You can absolutely operate (and even operate well) without understanding the formalisms involved in theory, but understanding them opens up a new level of fluen

You can absolutely operate (and even operate well) without understanding the formalisms involved in theory, but understanding them opens up a new level of fluency. I don't need to know how to name common progressions or modes in order to write a song, but if I have my theory down well enough, it's easier to transpose to a different key, or have ideas for chord voicings or variations on the progression that'll enrich the piece (and as far as I can tell, the theory involved here is a form of mathematics). I don't need to understand the mathematics of waveforms in order to twist a knob and get a different sound and make an aesthetic judgment, but having a model of how waveforms works in my head to correlate with specific sounds helps. This is what math does. It helps take ad-hoc implicit models -- which people (and now to some extent ML algorithms) can often get by on pretty well -- and move them into explicit formalisms, often providing new power and insight in the process. The fact that people can build internal models that make them some degree of effective composers/performers implicitly isn't a particular rebuke of the idea that a significant portion of what we call music can be well understood as mathematics.