I linked to a podcast (but here it is again) where Sean discussed the implications of using XLA and how that reduces the translation back and forth between pyth

I linked to a podcast (but here it is again) where Sean discussed the implications of using XLA and how that reduces the translation back and forth between python and the numerical libraries underneath. It sounds like the Axon team made some choices that are paying off in the in that respect and that other teams are migrating towards. https://podcast.thinkingelixir.com/102 https://podcast.thinkingelixir.com/102