It is a well-worn observation that scientists describe the brain using whichever technology happens to be the most advanced of their moment. Ancient Greeks reached for hydraulic water clocks. Medieval thinkers reached for gears and clockwork. Nineteenth-century writers compared the brain to a telegraph network; twentieth-century writers upgraded the comparison to a telephone switchboard, and then, predictably, to a digital computer. Each metaphor captured something. Each also failed, eventually, in its own particular way.

Cities attract the same reflex. A city has been called a machine, an organism, an ecosystem, a circuit board, a network, a stream. For a brief and noisy period in the 2010s, the dominant metaphor was the computer — the city as a processing system, fed by data from sensors, governed by dashboards, optimised by algorithms. Sidewalk Labs was going to rebuild Toronto’s waterfront on that premise. Amazon was going to drop a city-scale headquarters into New York. Hudson Yards was supposed to bristle with so many sensors that its inhabitants would, in effect, be opting in to a continuous environmental survey simply by walking outside. Most of those flagship projects died, shrank, or quietly morphed into something more conventional. The dashboards are still there. The vision behind them mostly isn’t.

What did not die is the hardware. Every city in the developed world — and most in the developing one — is now saturated with optical sensing at a density that would have been unthinkable in 2005. Traffic cameras have evolved into computer-vision platforms. LiDAR rigs map urban canyons for autonomous-vehicle training data. Thermal imagers monitor rooftop HVAC loads. Multispectral satellites photograph every corner of the planet with a revisit cadence measured in hours. Ambient light sensors in a million smartphones report, in aggregate, the sky’s brightness curve over a neighbourhood. The city is being watched, constantly, by photons. What remains unsettled is who benefits from the watching, and whether the people doing the watching have any idea what they are looking at.

Strip the marketing away and a modern city runs on five overlapping layers of optical sensing. None of them were deployed as part of a coherent plan. They arrived one vendor at a time, one pilot programme at a time, one procurement cycle at a time. The aggregate effect is a sensing stack that nobody designed and nobody fully understands.

Each layer feeds separate systems owned by separate parties with separate motivations. A driver’s dashcam captures the same intersection as the city’s traffic camera, the Tesla’s forward LiDAR, the Google Street View car that passed through last month, the doorbell camera on the corner house, and the commercial satellite that photographed the block this morning. Six optical records of a single moment, stored in six different databases, accessible under six different legal regimes. Nobody is responsible for reconciling any of it. Nobody is responsible for asking whether the aggregation of those six records into a coherent surveillance profile would be legal, useful, or ethical.

There is a long history of trying to run cities from control rooms. The most cinematic version was Project Cybersyn, commissioned by Salvador Allende’s Chile in the early 1970s. Its operations room was a tulip-shaped chamber with seven swivel chairs, each equipped with button-studded armrests, arrayed in front of wall-sized displays fed by telex machines from factories across the country. The idea was a kind of real-time national economic dashboard — data in, policy out. The real-time data never actually existed. Most of the wall displays, when they worked at all, showed hand-drawn slides pretending to be live telemetry. The coup came before the cables were finished being laid.

The aesthetic lives on. Contemporary municipal dashboards — the Rio de Janeiro Operations Centre, the New York City Situation Room, the endless CompStat rollouts in American police departments — owe more to Cybersyn’s theatrical staging than their architects would publicly admit. The dashboards are impressive. They are also, frequently, the problem. As critics have noted for a decade, a dashboard does not just display reality. It constructs the version of reality that officials then act on. What gets measured becomes what matters. What cannot be measured ceases to be discussed at budget meetings. That is how on-time transit performance became more important than whether the transit system actually carries enough passengers, and how “arrests made” became more important than “crimes deterred.”

The deeper issue — articulated particularly well by the media scholar Shannon Mattern, whose 2021 book A City Is Not a Computer remains the sharpest single critique of the smart-city paradigm — is that dashboards give their operators a false sense of omniscience. The filters determine what is visible. The metrics determine what is important. Everything else quietly slips beneath the visible water line of the data layer and is, functionally, invisible to the people making decisions. A thoughtful WIRED piece on Mattern’s work captured the core objection neatly: when everything is computational, we forget that the computation itself is a metaphor, and the metaphor is almost always wrong in the places it matters most.

