I've read the first three, and attempted every single exercise (besides the few he marks as "suggested research problems"). It took me four years to finish, an

I've read the first three, and attempted every single exercise (besides the few he marks as "suggested research problems"). It took me four years to finish, and I had a MSCS when I started. The first volume is probably the most math-intensive, but also the easiest, since a lot of the math is review. V1 is split into two parts: a pure math half, and then a programming half. All of the programming examples and exercises are done in a custom assembler that he calls MIX, but you can find emulators online to run your programs. I've read CLRS and Knuth and I definitely enjoyed Knuth a lot more - there's just something about his writing and exposition that makes the books compelling. They dare as well as entice you to understand them. It's not really as proof-heavy as CLRS IMHO, and TAOCP covers different ground that CLRS does; the only real overlap there is that volume 3 covers sorting algorithms.