Falandays and Smaldino explain how shared cultural forms can stabilise without presupposing a population that already thinks alike. Their agent-based model begins with cognitively plastic, initially uncoordinated learners who provide training input to one another; under specific conditions, their category structures spontaneously align and produce stable cultural attractors. The decisive theoretical move is to make alignment emergent rather than given. Noise, learning periods, population scale and network topology become active variables in the formation of collective cognitive order, with the counterintuitive result that noise can sometimes increase stability. Methodologically, the work extends unsupervised category learning into a multi-agent social system and makes cultural convergence experimentally tractable. Its wider importance lies in redefining cultural persistence as a dynamic equilibrium generated by repeated perception, reproduction and mutual adjustment. The paper bridges cognitive science, cultural evolution and network theory by showing how categories can acquire population-level durability even when individual representations remain contingent and continually renewed.