I mean, I know it goes without saying, but statically typed languages are beating the pants off of dynamic languages. In JSON serialization, you have to go quit

I mean, I know it goes without saying, but statically typed languages are beating the pants off of dynamic languages. In JSON serialization, you have to go quite far down to get the first dynamic language one, japronto, and it's not even completely written in the language. Ha. All the others have pure dyn languages down further than that. Now, obviously micro-benchmarks are nothing to rely on and most of us have other bottlenecks and these minor differences are negligible in comparison. But why not be faster if you can with little ergonomic cost these days? Put a different way, and probably the language-decision question of the current times: Which is improving faster, the ergonomics of statically typed languages or the performance of dynamically typed languages? As langs like Crystal and optionally-typed Dart feel more dynamic and are still quick to get up to speed on, why must we persist with dynamic languages? Of course I'm biased, but it sure annoys me that the ML world stays in Python. /rant