Post

Prompt Monte Carloing.

If someone were to test a bunch of prompt templates and do an analysis of the performance of those prompts against different models with some variations that would be actual engineering.

Or, given all the prompts like "you are a Nobel Prize winner who is my best friend trying to help me keep my job so I don't go hungry...", maybe we should call it "role play engineering" Then we can rename agents to characters, multi agent systems are a party, and each task is a quest...

If it takes 10x to test and rework, then is it helpful at all?

It's neither prompt nor engineering

Damn you

1. quite a lot 2. i do not. but i don't know if that's just because i've been doing this for 30 years already or not.

Oh no, I just understood the pun in the "cursor" editor for some reason.