In the beginning...

I started building Prompt Lab because I was frustrated with my own "prompt hoarding" habit. I had a Notion file full of prompts copied from Twitter, but whenever I tried to use them on different models (like the new Nano Banana or Flux), they failed miserably.

I realized that "collecting" prompts is useless if you don't understand the underlying structure. I built the Image-to-Prompt tool initially just to debug my own failed generations—to see what the AI "saw" in the reference images I liked.

Once I got the reverse-engineering working, I realized I could automate the "translation" layer too. That became the Prompt Optimizer. It's been a classic "scratch your own itch" journey.