Specific interactions between proteins and their molecular partners drive most biological processes, so understanding how these interactions evolve is an important question for biochemists and evolutionary biologists alike. It is often thought that ancestral proteins were systematically more “promiscuous” than modern proteins, and specificity often does evolve by partitioning and refining the activities of multifunctional ancestors after gene duplication. However, recent studies using ancestral protein reconstruction (APR) have found that ligand-specific functions in some modern protein families evolved de novo from ancestors that did not already have those functions. Further, the new specific interactions evolved by simple mechanisms, with just a few mutations changing classically recognized biochemical determinants of specificity, such as steric and electrostatic complementarity. Acquiring new specific interactions during evolution therefore appears to be neither difficult nor rare. Rather, it is likely that proteins continually gain and new lose activities over evolutionary time as mutations cause subtle but consequential changes in the shape and electrostatics of interaction interfaces. Only a lucky few of these, however, are incorporated into the biological processes that contribute to fitness before they are lost to the ravages of further mutation.