At that point, I'll usually sort of reboot and just start a fresh conversation with like, okay, here's where we've gotten. Let's use that as our new starting point. And then it starts to give me better answers again.