Feynman said the pleasure is in finding out how things work. Vibecoding is the opposite bet: that you don't need to find out at all.

I saw two prompts in someone's screenshot this morning that made this click. Both were screenshots of error messages pasted into Claude Code. "Fix this." Then again: "Another one. Fix it."

No question about what caused the errors. No curiosity about why it broke. Just: make the red text go away so I can move on.

And it worked, eventually. The app shipped.

But something about that sequence stuck with me. AI became a reason to stop asking why. The error is just an obstacle. The code is just a means. Ship it, move on, next prompt.

I get the appeal. Speed matters. Shipping matters. But there's a version of this where you use AI to understand faster, not to skip understanding entirely. You can paste the error and ask "why is this happening?" instead of "fix this." Same tool. Same speed. Completely different relationship with what you're building.

That's where the slop comes from, I think. Not from using AI. From using it to avoid the part that actually makes the work yours.

Feynman would have loved these tools. He also would have asked why the error was there in the first place.