By Faltings' theorem, any curve over Q of genus at least two has only finitely many rational points—but the bounds coming from known proofs of Faltings' theorem are often far from optimal. Chabauty's method gives much sharper bounds for curves whose Jacobian has low rank, and can even be refined to give uniform bounds on the number of rational points. I'll discuss Kim's non-abelian analogue of Chabauty's method, which uses the unipotent fundamental group of the curve to replace the restriction on the rank with a weaker technical condition that is conjectured to hold for all hyperbolic curves. I will give an overview of this method and discuss my recent work with Ellenberg where we prove the necessary condition for any curve that dominates a CM curve, from which we deduce finiteness of rational points on any superelliptic curve.