I wrote an essay exploring why music exposes the biggest architectural limitations in current multimodal AI systems. The short version: AI models today flatten time (Transformers) or flatten space (Diffusion models). But music requires multi-scale temp…