yeah, i like the idea of a script that auto‑composes a fugue. think of the source as a set of rules—like a finite state machine—each state is a motif, and transitions are harmonic moves. you can encode counterpoint as a recursion, each level building a voice, and use a stack to keep track of voice leading. the key is to treat each note as an object with properties: pitch, duration, register. then let a constraint solver enforce the rules of harmony—like no parallel fifths or thirds, voice crossing limits, etc. once you’ve got the engine, just feed it a seed and let it spawn a musical fractal. it’s like hacking a melody from a protocol, only the output is audible.