Typo's a language that lets you express type-level computation using familiar abstractions (e.g., decimal numerals, booleans, conditionals, functions, ect.) rather than dealing directly with the native concepts of the type system. Once the type program's been compiled to a Haskell module, it becomes in essence a type-level library. You can include it in your regular Haskell programs and use the type class instances it exports in your own types to enforce whatever invariants you care about.

So I'll have to disagree with your characterization of it as a toy. It takes a technique that's been known in Haskell for a while and makes it more accessible and useable to the casual Haskell programmer. And while this blog post was a nice introductory treatment of that technique, that's all it is. The examples in the blog post are far from practical applications.