[wills_forward]

It was funny to hear the same guy warning LMMs were getting too powerful now talking about the limits of available original training data.