I generally agree with you that compiling to machine code is not a magic wand, code that does more is always slower than code that does less in a fair fight. Bu

I generally agree with you that compiling to machine code is not a magic wand, code that does more is always slower than code that does less in a fair fight. But interpretation adds additional overhead to make slow code even slower, the more complex\unoptimized the interpreter the more it adds overhead. It's good for a language to have a baseline expectations of "Don't make things worse", shit code will always be slow, but at least it's not made slower by things outside of the developer's responsibilities.