Low token and highly specific

I finally remembered a technique I figured out by accident, I was messing around with the story generation GPT is capable of but, I was getting annoyed with the irrelevant extra meta talk. I asked to have an idea described in 3 sentences, (this number I reached after tweaking it a bit) it was just the right amount. I had to set this up by referring to what I wanted in three similar prompts, (similar to setting your functions or classes, but with three separate short prompts) basically repeating myself. But when I started using the three sentences method i gained very easy to digest but info packed outputs. It was pretty consistent, I did have to start each prompt with “in three sentences” or similar. It seems like extra work, but it ultimately saved me thousands of possibly wasted tokens. And it easly combines with almost all other prompt techniques. So it will maximize all techniques, obviously there maybe uses i am not thinking of. But I am generalizing.