A low-research day was saved by a great talk by Blakesley Burkhart (CfA) about turbulence in astrophysical MHD. She made an unassailable argument that if we don't understand turbulence (and we don't), we don't understand almost any astrophysical phenomenon. That is only slightly an exaggeration! And then she talked about a general framework for understanding turbulence: Simulate the hell out of it, build empirical statistics from those simulations, and use those statistics to measure (latent) physical parameters in observed systems (like interstellar molecular clouds and supernova remnants). She showed baby steps towards this ultimate formalism, but, in my way of thinking, this approach asymptotes to something like ABC or likelihood-free inference, in which we try to put posterior constraints on parameters of interest using statistical surrogates to obviate an explicit likelihood function. Right now, this seems like the only approach for something as complicated as turbulent plasmas.