If you read the tutorial, the broadcast iterators and GPU kernel are often implicit. As the CUDA array type hints means you don't need to explicitly define standard operations unless doing something custom.

I hope one day Julia does a cleaner version of Scala or Erlang/Elixir OTP languages. Clustering on OTP was certainly an area even seasoned gray beards tried to avoid. =3

https://youtu.be/lg-710Kk5CM?t=20