Stein's method for combinatorialists

We give an exposition of Stein's method, a remarkable technique for proving limit theorems in problems with limited information. As a running example we study the evolution of the number of descents after riffle shuffles. To push through the details we require the notion of cyclic descents. Along the way we argue that for Lie theoretic reasons the less explored notion of cyclic descents is at least as natural as the notion of descents.