The idea that code can be data feels right.
I'm in that weird boat where I 'get' how and some of the possible consequences of thereof, but never got to the point where it struck me as useful for something I'm doing or plan to make. Functional programming turned out to be the more intuitive one, for sure, but this kind of Lisp-ism trips the hell out of me.
Worked through a bunch of books and articles since that post, mostly from bfv's and (IIRC) flaga's recommendations, but at this point, I think it's a good time hang my gloves. Haskell can be a performance beast, Scheme was fun, but Lisp repeatedly proves itself to be beyond me right about the time you call for some metalinguistic gizmo to write a macro that's (IMO) less transparent than Duff's device. But, according to the author, makes everything better. All this effort did is made me see those are rarely empty promises, but that's about it. Damn shame, really.
In physics, I found that people seem to have a natural aptitude for either differential equations or linear algebra, but more rarely both.
Yeah, it was one of those observations that became obvious in hindsight. Just as to why we are rushed so 'senselessly' to at least start Spivak's Calculus on Manifolds while we're still freshmen. Then, around the time QM showed the transition of Heisenberg's and Schrodinger's pictures, it clicked for real. Took a while, less than Lisp though. :)