My AI prompt library used to be a mess, now it's a living doc with version tags

I used to keep every prompt in my head or in a notes app with zero structure, so every week I'd retype the same request for summarizing meeting notes and get a different result each time. Then I started adding a version number to each prompt, like v3 or v4, and keeping the older versions below the current one in the doc so I can roll back if an update ruins the output. Last month I hit a spot where a prompt I'd used for 4 months suddenly gave me garbage after a model update, and because I had the old version saved I could compare and fix it in like 10 minutes instead of blaming myself. That little habit of tagging each version with the date and model name has saved me at least 6 hours of rework so far, and it feels like having a manual for my own brain. Do you name your prompts or just trust that the latest one is always best?

0 comments

0 Comments

No comments yet

Be the first to share your thoughts on this discussion.