Glad to see that Hamiltonian / Hybrid Monte Carlo is gaining more interest, though I guess it's supposed to have had more interest in physics than statistics for a while. Given how much better these algorithms scale, I think they should get a lot more attention. The description of Parallel tempering was pretty good, I had heard it described before but didn't get it. I get a sense for why it's exciting to people. The section on Reversible Jump MCMC made me realize I don't understand model selection problems very well at all.