A fun observation: pulling models sends ~200mbit of progress updates to your browser