Cities will become increasingly instrumented and interconnected by the middle of the century, generating huge amounts of user-generated data. Mobility, openness, and connection will be more important than physical presence. Future city services will need to address massive complexity and interdependency. Challenges include assimilating big data at internet scale, modeling human demand, and accounting for uncertainty. ReasonableCity is an approach using learning systems to help diagnose issues in cities from diverse real-time data. Understanding online and physical communities can help plan services and characterize spaces. Robust water network management techniques address uncertainty in demand and infrastructure. The goal is to make city data and services as accessible as email.