[smaddox]

Looks like they use LLVM backend, so they'll get low level optimisation for "free". The GC is also interesting: fully concurrent and lock-free.