A Scalable Urban Design and the Single Building City

by Edward Swernofsky16 min read25th Jul 202015 comments


Urban Planning / DesignWorld Optimization

Crossposted from Medium. Initial description of proposed infrastructure.

Epistemic status: I've read a few books on the subject, spent a few months trying different things and working out the details, and am fairly confident this is at least worth a shot; definitely some kinks to work out; would appreciate writing and technical feedback.

-updated 07/26/2020 with more notes on sunlight, windows, and fires-

-transit network section updated 08/02/2020-

This article will be an exploration of some of the main issues people encounter with cities, the causes of those issues, and a comprehensive set of partial solutions and possible infrastructural innovations.

Problems with Existing Cities

One of the main issues people have with cities is the cost of housing. If you try to buy a house in a suburban or urban area, the suburban house will tend to be cheaper. In fact, if you just try to buy land, the suburban land will be cheaper. The land can be most of the property price difference for small buildings, though there may be other things at play like an artificially limited housing supply.

Why is space so expensive in a city? Well, it’s extremely valuable. Everyone wants to live there. Utilities such as water, electricity, trash, internet, and mail are much cheaper. There are more people around. You can buy whatever you want from tons of different stores. There’s a certain economy of scale about it. A lot of people and things in one place reduce a lot of the cost of moving around or bridging distances. There are more job opportunities. Everyone wants to put businesses there. It’s a lot easier to find people to work for you with more people nearby looking for a job.

Actually, why are buildings taller in a city? Higher is more expensive, right? Construction costs tend to depend most on whether the frame is wood, concrete, or steel, and whether there are stairs or elevators. Both of these mean higher numbered floors are more expensive. Despite this, if the land is valuable enough, it’s easier to make a 2nd or 20th floor than it is to buy more land and put a single story building up. In other words, buildings tend to be taller because the space they afford is worth the added expense.

Urban design, infrastructure, and regulations can also be uncoordinated, raising prices or imposing perverse restrictions. There are many different people with different aims: Max doesn’t want to live near a factory. Suzy likes driving fast. A borough, city, county, state, and country might each have a bunch of rules that try to satisfy conflicting interests, but sometimes those rules are really out of date, or the people that made (or keep, or enforce) the rules aren’t exactly looking out for everybody. New technologies that revolutionize different aspects of life are implemented haphazardly or fought over. People speculate on land and veto infrastructure.

The US, for example, is home to many building regulations and codes: Factory-built homes are required to have permanently attached chassis, have worse financing options and much stricter energy and fire codes, and are sometimes only permitted on certain monthly fee sites. The ADA requires an elevator for new buildings over 3 stories tall. Many counties in the US have lot size, parking, and apartment size minimums; height and square footage restrictions; and copious single use zoning.

Some places in the US let residents veto new development, and the residents take full advantage of this to raise their own property values. Land is valued for the kind of legal hoops it can jump through. There are huge farms just outside some cities that can’t be redeveloped into the housing that so many want. These restrictions can be fairly helpful in some cases, but the overall effect massively increases the cost of housing.

It doesn’t help that cities around the world prioritize cars in transport. Streets have grown wider and have free parking lanes. Gas isn’t taxed enough to factor in the harm to the environment. Half the land in many city centers is just road and parking. Have you ever had a hard time walking down a street because there were too many people walking? Doesn’t happen a lot, right? But that’s true for cars every day.

Transit throughput is a measure of people moved at some speed over time divided by the total width of space dedicated to the given travel mode. For example, a 4 foot sidewalk supports 2 people per second walking at 3 mph past a point. A foot of space can support 1.5 people-mph per second walking, 2 driving on streets, 3.5 highway driving, 4 cycling, 10 by highway moped, 50 by local rail, and 80 by commuter rail. Transit throughput is a good illustration of space usage for various long trip options. Higher throughput transit is made more useful in cities, where space is at a premium.

In the US people own cars, and cars have an average total cost of ownership of ~$10k per year. Cars heavily impact pedestrian safety, leading to a different civic environment. Some cities are so car oriented you can’t even walk to that many places from your home.

On the other hand, mass transit is so space efficient that it’s often cost effective to just bore a tunnel for it to use, despite this costing 5x as much as laying rail at-grade (at ground level), including stations. Cars are subsidized with free roads and parking and they’re still ~3x the total cost per passenger-mile of rail. Rail is unfortunately surprisingly expensive in the US vs other countries, as well as burdened by its own host of regulations and planning failures. Bus and train frequencies, despite being cheap to increase (or even just maintain off-hours), are too weak in many places to draw riders and empower the places they connect.

Cities around the world enjoy a wide range of densities. Here’s a density map of Paris:

The actual city boundaries aren’t so clear: The land around the city becomes valuable enough to develop as the city itself grows. Importantly, you can’t design for one density; density changes over time, and even a particular section will have spiking density around things like train stops.

In lower density areas, it can be pretty annoying to get around: Walkability is a measure of how friendly an area is to walking, and there are plenty of places that just don’t measure up. For example the LA metropolitan area, home to 13 million people at a population density of 2,700/mi²: Everyone owns a car, there’s 765 ft² of road per person, almost every public street is a stroad, and more land is set aside for parking than for housing. Higher density areas have more things reachable easily with your own two feet, and on top of that areas without cars or with tree shade are just nicer to view and walk around in.

Possible Solutions

OK, so cities and suburbs around the world have a bunch of issues impacting quality of life. What can we do about that?

A lot of regulations stifle the usable space market in perverse ways, leading to far higher prices. Others give undue priority and subsidy to cars. There’s not much you can do without letting many of these regulations go. It’d be nice to make an informed legal overhaul in a particular municipality, but existing systems aren’t usually amenable to such drastic change. There are still incremental gains to be had by working to reform regulations over time and implementing a few of the things to be presented here, but there’s another approach:

Just make a new city.

It actually happens fairly often already! It doesn’t always go that well, but people get into this sort of thing with a lot of different ideas and they aren’t always in line with what’s possible, both economically and socially. Still, Paul Romer, the Charter Cities Institute, Pronomos, and Apollo Projects outline some of the possible (mostly institutional) advantages and are hoping more of this kind of thing will happen soon.

We’ll focus on the infrastructure side of things in this article, looking at possible solutions to our whirlwind of issues plaguing cities. The best place to put a new city is usually right outside an old one, as close as you can get while still being cheap enough to buy all the land you need. You get to take advantage of a bunch of existing demand and services. Suburbs have the right idea, but the wrong vision. I’d like to make a note here that many of the following things will have to be modified, or in fact may not work altogether, for some places like very poor countries or extreme climates.

Land Ownership

First up is a Land Value Tax, which is a way to incentivize building higher and to disincentivize empty lots or land speculation. Tax the land based on regular appraisals of value either per plot or on a zone basis. There’s often an issue where appraisals only happen at sales and market value diverges wildly from this, leading to wildly under-appraised land or property. The current major source of municipal income in the US is a combination of property tax and sales tax. A land value tax is a nice alternative because it consolidates taxes onto something more straightforward which doesn’t decrease as a result of being taxed.

An alternative is a Harberger Tax, which works roughly like this: The owner of a plot states a price that they are willing to sell the plot and everything on it for (likely including a buffer for ownership frictional overhead). They’re then taxed a fraction of this price regularly. Anyone can buy the plot for the stated price at any time. Now market forces set appraisals and the feedback is faster, another fix to diverging market values. It disincentivizes density more, while being easier to maintain.

One of these taxes being implemented should lead to noticeably faster and more even intensification, or increasing investment, in valuable areas.

Transportation Infrastructure

Next up is cars. Large vehicles are necessary for some things: Specifically, construction and large deliveries. Trash can be handled with an automated vacuum collection (AVAC) system cheaper than with trucks. We’ll address mail and packages later. Let’s go with single one-way 10 ft lanes just to one side of streets that allow vehicles. Drop in an occasional few paid parking spots adjacent to the street for stops. Vehicles slow significantly in thin lanes with no shoulders and near pedestrians, contributing to safety. You might need to further limit speed somehow for very small vehicles.

Now we just need a congestion tax, implemented with automatic tolls around the city perimeter that charge either based on how long the vehicle is in the city or total street usage. Tune prices for low congestion, but high overall usage.

The main form of long-trip transportation here will be low speed highways at first. Each highway can be replaced with mass transit when there’s enough demand. Straighter roads affording higher speeds are more expensive, so with higher per-capita income highways can be higher speed and replaced with mass transit earlier.

(boundaries not to scale) Increasing street layout intensity: Purple dots are stations / highway access, dashed lines are transit, solid lines are streets; Left shows initial infrastructure spread and one-way direction arrows; Middle shows initial intensification; Right shows full saturation of streets.

An approximate street grid (following natural terrain limitations) maximizes walkability and service area by stations or highway entrances while minimizing street space and expense. If we add in streets just along the mass transit lines, we get another 20% reduction in average trip length to the train stop (due to the 10 min serviceable area becoming an octagon) with the tradeoff of less homogenous street traffic. (I’m sorry Jane Jacobs, for I have sinned) You can have the streets zig zag just enough to erase the endless street visual effect for 1% more street.

As train technology was introduced, rail has been implemented in major cities to great effect, but in a bit of a haphazard way: New lines are planned every so often and in large bursts over years, and construction of these lines is made more expensive by high land prices, crossing existing transport networks, and lack of coordination in switching transit modes.

In order to make scaling transport infrastructure cheaper, future transportation requirements should be mapped out in detail. Land for this future infrastructure can be instead leased for a period of time, or sold with clauses for construction standards permitting unimpeded future development. For example, a lot where a certain volume necessary for rail isn’t built through and can be reacquired by the city cheaply. This can still accommodate intensive development long before land is necessary and cheaper uses in years leading up to construction.

Here we’ll examine the case of transportation where the development in question is a suburb of an existing city. Long trips will be mostly be either commuting or local transportation. Demand will likely be significant to and from the parent city center as well as homogenous locally.

A grid-like network with one axis aligned with commuting will work well for this scenario. The transit network can extend past the initial new city land boundaries to near the overall metro area center (or until existing mass transit stations). At the lowest level of demand, highways within the development handle long trips locally and existing metro-area wide infrastructure handles commuting. When demand is high enough, buses and later rail connections to the parent city center can replace the highways.

It may be helpful to make a deal with the parent city to collectively purchase the land for a future rail connection early and share the cost of constructing it. This connection benefits the parent city as well, allowing stops along the line to enrich intermediate land.

If new city population and land ever contests the parent, or if this development is not a suburb at all but a standalone development, transit lines can be planned radially to better scale with demand.

For those who have only ever seen infrequent or weekday-only mass transit, be aware that it’s not particularly difficult or expensive to have high frequency trains or both local and commuter service at most stops. Maybe you wouldn’t consider taking the train if it came every hour, but what if it came every 3 minutes? With enough ridership, that frequency isn’t just feasible but actually quite manageable. Local service may only average 35 mph, but commuter service that makes fewer stops (one every 4 miles or more) may average 60+ mph even in dense areas.

There’s actually another approach to transit that eliminates walking to or waiting for trains altogether, or any kind of figuring out the (already quite simple) service network. It doesn’t make any stops, can be extremely fast, and achieves similar throughput to mass transit. In fact, it’s a kind of transit pretty similar to a fully autonomous car system if the cars were a lot smaller and had no headway. Personal Rapid Transit (PRT) unfortunately also is relatively young and hasn’t been realized at scale anywhere yet, and for that reason we’ll continue with mass transit in this article.

Raised Streets

Earlier we covered street land waste. Manhattan isn’t particularly bad, but 36% of the land is still streets. Those streets are the veins of the city, letting people move between buildings and delivering shipments to businesses.

But Manhattan is 23 mi² and the land is worth an estimated $1.7 trillion, so an average of $3k/ft². That’s a lot of valuable space; Why not just raise the street? Maybe it’s more expensive to raise car lanes due to vehicle weight (try putting a semi on an asphalt shingle roof), but it’s not hard to raise everything but those lanes. Roofs you can walk on are quite common and only a bit more expensive over time. If we’re raising the streets, why not parks as well? A green roof is just a sturdier roof with a drainage system and a bunch of dirt and plants, still only around $40/ft².

Several street heights and transitions with faint skyline outline; Numbers indicate height of street in floor #.

So raise the pedestrian street and sell the constructed space to lot owners on either side if the floor heights line up. If it’s wide enough, you can sell the space separately. This is where having standards or incentives for floor heights can help a lot for a greenfield development: Ensuring floors in buildings exist at the street level allows shops and spaces to line the street. Equal neighboring rooftop heights allow continuous parks (just step from the edge of one building to the next). Not only that, but because buildings on opposite sides of the street are now touching, you can do things like allow shared stairs or elevators that service both buildings, lowering costs further.

Several street cross sections; Numbers indicate floor #.

What about construction? Trucks and cranes need space to actually build and tear down buildings, right? OK, give them that space. They don’t need the whole street. A given dedicated shaft for cranes can service anywhere the crane can reach, requiring much less space relative to the service area.

What about sunlight and windows? People want a certain amount of contact with the outside. Keep in mind here that people are different and that some aren’t nearly as concerned with sunlight or windows in their own home. There are also those that desire window contact but not sunlight, and perhaps vice versa. Imposing one view of peoples’ needs is premature here, and extremely expensive. Let the people that are willing to make the tradeoff make it. Full-spectrum LEDs can emulate sunlight in sufficient intensity with minimal space requirements.

