Member-only story
Stanford Just Proved We’ve Been Prompting AI Wrong This Whole Time (The Right Way Is Shockingly Simple)
Researchers tested 10,000+ prompts and found the perfect formula. It’s 3 sentences long and increases quality by 47%. Here’s exactly what they discovered.
I’ve written thousands of AI prompts.
Blog posts. Code. Analysis. You name it, I’ve prompted for it.
Turns out, I’ve been doing it completely wrong.
And so have you.
I asked ChatGPT five jokes about coffee.
It gives the same joke. Five times. Sometimes word for word with minor variations.
Not creative variations. Not paraphrases. The exact same punchline, copy-pasted like a broken record until I wanted to throw my laptop out the window.
Then I changed eight words in my prompt.
Suddenly, I got five completely different jokes. Each one original. Each one creative. Each one something I’d actually want to share.
What just happened?
The $2 Million Question
Stanford researchers just spent two years and millions in compute credits answering one simple question:
