Why Cities Don't Die
Organisms grow to a characteristic size, maintain homeostasis, and die. Companies grow, plateau, and die on mathematically predictable schedules. Cities do neither. They grow. They sometimes shrink. They transform. But they do not die of natural causes. In West and Bettencourt's data, the death rate of cities from internal causes is effectively zero. Rome has been sacked, burned, occupied, and depopulated by plague — yet it persists, with a population today exceeding that of the Roman Empire's capital at its height. Detroit lost sixty percent of its population between 1950 and 2010, but it did not die; it shrank, painfully, then began to grow again differently. Cities are destroyed only by external catastrophes — Pompeii by Vesuvius, Hiroshima by nuclear weapons. Absent such interventions, cities persist indefinitely. There is no mathematical equivalent of the biological sigmoid curve for cities, no built-in deceleration leading to stagnation and death. The growth dynamics of cities are, in principle, open-ended — a property produced by their superlinear exponent and the open network topology that generates it.
In the AI Story
The distinction between city-immortality and organism-mortality is not a matter of degree. It is structural. When outputs scale superlinearly with size — when doubling the population more than doubles the economic output — the system generates increasing returns. Each additional unit of input produces more than one additional unit of output. The surplus feeds back into the system, attracting more people, generating more interaction, producing more surplus. The cycle is self-reinforcing. Unlike organisms, which reach stable size because their sublinear metabolism cannot support further growth, cities never reach equilibrium. They are always accelerating.
The mechanism, in West's analysis, is not technological or economic in the first instance. It is social. The superlinear exponent of approximately 1.15 emerges from the density of human social networks — the number and diversity of interactions per person per unit of time. A city of one million does not merely contain twice as many people as a city of 500,000. It contains, effectively, more than twice as many connections between people.
The practical significance of this finding for the AI transition is profound. If AI increases effective cognitive density without requiring physical proximity — if it allows a developer in Lagos to experience connection density exceeding San Francisco's — then the city-like properties of open-ended growth and persistence become potentially available to networks that were previously constrained by geography. Organizations that adopt city-like topologies may acquire city-like longevity.
But the framework carries an equally important warning. Cities do not die, but they can become uninhabitable. A city that grows without building infrastructure — physical, social, institutional — commensurate with its growth does not collapse suddenly like a company. It degrades gradually. The pathologies outpace the adaptive capacity of its institutions. The creative class that drove the superlinear returns is priced out. The social fabric that held the dense network together frays. Persistence is not flourishing.
The cities that thrive are not the ones that grow fastest. They are the ones that build institutions commensurate with their growth — transit systems, public health infrastructure, educational institutions, social safety nets, zoning laws, cultural norms — at a pace that keeps up with the superlinear acceleration of both their brilliance and their dysfunction.
Origin
The finding that cities show no intrinsic mortality emerged from Bettencourt and West's analysis of U.S. urban data in the mid-2000s. The specific observation — that cities persist across millennia while companies die within decades and organisms within years — was given mathematical precision in their 2010 and 2013 papers on urban scaling dynamics.
Key Ideas
No intrinsic mortality. Cities are destroyed only by external catastrophes; absent such interventions, they persist indefinitely.
Superlinear growth dynamics. The self-reinforcing nature of superlinear returns produces open-ended expansion rather than the sigmoid deceleration of sublinear systems.
Density of connection is the mechanism. The superlinear exponent emerges from the density of human social interaction — the collisions between minds that produce innovation.
Persistence is not flourishing. Cities can persist while becoming uninhabitable; the pathologies scale at the same exponent as the innovations.
Institutional adaptation is required. Thriving cities build infrastructure — physical and social — at a pace commensurate with their superlinear growth.
Appears in the Orange Pill Cycle
Further reading
- Luis Bettencourt et al., Growth, innovation, scaling, and the pace of life in cities (PNAS, 2007)
- Luis Bettencourt, The origins of scaling in cities (Science, 2013)
- Geoffrey West, Scale (2017), chapters 7–8
- Jane Jacobs, The Death and Life of Great American Cities (1961)