A Computer Might Be a City

For most of the twentieth century, the computer had a relatively stable cultural role. It was a machine for calculation, storage, optimization, classification, and command execution. It was powerful, but also conceptually contained. It processed inputs. It produced outputs. Even when it behaved in ways that seemed complex, its complexity was usually understood as the expression of formal rules, however difficult those rules were for humans to follow.

That computer mattered enormously to urban planning. It helped planners model transportation systems, allocate infrastructure, evaluate land use, run demographic projections, and visualize possible futures. It also supported a persistent temptation: that the city itself might be made sufficiently legible to become computable. If enough data could be gathered, enough variables specified, and enough relationships modeled, perhaps the city could be optimized like other technical systems.

Urban theory has spent decades warning against that temptation. Christopher Alexander's "A City Is Not a Tree" criticized the reduction of urban complexity into overly neat hierarchical structures. Shannon Mattern's A City Is Not a Computer extended a related critique into the age of dashboards, platforms, sensors, and smart-city rhetoric. Their point was not that computation is useless but that cities should not be reduced to machines for processing information. Cities are dense, situated, messy, political, embodied, and contested arrangements of people, institutions, capital, regulation, and conflict.

Those warnings remain essential. But the object they were aimed at is changing.

The computer implied by this critique was, for the most part, a deterministic and instrumental machine. It was a machine of abstraction, simplification, command, and control. Even when it contained simulated entities that resembled agents, those agents were usually confined to narrow rules, limited environments, and explicit objectives. The computer could model a city, but it was not itself urban; it could represent interaction, but it did not meaningfully participate in social life; it could simulate emergence, but it did not usually generate open-ended, long-horizon behavior.

As of late 2025, it is harder to describe computers in those terms.

Large multimodal models, tool-using agents, autonomous AI systems, generative simulations, and networked fleets of computational actors no longer behave only as passive instruments. They infer, approximate, communicate, retrieve, remember, summarize, generate, navigate, plan, coordinate, and adapt. They operate across text, image, code, maps, video, sensor feeds, application interfaces, and - if given access - even physical environments. They can pursue long sequences of tasks without continuous human instruction. They can call tools, delegate subtasks, observe intermediate results, revise plans, and interact with other human and nonhuman actors.

This does not mean that computers are alive, conscious, or equivalent to people. It does not mean they possess political rights, moral agency, or civic membership.

But it does mean that the older category of "the computer" is becoming insufficient. A machine that converses, plans, senses, predicts, improvises, and coordinates across environments is not simply a faster calculator. It is a different kind of actor in the sociotechnical world.

The question, then, is not whether the city is a computer. The more difficult question is whether some computers are beginning to acquire properties that urbanists have historically associated with cities.

If It's Measurable, It Can Be Optimized

The old critique of computational urbanism was largely directed at reduction. It challenged the idea that the city could be compressed into a dashboard, a model, a control room, a set of indicators, or a tree-like structure of ordered relations. It resisted the managerial fantasy that urban life could be optimized from above if only the data were complete.

This critique was also epistemological. It asked what forms of knowledge become privileged when planning sees through computation. Quantifiable flows become easier to recognize than informal practices. Model-visible populations become easier to govern than communities whose knowledge is oral, local, fragmented, or resistant to measurement. Efficiency becomes easier to value than care. Prediction becomes easier to fund than participation. The technical frame does not merely describe the city; it reorganizes what counts as evidence and what becomes easy to know.

That argument remains powerful. Yet it depends, implicitly, on a particular image of the computer: a machine that is formal, instrumental, deterministic, and externally directed. It is a computer that acts on the city from a distance, converting urban life into data and returning recommendations, predictions, or controls. It is the computer as epistemic machine where the way of knowing is the way of governing.

Contemporary AI does not erase that critique. In many ways, it makes the critique more urgent. But it also complicates the object central to the critique. The AI-era computer is no longer only a representational device in the toolbox of urban governance. It is increasingly embedded inside institutions, workflows, communications, simulations, design processes, vehicles, buildings, public interfaces, and decision systems. It does not merely calculate after humans define the problem. It helps define the problem, frame the evidence, draft the alternatives, mediate the conversation, and enact the decision.

