Of all the Paul Samuelson obits and reflections that I have seen, I liked this one by David Henderson best. A remarkbale man, economist, body of work as well as the state-of-the-discipline are all touched on, including Samuelson's "tin ear" re the fraudulent economic news from the Soviet Union.
Henderson touches on the tension between economists who rely on mathematics as opposed to those who look to history. Obviously, both have much to offer. But when it comes to being wrong about the most auspicious economic event of the twentieth century, the fall of communism (and the demise of Marxist economics), does that tell us something about how mathematical economics leavened by a deep understanding of history would have been helpful?
History and mathematics are each glorious, but the pendulum in economics graduate curricula has been pretty much stuck at the math extreme for many years. A pendulum is, of course, a model that fails to account for stickiness.