Economies are so high dimensional and non-constant that many features of models cannot be derived by prior reasoning, intrinsically involving empirical discovery and requiring theory evaluation. Despite important differences, discovery and evaluation in economics are similar to those of natural science. Fitting a pre-specified equation limits discovery, but automatic methods can formulate much more general models with many variables, long lag lengths and non-linearities, allowing for outliers, data contamination, and parameter shifts; select congruent parsimonious-encompassing models even when there are more candidate variables than observations, while embedding the available theory in costless search; then rigorously evaluate selected models to ascertain their viability. Interdependence makes it essential to tackle all the complications jointly, and automatic selection can do so on a scale well beyond the powers of humans alone. Live computer illustrations using Autometrics show the remarkable power and feasibility of this exciting approach.