For others, you can bring columns of outside enclosed with windows and textured facades down into buildings periodically. Or for more space and sunlight or a more interesting view, you can limit some buildings to dense thin rows lining the streets or even just towers with lower construction filling in the gaps (or use some kind of per-floor area ratio limiting density but allowing a variety of shapes). This is an area worth further exploration. Assessments on the value of spaces with or without windows, sunlight, or nice views can allow charging negative externalities to existing property caused by new development to the new owners, allowing the market to regulate more of this.

What about fires and fire engines? Here we must be very humble: Even glass and metal skyscrapers can go up in flames. Fires are a force of nature to be dealt with with the utmost caution. All buildings will need to be fitted with fire sprinklers at the minimum. Firemen will thankfully be close enough not to require fire engines or emergency lanes, but it’s possible many more firemen will be needed in extreme emergencies. This would be an area of research with large potential wins. More metrics and more confidence is needed in the strategies available to fight fires in dense city sections. A few possible techniques to start with are CO2 emergency networks, brick or solid stone firebreak building sections, and more extensive fire drills. This is a critical thing to address, but I believe it is possible.

Package Delivery

Let’s tie up a thread from earlier: Package delivery. Today we use small trucks to deliver packages in many places. Last mile delivery is about half the cost of shipments in the US, doesn’t happen on Sundays, and adds a day latency for in and out-going packages. There’s a startup called Starship with a bunch of $5k robots that use machine learning and a bunch of logic to navigate sidewalks and mostly deliver food for $2 a delivery.

Hey, that’s actually not a bad idea. Let’s just let the bots share the one-way lanes at first and give them a separated lane under raised streets. That way they can go faster to start with, a lot faster in the separated lanes, and don’t need nearly as much logic, machine learning, sensors, or wheels to navigate. We can even have one big lane and one small one or otherwise mix smaller robots in for smaller packages. Actually, this is pretty much exactly how PRT works, just with packages instead of people. You could even upgrade the network later to let people ride if it’s reliable enough. Robots deliver to lock boxes or directly to an owned space. Delivery fees should be… under a cent for smaller packages?! Hold on…

But that’s exactly it: An autonomous toy car can deliver your food for less than a cent, and in a minute or two along separated lanes. It can also deliver similarly beer, medication, toilet paper, and just about anything small you might get from Walmart. Or Amazon. A larger 2 foot wide bot could do it for a few cents. It’s just a box with a motor, battery, computer chip, camera, and wheels.

The effects this would have on retail and food are enormous. Convenience stores and cheaper restaurants have the option of moving to much cheaper locations and just serving all their products remotely. There’d be no need for waiters or customer-friendly warehouse aisles. These stores would serve a far wider area, competing for customers that can choose from any of them. More stores would pop up serving niche items previously impossible to justify economically. If the food arrives in a few minutes, it’s still fresh and you can just send dishes back with another bot. There’d still be space for higher end restaurants, and a new market for space to sit and eat decoupled from cheaper eateries.

An Arcology

We talked about density earlier. Manhattan has an average of 72k ppl/mi². Paris: 56k, Tokyo: 16k, Barcelona: 42k, New York City: 27k (remember how city boundaries aren’t well defined?).

There’s also the amount of built space (including roads and parks) per person in a city. This is distinct from inverse density: Imagine a single occupant 2-story house. The built space is twice (briefly ignoring usable space ratios) the land the occupant is using. In NYC the average built space is 1010 ft²/person (but Manhattan is likely lower), Manilla: 250, Paris & Tokyo: 500, LA: 3660. (note differing Paris and Tokyo densities)

Keep in mind that it is usable space per person in a given setting and not land population density that determines how crowded it feels, as well as the fact that density and crowding are primarily a function of demand and the market as a whole. Private interests self regulate crowding: Cities only get as crowded as Paris if people and businesses want to be there. Overcrowding is partly psychological: Cultures have varying ideas of personal space in different contexts. The design here just allows the density of land use (and says nothing of crowding) and infrastructural capacity to increase beyond what is currently achievable.

We also mentioned raised streets earlier. Let’s examine a valuable district where all the streets are raised and incentives on rooftop height coordination have led to an even set of rooftops with a considerable amount of street inset 1–3 floors from the collective rooftops.