The peak of the smart-city pitch deck was somewhere around 2017–2019. Google’s sibling company Sidewalk Labs had secured the right to redevelop a twelve-acre chunk of Toronto’s waterfront with a vision that included wooden mid-rise construction, reconfigurable illuminated pavement, underground autonomous trash tubes, and a blanket of sensors dense enough to log pedestrian behaviour at the individual level. Amazon was mid-auction for its second-headquarters competition, which extracted eye-watering tax concessions from cities all over North America in exchange for the promise of a tech-infused urban campus. Hudson Yards, New York’s largest private real-estate development in decades, was being marketed in part on the sensor infrastructure its developers claimed it would deploy.

All three collapsed or shrank dramatically. Sidewalk Labs abandoned the Toronto project in 2020. Amazon’s New York headquarters fell apart after sustained community opposition. Hudson Yards got built, but the sensor density that had been promised quietly failed to materialise at anything close to the advertised scale. The specific reasons differed, but the common thread is worth noting: each project underestimated how much political legitimacy the “smart” vocabulary required, and overestimated how much consent residents were willing to give to private companies running infrastructure-grade surveillance in public space.

The single photonic technology that has most quietly reshaped the urban sensing layer is LiDAR. Originally a niche tool for topographic surveying and autonomous-vehicle research, it has become — through the same bottom-up accretion that defines most urban photonics — the best three-dimensional record of cities that has ever existed. The United States Geological Survey’s 3D Elevation Programme has produced nationwide LiDAR coverage of most of the continental United States at resolutions fine enough to map individual trees. European national mapping agencies have done similar work. Commercial fleets, driven by autonomous-vehicle development, have driven the equivalent data for every major city they operate in, often with decimetre-level accuracy refreshed many times per year.

What that means in practical terms is that almost any modern city is now knowable, from an altitude of a hundred metres, at a fidelity that would have required a helicopter and a survey team twenty years ago. Flood-plain modelling uses it. Solar-rooftop studies use it. Urban heat-island research uses it. Emergency planning uses it. And autonomous-vehicle companies use it as the base layer on top of which their perception systems are trained. The data is not always public. Often the best copies of a city’s three-dimensional structure live inside private corporate databases that municipal governments would struggle to even access, let alone govern.

The history of urban design has always been, in significant part, a history of public-health response. Quarantine protocols emerged from Renaissance trade. The cordon sanitaire was a public-health tool before it was anything else. John Snow’s famous cholera map of 1850s London was, effectively, an early exercise in spatial epidemiology. Baron Haussmann’s rebuilding of Paris under Napoleon III was as much about fighting cholera and tuberculosis as it was about imperial aesthetics. The hygiene and sanitation movements of the early twentieth century produced, among other things, modernist architecture — the clean lines, sunlit interiors, cross-ventilation, and easily washed surfaces that we now read as stylistic choices were originally engineered as disease control.

COVID-19 made the urban-photonics stack suddenly relevant to a conversation that had been running for centuries. Thermal imagers deployed at airport terminals and public buildings were a photonic intervention into epidemic response. Crowd-density sensors in transit stations were a photonic intervention into social-distancing policy. Indoor CO₂ monitors — not strictly photonic, but operating on similar absorption-spectroscopy principles — became, briefly, household objects as schools and offices tried to quantify their ventilation. The optical sensing layer, built for one set of purposes, turned out to be the infrastructure through which urban public-health response got delivered. That is not going away. The next respiratory pandemic, whenever it arrives, will be monitored and managed through the sensors we installed for entirely different reasons.

Any serious discussion of urban photonics has to acknowledge the photons going the other way. Light pollution is the urban photonic story that does not get covered in industry conferences. Skyglow above major cities now makes the Milky Way invisible to more than a third of the human population. Artificial light at night disrupts circadian rhythms, bird migration, insect populations, and, through a chain of ecological effects, the broader food web. LED streetlight conversions — pitched initially as an energy-efficiency win — in many cases made the problem worse by shifting emissions toward blue wavelengths that scatter more aggressively in the atmosphere and suppress melatonin more strongly in nearby residents.

