Haha I was talking about this to a colleague the other day. That if it was like 1987 and we wanted to use a fast Fourier transform, we would pull a book off the

Haha I was talking about this to a colleague the other day. That if it was like 1987 and we wanted to use a fast Fourier transform, we would pull a book off the shelf like numerical recipes in C hoping it was there and implement what we find there. Then after a couple weeks of making sure that worked we would realize it's not fast enough so we would probably do things until it does what we want. Today you just use any one of the one liner libraries that offer it like scipy, pytorch, etc. but that's amazing because it offers us a sophisticated solution that we can use out of the box to design our own sophisticated solutions.