> Since then, all improvements came at a tradeoff, and there was a definite flattening of progress.

Idk, that sounds remarkably similar to these AI models to me.