Spent the whole day hacking, with Long and Richards and (returned from DC) Bloom (Berkeley). We had the standard experience that when you try to apply your good ideas to real data, you make many improvements. That is, real data are a much better testing environment than fake data! We really want to have everything run (that is, have the model optimized on every one of the lightcurves) by tomorrow. Also Brendan Brewer (UCSB) showed up and we ran into Lisa Randall (Harvard) at lunch.