SICP blew my mind the first time, but I think it's most useful when you already have some hands-on coding experience (as I did). It's why we don't start teaching math with set theory.
Yeah I taught myself how to program for about 1-2yrs, but I didn't really learn what it meant to be a programmer or the ideas unpinning programming languages until I fully dug into SICP and watched the classic 1980s videos of Sussman's class.
But I don't think I would have appreciated without my prior exposure to programming. That said, I've also watched some MIT 101 math classes and they also seem like courses better suited for people who know/appreciate the basic content (beyond typical high school math). So maybe MIT selects for kids smart enough to grasp it.
Not so much about smarts but about exposure. Most of the kids who get into MIT for math didn't start doing proofs at MIT, but in maths olympiads and such at a much younger age. Still, even given the formidable qualifications of the MIT undergraduate CS class, they probably overshot at the time with SICP. (I also read SICP for the first time well into my programming career, and completely agree with your assessment.)
But I don't think I would have appreciated without my prior exposure to programming. That said, I've also watched some MIT 101 math classes and they also seem like courses better suited for people who know/appreciate the basic content (beyond typical high school math). So maybe MIT selects for kids smart enough to grasp it.