This chapter debates on the macro-micro dichotomy in search of appropriate scales to build digital twins of cities. It argues that under nonlinearity the macro properties of a system cannot be directly read off from the microstates of the agents at any point in time, and macro emergence may not be directly derived from the properties of the parts and their interactions at the micro level. The entire system evolves along the dynamic interactions of the micro and macro over time, with the micro providing what is possible from bottom up and the macro conditioning the micro-level parts and their interactions from top down, attempting to strike a mutually beneficial coordination of both levels. The macro-micro division of analysis also excludes the meso domain which may be the center of change in the adaptive complex urban systems. By equipping digital twins of cities with the macro-, meso-, and microscopic perspectives, the simulation, prediction, and control of the agent-based urban systems may be made more holistic, flexible, practical, and interpretable. To carry the scale argument a step further, we might need a multilevel construct to comprehensively understand and model urban dynamics and complexity via digital twins of cities.