A guy at a Chicago ai meetup reviewed my chatbot and said my prompts were "polite to a fault"
I had been writing these long, gentle prompts with please and thank you, and the model kept giving vague answers. He showed me his version where he just said "list 5 reasons this fails, be blunt" and the output was way sharper. Do you add manners to your prompts, or did you drop them too?
So when you switched to the blunt version, did you change anything else at the same time? Like did you also cut the extra context or background you were giving it? Curious because I tried dropping the please and thank you stuff and my results barely changed. But then I also shortened my prompts from like a full paragraph down to two sentences and that made a way bigger difference than the manners part. Maybe the politeness is not really the thing that matters, maybe it is all the padding around it. What does his blunt prompt actually look like compared to yours?