Yeah, when I was inexperienced what impressed me about Carmack was all the "cutting edge" techniques and tricks he used to push hardware to its limits. As I've became more experienced what impresses me is how iD was able to crank out so much software so quickly, have time to research these more advanced methods, quickly triage which were promising and which were not, and get them working and well optimized in production software.

I can read research and incorporate it into my code, but it takes me a while, and I usually have multiple dead-ends where it isn't until after I've tried using a technique with real data that I realize it has weaknesses the researchers didn't point out and which can't easily be fixed. I can crank out quick code, but it isn't going to be advanced or optimized. I struggle finding the right balance of how to trade between these three things, and yet Carmack and iD were able to squeeze in all three at once.