Post-2002 developments that come to mind that hadn't occurred at the time of writing: anything involving Fourier analysis of the boolean hypercube, like Ryan Williams' circuit lower-bounds & the Unique Games Conjecture, and work on pseudorandomness.

More broadly, I'm always fascinated how interaction of (probabilistic) functions is the underlying concept across such a broad swath of recent developments in complexity theory. Looking forward to seeing where the next paradigm shift comes from