And then their O(10^7) LoC increment the exponent, as Go has fuck all higher level abstractions.

    > Sounds like they should use Go.

You're being too pithy and probably a bit snarky, but I think there's truth in this. I think Marius underemphasizes the problems of incidental complexity that plague Scala, and overemphasizes the importance of futures as a mechanism of composition. Specifically, I think all of the advantages he attributes to futures and async methods can also be equally well achieved with coherent synchronous code, which is the domain of Go. Generalizing, I think the (good and interesting!) things he outlines in this presentation can not only be achieved in Go, but in a way that makes them even easier to reason about, manipulate, extend — all of the things he wants to optimize for.

Maybe Scala was a good, or even the best, choice of implementation language for concurrent application services six or eight years ago. I don't believe it's a good, and certainly not the best, choice today.