About
A prompt you have not run
is a guess.
There are lists out there with ten thousand prompts on them. Nobody wrote ten thousand prompts. They collected them, and most were never run against a model once.
Why This Exists
I kept copying prompts that produced worse output than just asking the model plainly. That is the actual failure mode of prompt content, not that the prompts are bad, but that nobody checked whether they beat the obvious version.
So every prompt on this shelf has been run, compared against just asking directly, and kept only if it won. If it did not win, it is not here.
How A Prompt Gets Shelved
Run it against the model it names. Run the plain version of the same request. Compare them on something I actually needed done, not a toy example.
Then write down why it works, because a prompt you understand is one you can adapt, and a prompt you copied is one that breaks the moment the model updates.
What You Will Not Find Here
- Mega-lists of prompts nobody has run
- Prompts with no model named
- "Act as a world-class expert" openers
- Screenshots of output with the prompt cropped out
Who Writes It
Kevin Gabeci. I build software and use these daily, which is the only reason I noticed how many popular prompts lose to just asking.
Read the posts