AlphaFold2 revolutionized structural biology by accurately predicting protein structures from sequence. There remain however several outstanding tasks for which AlphaFold2 does not work well or is in the process of maturing. These include the prediction of (i) multimeric complexes and interfaces, (ii) structurally impactful mutations and allelic variants, (iii) alternate protein conformations and conformational ensembles, and (iv) binding of small molecule ligands to proteins. This talk will provide a status update on the current state of the art for tackling these challenging problems and key future directions, including the open platform, OpenFold, for developing new variants of AlphaFold.