It's an interesting post but I disagree with him on (at least) four counts:
(1.) Nobody really thinks (well, maybe Cochrane does) that the NK consumption function is exactly right. We bring in intertemporal optimization because people do plan for the future, but people still think that current or at least near/medium term income matters for psychological/behavioral reasons if nothing else,
(2.) so with MPC = 0 if government spending doesn't crowd out you've still got a case for government spending. The problems only come in if government crowds out. Insofar as government commits to major investment programs, and insofar as we think investment is governed by expectations of future demand, you get positive multipliers again (of course, by a different mechanism than the Old Keynesian multipliers).
(3.) The whole MPC/income thing goes out the window when we consider credit constrained, paycheck-to-paycheck people, which leads me to
(4.) It's true politicians talk like Old Keynesians. I think economists do this less - they focus on my (2.). When we do talk about MPC it's almost always with reference to something like (3.).
So it's a good post laying out the issues, although I don't buy all the "you silly Old Keynesian" implications. I agree with Krugman that the Old Keynesian intuitions hold up surprisingly well, even if the mechanisms do change.