Kathryn Johnston (Columbia), Ana Bonaca (Yale), Andreas Küpper (Yale), and Adrian Price-Whelan (Columbia) stopped in for the day to discuss all things streams-in-the-Milky-Way-Halo. We discussed objective functions, probabilistic inference, action–angle space, chaotic and integrable potentials, biases and functional testing of inferences, tidal stripping, and other matters of great importance. We spent quite a bit of time talking about how to make more realistic or asymmetric potential models and fit them, and how to test the evil done by the assumption (that almost everyone makes) of integrability (which cannot be true). We also discussed work by Sanderson and Helmi (Groningen) that makes use of information theory to do inference in a similar context; indeed information theory appeared multiple times in the conversation. We decided that there are many overlapping papers to write. My job in all this is appearing to be: Turn a sensible data analysis idea into a probabilistic generative model such that we can fire up emcee. Love that!