The practical counterpoint to Dex is Jax, by many of the same authors: https://github.com/google/jax https://github.com/google/jax I’m optimistic that we’ll ev

The practical counterpoint to Dex is Jax, by many of the same authors: https://github.com/google/jax https://github.com/google/jax I’m optimistic that we’ll eventually be able to figure out how to build practical tools for deep learning (and scientific computing for broadly) based on these sorts of ideas. The biggest challenge may be trying to fit them into Python syntax!