Basically, Feynman starts with microscopic approach. We want to describe interaction where each piece of mass/energy is attracted to each other. So lets introduce a boson (whole number spin - 0, 1, 2, ...) to carry interaction. Then he talks about spin properties, rejects spin-1 boson (aka photon), somehow rejects spin-0 boson (I don't remember arguments) and start building microscopic theory of interacting via spin-2 massless bosons. Because they carry energy there is a self-interaction. He builds interaction rules, Feynman diagrams, and then goes to classical limit, where he derives macroscopic classic GR from microscopic quantum interactions. That's what I remember from reading this book. Apparently he did it on the fly, lots of errors in derivations which were noted and corrected in footnotes. Nevertheless, very entertaining approach...
