I attended one of his talks (but not the ones you mention), during which he described how he was approaching the topic. He said he was reading the most important papers on sat problems in chronological order, and implementing each of the algorithms he found along the way. At the time of the talk, he was half-way through the ‘80s, IIRC.
A lesson in commitment and perseverance for me. And a very funny talk as well (Knuth likes jokes!)
knuth only touches on the condensed matter (survey, beliefprop) in like one section in the whole book, i recall. beautiful presentation, nothing about, say, the d1RSB and s1RSB transition or anything like that. try montanari's book
I'm very interested in SAT solvers, as this has deep connections to statistical mechanics and condensed matter physics.