I disagreed strongly with this post and was about to post a rebuttal, but then I realized there actually is a situation where untyped languages enjoy a concrete, practical advantage because of the lack of compilation dependencies. That situation is when you are writing a module (something that would be a compilation unit in a typed language) that calls missing or nonexistent library interfaces, and you want to test parts of the module that do not depend on the missing APIs. In a typed language, you can't do it. You would have to satisfy the dependencies first so your code would compile, or you would have to copy snippets of your code out of the compilation unit into a REPL or another file so they could be compiled separately. In an untyped language, you can experiment with parts of the module and even start writing tests without having to make sure all of the module's dependencies are satisfied.

Aside from that one advantage, though, it's the same with types or without them. For a module as a whole, the compilation dependencies that exist in a typed language are the same as the runtime dependencies that exist in an untyped language. Whether you're using a typed or untyped language, you can write whatever code you want, but you can't do anything with it until you satisfy its dependencies. With that one exception, of course: in an untyped language, you can run code that doesn't depend on the missing APIs but happens to be in the same module as code that does depend on them.

Overall, I don't think that amounts to much, especially compared to a typed language like Scala where you can copy snippets into a REPL. Or even better, Ensime is an Emacs mode that is supposed to let you work with Scala the way Slime lets you work with Common Lisp. (I haven't tried it myself yet, but it exists, and people are using it for real work.)