A guy in a Discord server said my prompts were lazy and it actually stung
I have been messing around with local AI models since March and I thought I was hot stuff because my prompts were short and quick. Then last week a guy in a small Discord server read one of mine and just said "you are writing commands, not context, the model has no idea what you want." I got defensive at first but he was completely right, so I started adding a role, an example of the output I wanted, and a note about who the answer was for. My test prompts went from getting maybe 40 percent usable answers to like 80 percent just from that change. The weird part is it takes me longer to write each prompt now, but way less time fixing bad answers after. Anyone else here change how they build prompts after getting called out like that?