If I’m investing in frontier AI I would underwrite it mostly on new discovery and its applications like medicine because coding and most of knowledge work is quickly being saturated by sub-frontier models.