(area near a train stop) This is what peak efficiency looks like: Streets not filled in and spiky high-rise demand gradient not shown. The outer space is also buildings. (it’s buildings all the way down)

Imagine a world with a massive single building center near a train stop, a sort of groundscraper. However, the “single building” is secretly many adjacent buildings masquerading as one. The roof has plenty of parks and trees and there are inset streets leading everywhere, with elevators you can take down into the building strewn throughout the rooftop. You can sort-of think of the whole place reversed, as a massive underground building extending into the ground from the vantage of a set of streets and parks. The building has many skylights and roof fans bringing light and air into it, and at the sides of the building are elevators leading to the streets below.

Density and the buildings themselves may be by no means this uniform, but consider this: A place like this with a height of 35 floors (the height of Sidewalk Labs’ PMX, the modular plyscraper), 500 ft² built space per person, and 1.4 gross : usable square feet would give 1.4 million ppl/mi², or 20 times the average density of Manhattan. Even a height of 5 floors, a traditional limit for wood construction and requiring no elevators, would give 200k ppl/mi², which is denser than most parts of Manhattan. A height of 12 floors with streets outside around floor 6 might have an average of 10 floors of usable space and still not require elevators.

A single street level actually isn’t enough to handle some of these densities, even with the affordance for higher street to floor ratios granted by raising the streets. You’d need multiple floors similar to what malls have, but with more mixed use development and integrated into the street network. Periodic half-floor offsets can promote varying paths and aid exploration among citizens, driving street sociability and discoverability. You don’t need that many street levels, either: 1 street level near the roof per 10 floors total height should be sufficient.

And indeed, the densities involved here easily dwarf existing cities without much expense. You could get anywhere in no time; There’d be so many things reachable you’d never have to go far at all. Every shop would have hundreds of thousands of potential customers regardless of location. Every school, library, and park would as well. The economic opportunity is unreal in cities with this level of demand, but you have to build the infrastructure for it, and you need the demand first.

Demand for Raised Streets

How much demand? At some point the average land value assessment along a particular section of street reaches a threshold where the city figures it will be profitable to raise the section. The financial model I’m imagining here is: The city sells the section to a large developer that then works with and sells space to adjacent building owners and other parties.

In order to lower construction costs overall, the city can set standards that specify that any building to the truck lane side of the street use a wall capable of bearing more weight than necessary. All that’s left to raise a section of street is to construct a long building with a walkable roof. The new building would be built attached on one side to the special load bearing walls. The special walls are an investment pool that gives a return when the city builds the raised street and repays building owners.

For example, if street about to be raised is 30 ft wide, such a planned building would then have a width at its base of 20 ft (to accommodate the truck lane) and higher floor widths of 30 ft. Since the new building space can be sold, the actual cost of raising the pedestrian street is just the difference between the cost of placing it on a roof vs on the ground.

This shows the demand threshold for raising a street is just about when the land is valuable enough for private investors to want to build a multi-story building with a walkable roof.

That demand doesn’t come so easily: If everyone wanted to live in a new town, suburbs would fill instantly with starry-eyed folks. Reality is much more gradual and uncertain. Economic incentives may greatly aid this process, but nothing happens overnight and infrastructure must accommodate the world as it is, not as how it wishes it could be.

Lower Densities

However, even starting with large residential plots and cheap land, the aforementioned infrastructure performs admirably: At first, while density is low, cars are perfectly suitable. As density increases, people can sell their cars and begin biking to transit. Let’s assume a constraint on mass transit stops where people want to be able to let their kids walk to a primary school near the stop. That gives a population floor of about 4000 people per stop. People can bike to transit above a density of 900 ppl/mi² if there are streets along rail lines increasing reachable area. For walking, this gives a figure of 9000 ppl/mi². Because the infrastructure is incremental, the city can add rail lines, expand train stations and streets, and even add more streets and parks as demand appears.

This infrastructure is also legitimately cheaper and better than typical car-oriented suburbs at being, well, a suburb. People use suburbs to refer to a wide range of densities, from 100–2200 ppl/mi² with a median of 1800–2000, with the variation mostly from residential lot sizes. The rail system isn’t very helpful below 900 ppl/mi², so we’ll only compare denser suburbs. With the proposed changes the streets are safer, the utilities (mostly trash & mail) cheaper, the financial incentives more aligned, the retail even more convenient, and there is a ready path of intensification to a full city.


Let’s take all these things together: In a world where you can get cheap housing in or quite close to a massive hub of culture. A world where all shipments and returns are cheaper and faster, where you can order food from anywhere quickly for no cost, where you can travel anywhere in the city with exceptional speed. A world where all the streets have shade and no cars, there’s considerably less pollution, and the view is amazing even in the densest sections.

That’s a world I’d like to see.