There was a time when cities were blank canvases for urban planners to unleash their wildest imaginations. Brasilia, designed from scratch in the cerrado by Oscar Niemeyer and Lucio Costa in 1960, was shaped like an aeroplane—a modernist statement that prioritised a monumental aesthetic over everyday life. Barcelona, with Cerdà’s Eixample in 1859, imposed an orthogonal grid of octagonal blocks to tackle overcrowding and disease, but also to enforce social control. The Ciudad Lineal, envisioned by Arturo Soria in 1882 around Madrid, sought to extend the city along a tramline, blending countryside and urban life into a five-metre-wide strip. Meanwhile, Le Corbusier dreamed up his 1935 Ville Radieuse—skyscrapers set within a park, sacrificing the traditional streetscape on the altar of rationalist efficiency.

Whether weaving stylistic flourishes or pursuing strict rationalism, these planners relied solely on instinct and emotion. However, the 1990s and 2000s ushered in a period of rapid urban growth dominated by a far less planned development model.

And what about today? In our time, we have the opportunity to reinvent the urban model through technology, with environmental and social sustainability as our primary objectives.

Unsurprisingly, digital solutions of every stripe have long been integrated into urban environments. It is virtually impossible for any citizen to have escaped public officials talking about "smart cities"—a catch-all concept that covers a vast range of different initiatives.

Local authorities have consequently shifted from a piecemeal approach to technology—defined by connected streetlights, sensor-equipped bins, and isolated pilot projects in mobility or public space maintenance—towards the deployment of urban digital twins. These are designed to operate as dynamic, cyber-physical replicas of the cities themselves.

These urban twins are extraordinarily useful because they continuously ingest massive streams of data from land registries, weather stations, transport networks, and utility telemetry to simulate future scenarios with pinpoint accuracy.

Thanks to them, we can anticipate what will happen to traffic during major crowd-pulling events (such as the next papal visit to Madrid or a World Cup), which areas of a city will bear the brunt of an historic snowfall like Filomena, or where anti-social behaviour might flare up under specific triggers.

Yet, this infrastructure remains a simulation of an already established backdrop—a sketch that can only be partially modified or influenced. If we take it a step further, however, we can use this exact knowledge base to integrate artificial intelligence and reach a frontier that Le Corbusier or Niemeyer never could: designing the perfect city.

Can artificial intelligence serve as a sort of invisible architect for our metropolitan areas? While it may sound like an overly bold proposition, AI is undoubtedly becoming an indispensable tool for balancing the economic, energy, and political priorities embedded within complex mathematical models. Ultimately, these models give birth to urban planning blueprints or the design of entirely new neighbourhoods in our cities.

This qualitative leap transforms the algorithm from a mere three-dimensional visualization tool into a lever of political and economic power.

In this article:

• From the smart city to the algorithm-designed city
• Singapore and Helsinki as urban laboratories
• Algorithms and social impact
• Blending algorithms and cities for the greater good

The first wave of smart cities between 2009 and 2019 built its strategy on the techno-centric premise that a massive rollout of sensory hardware was enough to trigger efficient territorial management. To nobody’s surprise, this approach revealed severe structural limitations. It turned out to be closer to corporate marketing than profound change, proving entirely unable to shift the deep dynamics of local governance.

Ignacio Alcalde, an urban planner with three decades of hands-on experience and recipient of the European Urban Plan Award, shares this view: “The smart city model has failed because it has been too much ‘smart’ and not enough ‘city’. If your city’s values aren’t clear and you haven’t defined a future vision, technology ends up serving the interests of tech firms rather than those of the community.”

A staggering amount of investment surrounds this conceptual umbrella. According to Market Research Future, the smart city sector is projected to reach $73.5 billion in revenue by 2028. In turn, this expanding ecosystem requires storing and processing petabytes of real-time information, shifting a de facto share of power to major cloud infrastructure providers.

As a case in point, the metropolis of Singapore alone poured more than $100 million into Virtual Singapore, turning the project into the global benchmark.

In this regard, Alcalde argues that urban intelligence must be constructed across three consecutive, non-negotiable layers: “Values come first. Upon those values, we build the vision, and only from that vision do we layer on the technology. This is completely opposite to the traditional smart city approach, which starts with the tech—and in my view, that is its fatal flaw.”

It is a question of maturity that should lead us to use technology as a means rather than an end when designing and understanding human habitats. In a systematic review published in Sustainable Cities and Society, researchers Charlotte Weil, Simon Elias Bibri, Régis Longchamp, François Golay, and Alexandre Alahi analyse how urban computing demands an architecture capable of processing the continuous interplay between the built environment and social dynamics.

Only when a predictive model reaches this level of maturity does it cease to be a laboratory experiment and become a genuinely effective tool for resource allocation. At that point, it can direct private investment flows, justify traffic rerouting that disrupts local commerce, or prioritise interventions in specific districts under strict criteria of mathematical efficiency.

As noted, Singapore is the most advanced exponent of semantic simulation applied to extreme density. Through Virtual Singapore, the city-state maintains a geospatial database under the international CityGML standard at a LoD3 level of detail. Thanks to this vast repository of data, its system can semantically recognise the material properties of every façade, window, or road infrastructure component from LiDAR point clouds.

Operating on top of this infrastructure is Cooling Singapore 2.0, a digital climate twin designed to combat the urban heat island effect. Led by the Singapore-ETH Centre with a team of more than twenty researchers, these experts have integrated microscale meteorological models with anthropogenic heat emission data. This enables the deployment of coordinated packages combining urban morphology, district cooling, and adaptive greenery.

This approach yields a temperature reduction of up to seven degrees in critical areas, cutting cooling energy demand by 8% and decreasing residual waste heat emitted into the atmosphere by 22%.

The symbiosis between cities and artificial intelligence seems written in the stars. It is telling that just eight years after the concept of AI was coined at the Dartmouth Conference in 1956, the celebrated philosopher Henri Lefebvre formulated his own landmark concept: the “right to the city.”

Laudably, Lefebvre argued that the right to the metropolis is not merely about accessing its services, but about holding the power to actively participate in the production of urban space. This was an idea that Manuel Castells would later expand upon in The Urban Question, demonstrating how the collective consumption of services forms the invisible infrastructure that shapes structural inequality.

Today, the artificial intelligence ecosystem is the new great medium of collective consumption. Yet we are delegating design and decision-making to these probabilistic black boxes, stripping citizens—not to mention visionary urban planners—of their agency to shape their own habitats.

The concept of the Black Box City, championed by Sarah Williams of the MIT Senseable City Lab, describes urban centres where decisions are made within opaque systems. Consider this scenario: if a model solely prioritises traffic flow or the maximisation of land values, it will inherently blind itself to variables like social cohesion or pedestrian accessibility.

This humanist perspective collides head-on with mathematical simulation when reviewing the principles of Jane Jacobs. In The Death and Life of Great American Cities (1961), Jane Jacobs fought against top-down boardroom planning, arguing that a neighbourhood’s vitality depends on “eyes on the street” and the complex, organic choreography of its pavements.

Applied to our current reality, the current design of digital twins is simply incapable of parametrising this serendipity. A predictive model can calculate densities and flows, but it cannot encode creative friction or the community’s instinct for mutual care. By attempting to optimise the landscape by purging its inefficiencies, algorithmic urbanism risks repeating the same sterile, hyper-sanitized errors that Jane Jacobs fought against: treating the city as a simple engineering equation.

This clinical detachment inherent in algorithm-driven urban design even clashes with the popular European framework of the “15-minute city.” Coined by Carlos Moreno, this model proposes that all essential daily services should be accessible within a short walk or bike ride.

Digital twins can easily optimise this theoretical distribution on paper. However, a poorly calibrated algorithm could easily design a 15-minute metropolis that is perfect on the stopwatch but utterly sterile in terms of its social fabric—accelerating gentrification processes by prioritising spatial efficiency over community equity.

Furthermore, navigation apps that reroute thousands of heavy vehicles down quiet school runs, or ride-hailing platforms that spike prices via dynamic pricing during emergencies, expose the clear dangers of outsourcing territorial management to code that remains blind to the public interest.

Despite these debates, the dual city—a physical city and a digital city coexisting and influencing one another—is the reality we face. Digital twins can either widen or bridge the gap between the two, depending entirely on who controls them and what they are used for.

The most ambitious proposal for maximizing the benefits—and minimizing the harms—of this combination comes from Rashid Mushkani. In his August 2025 paper, Urban AI Governance Must Embed Legal Reasonableness for Democratic and Sustainable Cities, he introduces the Urban Reasonableness Layer (URL).

The concept of the “reasonable citizen” originates in 19th-century English common law, where it served as a benchmark to evaluate the conduct of an average person. Mushkani proposes extracting this concept from the courtroom and embedding it directly into the core code of urban AI.

How exactly would this work? What Mushkani proposes is the creation of a sort of automatic circuit breaker that audits software decisions before they are executed. Imagine a digital twin detects a pollution spike and automatically decides to reroute all bus traffic through a low-income peripheral area, drastically increasing local accident rates. A machine lacks the contextual awareness to understand that this outcome is unjust.

This is precisely where the Reasonableness Layer comes into play. It would pause the process, evaluating the mathematical decision against a human standard previously encoded through public civic debate. If the measure imposes a disproportionate cost that a reasonable person would reject, the system blocks the optimisation.

In other words, it forces artificial intelligence to pass a community empathy test. This aligns perfectly with the point Ignacio Alcalde raised at the outset: “If we first know what our values are, and have built a vision matching those values, we can then approach the tech industry and ask them exactly how they can help us achieve that vision.”