I think it was easy to follow if you focused on the aspects of general interest (= what he was doing in principle, e.g. divide-and-conquer). But sure, if you wanted to understand exactly how the merging of "bitonic globs" worked, then no, that's not easy to follow.
I remember thinking when he started talking about merging the "bitonic blobs" that "Yea, yea, sure, you can divide-and-conquer, but it's just about as much work combining two of those partial solutions as solving the original problem! There's got to a be a better way...", and I basically stopped listening to the details.