As a tensorflow dev, with several years experience writing clojure at a previous startup, I believe this is the most natural language for writing tensorflow models (or theano. basically any graph building dsl that has any real complexity and has been bolted onto python).
Suddenly TF's cond and while_loop and context controls all fit naturally into the language.
Suddenly TF's cond and while_loop and context controls all fit naturally into the language.
My favorite technical post in recent memory.