Proving lower bounds on complexity and inventing algorithms that have theoretically better complexity but are most times useless in practice (for example the O(n^2.376) algorithm for matrix multiplication).