I do have pipe dreams of making Alda way, way more extensive than it is right now. MIDI ended up being the easiest thing to implement as a sort of proof of concept, but I am completely open to (in fact, am really hoping for) a wide array of music generators at the disposal of Alda's syntax. The next major milestone will be waveform synthesis, probably via a Java library like Minim (my original idea was to control Supercollider via Overtone, via Alda, but it seems there are too many setup issues involved, when compared to a self-contained Java library that I can use from Clojure).
Being able to generate code in other audio programming languages is also a thought I've had -- if not for Alda, then for a spinoff project. The primary thing that inspired me to create Alda was that I love the unbounded potential behind these more robust languages like ChucK and Csound, but the syntax of these languages could be greatly simplified and adapted for use by composers with little-to-no programming experience. I guess my point is that Alda prioritizes aesthetics and simplicity over being a Turing-complete language that lets you do everything you want.