It shifts from an instrument to a participant.

A More Urban Machine

What would make a computer city-like? Cities are systems of systems in which many heterogeneous actors coordinate without being fully coordinated, adapt without being fully planned, and produce collective patterns that no single actor controls.

By that standard, recent AI systems invite a strange comparison. They involve distributed agency across models, tools, users, datasets, vendors, institutions, sensors, applications, and physical devices. They are only partially legible, even to their human operators. They are stochastic rather than strictly deterministic. They produce outcomes through interaction, not command alone. They can generate emergent patterns when many agents interact over time. They can develop conventions, amplify biases, stabilize norms, and coordinate behavior in ways that exceed any single rule or command.

This is not a metaphor to accept too quickly. Cities are made of bodies, histories, obligations, land, law, labor, violence, property, and politics. AI systems are made of software, hardware, data, energy, capital, protocols, interfaces, and institutional arrangements. The two are not the same, and the stakes are not comparable.

But the comparison can become useful. A city is not fully designed, yet it is not undesigned. It is planned in parts and self-organized in others. It is governed, but not controlled. It is legible in some ways and opaque in others. It is shaped by intentions, but also by unintended consequences. It contains local rationalities that do not sum neatly into a global rationality.

Some contemporary computational systems are beginning to exhibit a similar tension. They are engineered artifacts, but their behavior is no longer fully specified by explicit rules. They are trained, prompted, constrained, fine-tuned, monitored, and evaluated, yet their outputs often remain approximate, context-sensitive, and difficult to reproduce exactly. They are designed systems that behave in ways their designers cannot always predict, and cannot always control with the precision the old vocabulary assumes.

In older terminology, computational uncertainty was often treated as a problem to be reduced. In contemporary AI, probabilistic behavior is an inherent aspect of the mechanism. Approximation is what allows these systems to understand and generate language, images, code, scenarios, routes, explanations, and simulated behavior. The same property that makes them creative also makes them unstable. The same flexibility that makes them useful also creates bias, hallucination, surveillance risk, accountability gaps, and misplaced authority.

Planners already have a deep vocabulary for systems that are partially knowable, socially consequential, institutionally embedded, and impossible to control completely. That vocabulary may now extend to computation itself.

The Computer and the Built Environment

Planning has often evaluated computation through technical criteria: accuracy, efficiency, transparency, explainability, robustness, and bias mitigation. These remain necessary. But they are insufficient if AI systems become active participants in planning processes.

Transparency, for example, is not the same as accountability. Seeing more of a system does not automatically tell us how to govern it. A planning agency may publish prompts, model cards, audit logs, or evaluation metrics and still fail to answer the more important question: who is responsible for what this system does inside an institution?

Explainability has a similar limit. A model may explain why it produced a recommendation, but that explanation may be partial, post hoc, or rhetorically persuasive rather than institutionally meaningful. Can affected communities contest the system's role? Can they challenge the data? Can they understand where judgment entered and where it was delegated?

If advanced AI is treated only as software, these questions may appear secondary. If it is treated as an institutional actor, they become central.

This does not require personifying AI. We do not need to pretend that computers are citizens. Rather, we need to recognize that computational systems can occupy institutional positions. They can mediate between actors, organize attention, allocate administrative burden, generate official language, prefigure decisions, and change what professionals learn to notice.

This is where the built environment matters. Once AI systems summarize public testimony, generate scenarios, simulate users, route services, support permitting, or coordinate autonomous mobility, they are no longer simply tools used to think about cities. They become part of the machinery through which cities are imagined, negotiated, and governed.

The question, then, is not whether planners should use AI. They already do, and they will continue to do so. The question is sharper: what kind of computer is planning now inviting into the city?

If the answer is merely a faster instrument, the old rules may be enough: audit it, explain it, constrain it, supervise it. But if the answer is a semi-autonomous, partially legible, socially embedded system capable of coordination and emergence, then planning needs a different posture. It must ask not only what the computer can optimize, but what kinds of urban relations it helps produce.

The city was never a computer. That remains true. But the computer is no longer only the machine that urban theory learned to criticize.

So perhaps the question is not whether the city is a computer, but whether the computer is becoming urban enough that planning must learn to govern it as part of the city itself.