“We don’t need likelihood functions; we just need to know how to simulate from (…) We don’t need models with sufficient statistics; we just need summary statistics (…) We don’t need to be Bayesian; we just need to be approximately so. We don’t need theory to tell us our method works; we just need