I have a Pie in the Sky wish too, and that's for a synthesizer built into the SDK. I have no illusions that that's an easy request, and in fact, I'm starting to work on resolving the issue to help simplify Sonic Pi and other live-code concepts in Ruby (and Crystal). I'll talk to y'all about it as my progress matures.
But I'd give anything for a GDK with a synthesis engine baked in so that I can create a generative soundtrack based on user input.
Believe it or not, I do want to add the capability to create "chip tunes" on the fly :-)
I'm working on my Ruby "201" right now specifically to fork Sonic Pi into a (literally) modular system with a (hypothetical) JIT/AOT compiled synthesis description language written in Crystal (an up and coming Ruby-like language for LLVM). I love Sam Aaron to death as he's essentially my first Ruby mentor, but I can tell his goals for the project as an educational tool are very different from my own as a glitch/breakcore/chiptune/electro-jazz performer.
It's an ambitious project but I hope to have the pieces together for Ruby to sequence a compiled synth with minimal outside dependencies within a year in order to build both the DSL for the synth and a library that will send build directives to Crystal/LLVM from Ruby and then magically connect to the resulting optimized binary in the background. I have a feeling I'll have to do the former myself, but I think when I can show a proof of concept about a live coding a compiled language, live coders and web developers who hate building between updates will see the value. I noted that Matz unveiled an MJIT in Ruby 2.6 and what he's doing here is basically the same idea with a pure Ruby twist and automatic decisions about what to compile.
Long story short... OMG that would be awesome and if there's any way I can help to enable that within the Ruby ecosystem, I certainly will. By the same token, if you have a lib out there in a Ruby-like syntax that generates a waveform with only minimal c dependencies, I'd be grateful to know about it as that would save me a lot of fundamental work and let me focus on the implementation language.