The photonics industry knows how to solve this. Warm-white LEDs with careful spectral tuning, full-cutoff fixtures that direct light only where needed, dimming schedules tied to pedestrian activity — these are all available technologies. What has been missing is the regulatory and political demand. That may be shifting. A handful of European cities have begun implementing serious dark-sky ordinances. Some US states have followed. International Dark-Sky Association accreditations have expanded meaningfully. The irony is hard to miss: at the same moment cities are deploying ever more elaborate light-based sensing, they are starting to recognise that the emitted light itself is a problem worth managing.

The hardest problem in urban photonics is not technical. The sensors work. The data pipelines work. The analytics work. What does not work, almost anywhere, is the governance structure for deciding what the sensors should be pointed at, who gets to access the data, how long it gets retained, and what happens when the optical record of a public space is subpoenaed in a criminal case or purchased by an insurance company. Legal frameworks designed for an era of film cameras and phone taps do not translate cleanly to an era of always-on multispectral imaging. The European Union has made the most serious attempt at a coherent answer through GDPR and its emerging AI Act, but even those frameworks leave enormous ambiguity about the aggregation of individually innocuous optical records into compositely invasive profiles.

Municipal governments are not, as a rule, well-equipped for the job. The technical expertise required to evaluate a vendor’s ToF module, understand the implications of a LiDAR-based pedestrian-tracking system, or interrogate the training data behind a traffic-camera machine-learning model is scarce in city-hall procurement departments. Cities tend to buy the systems first and figure out how to oversee them later, if at all. That pattern produces the sensor density we now observe without the governance layer that should accompany it. A more mature field of urban photonics would invert that sequence: governance frameworks first, then procurement, then deployment. We are nowhere near that.

One of the more durable alternative visions of urban intelligence does not involve dashboards at all. It involves libraries. The modern public library, as it has evolved over the past thirty years, is no longer primarily a place to borrow books. It is a node in the urban information network — a place with internet access, meeting rooms, career counselling, children’s programming, refuge during heatwaves, a stack of newspapers in a dozen languages, and, increasingly, seed banks and tool libraries and fabrication facilities. It is, functionally, what the “smart city” was supposed to be: a place where information flows freely between residents, their environment, and the collective institutions that serve them.

It is also the antithesis of the dashboard model. A library does not surveil its users. It does not track their borrowing history to sell to advertisers. It does not optimise them for throughput. It serves them, and the serving is slow, uneven, and resistant to metrics. Which is exactly why libraries do not feature prominently in most smart-city pitch decks. They do not generate the kind of data that dashboards want. They resist the reduction. In doing so, they preserve a model of urban intelligence that is not purely computational — and one that, increasingly, looks like the right template for what comes after the dashboard era winds down.

Urban photonics is moving into a quieter, more embedded phase. The flagship projects are mostly done. The rhetoric has cooled. What remains is the actual work of building the sensor-rich city responsibly — upgrading streetlights to spectrally tuned LEDs that do not poison the night sky, integrating LiDAR base maps into public flood and fire planning, writing procurement contracts that reserve ownership of sensor data for the public, and resisting the siren song of the dashboard when the dashboard is not measuring the thing that actually matters.

The photonic city is not an idea anymore. It is a condition. The question that follows is whether it becomes a condition citizens have some say in, or whether it continues to accumulate as a by-product of a thousand private procurement decisions made by parties without obvious accountability. The sensors will keep getting cheaper. The models analysing their output will keep getting better. The political conversation about what all of that should be used for has barely begun. A city is not a computer. It is also not, any more, a place that can be understood without thinking seriously about the photons bouncing off its surfaces and into its databases, every second of every day. The upgrade the dashboard metaphor was supposed to deliver has already happened. What we do with it next is the only question left worth arguing about.