I will present the GLASS software, which is a framework for carrying out useful simulations for cosmology. GLASS simulations take onion-like shells of matter density and add subsequent physical effects (such as weak lensing) or observational effects (such as spatially varying survey depth) by forward-modelling. The source of the initial matter density can be simple lognormal random fields or large N-body simulations. I will discuss how we go from there to realistic simulated Stage-IV galaxy surveys such as Euclid or LSST, highlight some of the open questions, and show how this approach enables new ways to do cosmology with Simulation-Based Inference.
