We don't. You could freeze model development today and we would be able to use our current models to effectively optimize existing code for years to come.