Hoping for a follow-up with Sparse Autoencoders.