> better type system
> better type inference (bidirectional)
how do you quantify "sane" and "better" here ?