Which is good. At Reddit they would have tried to rewrite everything on reasonML and then tried to prove at the end that it is now faster