In chatgpt, I continued on old conversations, rather than start a new one. So if I asked about martial arts, last month, I can just doto the end of that conversaion and ask more. Then it will at least remember that area and last questions without having to go through the same initialisation sequence again.
You can also instruct chatgpt what "type of answes" you want. I have noted there that I genereally want detailed answers, that focus on mechanisms, and that I always what chatgpt to provide an url list of references to back up it's answers. I also instruct it to avoid causal or ambigious language. This at least makes it better.
The problem I noticed is that chatgpt seems to have a limited "memory" in the sense of how "wide perspective" it has. When asking about complex phenomena, it often truncates the reasoning ignoring details - unless you in the question explicitly add wordings and phrases that you know are always used in for example papers where the details are dicussed. So just like you need a certain search technique on google, on eventually learns how to ask the questions, so that you get not just the blunt generic answers from AI, but more sharpened and specific answers. For MA, I noticed it confused kicks and punches at times, but that is probably because the labelling of techniques vary between styles, so in the training data set, the terminology isn't consistent either.