A Yamaha Disklavier can play notes faster than any human hand. The question is what to compose for it. Three traditions have developed answers independently: Nancarrow's tempo canons (interlocking pulse trains at irrational speed ratios), Xenakis's stochastic distributions (statistical textures), and L-system grammars (recursive self-similar structures). A new system called Amanous unifies them through distribution-switching, where grammar symbols select entire distributional regimes rather than modulating parameters within a fixed family.
The structural discovery is a saturation transition. At 24-30 notes per second, single-domain melodic metrics — pitch contour, interval distribution, rhythmic regularity — lose discriminative power. Below this threshold, you can tell sections apart by analyzing any one dimension. Above it, the texture is so dense that individual dimensions become statistical noise. Only cross-domain coupling — relationships between pitch, timing, and velocity simultaneously — can distinguish one passage from another.
This is a phase transition in analytical capacity, not in the music itself. The Disklavier doesn't change its behavior at 30 notes per second. But the tools designed to characterize music do. The analytical framework, built for human-playable densities, breaks at the boundary of human performance. Superhuman texture requires superhuman analysis: metrics that couple domains the way the texture couples notes. The instrument surpassed the player decades ago. Now the analysis has to catch up.