“can we just jump a little bit to practical advice on sort of prompt ops or working with foundation models?”