It was a science-filled day, with conversations with Price-Whelan, Jagannath, Maffucci, Blanton, and Jiang about various projects underway. But the best moment was when Dmitry Malyshev (NYU) came into my office and proposed a project (out of the blue) to determine the statistical properties of the distribution of sources too faint to be individually detected in the Fermi data. He had some good ideas about working in Fourier space, and I suggested some overly complicated ideas in real space. There ought to be a lot of information in the data set, because the photons are individually detected and there isn't a huge background. That is, there are many sources there at a significance too low to claim but which should show themselves in aggregate. One approach we did not talk about is doing the whole hyper-prior Bayes thing to find them.