Statistical models connecting galaxies to their dark matter halos are quite powerful: they can be employed to constrain cosmological parameters, and inform detailed physical modeling of galaxy evolution. However, recent observational and theoretical developments have revealed that traditional statistical models such as the HOD and CLF are too simplistic to reliably model the galaxy-halo connection. In this talk, I discuss how this inadequacy presents a new challenge to the theory of large-scale structure and the precision cosmology program. To address this challenge, I present Halotools: an open-source, user-friendly python package for building and testing cosmological models of structure formation with high-resolution simulations. Halotools is community-driven, and already includes contributions from over a dozen scientists spread across numerous universities. Designed with high-speed performance in mind, the package generates mock observations of synthetic galaxy populations with sufficient speed to conduct expansive MCMC likelihood analyses over a diverse and highly customizable set of models. I conclude the talk by describing how Halotools can be used to analyze existing datasets to obtain robust constraints on galaxy formation models, and by outlining the Halotools program to prepare the field of cosmology for the arrival of Stage IV dark energy experiments.