Some of these compilers are more toys than anything else: people always get a kick out of translating idioms between languages. That's fun, if a little boring after a while. But a lot of these tools are aiming at building abstractions for web programming, which is much more ambitious.
I still think these abstractions are an important goal, but the hard part isn't the compilers. It's understanding the models, i.e., the implicit specifications for web browsers. Apparently Ian Hickson has done a lot of great work compiling empirical data about the behavior of existing web browsers to infer the implicit specs for web standards. This kind of work is messy and tedious, but important.