> I have seen vibecoded intern prototypes that took half a day to vet and dismiss because they were completely useless.

They weren't useless, they proved if the direction that the prototype was exploring was worthwhile. I've personally made many completely shit code prototypes in the years before we had LLM's, of course they weren't magically production ready, that's not the point of a prototype.

The example I'm thinking of did not. It was just the dumbest way to execute an idea we knew was easy to execute (scanning a parameter space).