The Diffusion Layer.

Physical AI runs on silicon, and the world cannot design or make enough of it, because the binding constraint is people. The industry needs on the order of a million more skilled workers by 2030, and the hardest roles are exactly the ones that cannot be taught quickly. This track studies that as a diffusion problem. The knowledge to design chips, drive the tools, and work a fab spreads through a workforce along an adoption curve, and the rate of that curve is a variable we can move. Some of the skill is generic and travels fast; the highest-value skill is esoteric and tacit: it lives in practice and moves slowly, which is why a leading-edge fab is so hard to copy. The Diffusion Layer maps where that line falls, who pays to cross it, whether skill learned on one process node transfers to the next, and whether an AI copilot can carry an expert's tacit knowledge to a novice, compressing the curve that sets how fast the workforce — and the silicon — can scale. It is a new competency of the Human Layer, turned toward the makers.