AlphaFold is probably the greatest recent advance in therapeutics R&D. So much more work to do, but AlphaFold and related methods have already changed the game.

There are things our current combo of LLMs and data are bad at. There are reasons why AlphsFold works really well, but we can't, for example, do similar with non-canonical amino acids. I am collecting examples of significant problems our current AI tech can't solve (possibly because we lack deep data) if anybody has any they want to share.