Reinventing Gridlock

IntroductionA Future by Design

Picture a downtown street five years from now, at the evening peak. Traffic crawls, as it always has. But look closer, and something has changed. A large share of the vehicles carry no passengers and are not trying to park. They are circling. They are autonomous, between fares, and they have worked out that cruising the block costs less than paying to stand still. This is not a scene borrowed from science fiction. It is the logical result of a technology already operating on American streets, colliding with a set of parking and curb rules written for the age of the human driver. Those rules trace back to a single invention: the parking meter, first planted on an Oklahoma City sidewalk in 1935 to turn over the scarce curb in front of shops. Nearly a century later, much of the curb still runs on the same logic.

A great deal has changed since 1935, and the pace of change has never been faster. The hard part of keeping up is not engineering, and it is not even economics. It is imagination. The instinct, when a new technology arrives, is to bolt it onto the system we already have and trust that the old rules will still fit. They will not. To move intelligently into an autonomous future, cities will have to set down the habits of the human-driver century and ask a more demanding question: not how to manage the new machines, but what kind of place we want them to help us build.

That question has two parts, and this paper addresses the first: the rules, the pricing instruments, and the policy framework that give cities the leverage to shape the autonomous transition before it shapes them. The second part — what the programmable street actually looks like on the ground, how buildings meet a curb that no longer stores cars, and how the physical design of the mobility court, the neighborhood block, and the evolve-ready garage varies across urban conditions — is the invitation for an open dialogue with the design community. Policy sets the terms. Design delivers the place.

None of this waves away economics. The economics are decisive, and much of this paper is about them: how a city prices its curb, what it charges for occupying the public right-of-way, where it stops quietly subsidizing the car. But economics are the means, not the end. Our aim in the pages that follow is to lay a foundation for a forward-thinking approach, a way for cities and communities to harness the upheaval that autonomous vehicles are about to bring and convert it into the mobility underpinnings of genuinely vibrant places that enable robust communities. The encouraging part is that the tools to do this are largely ones cities already hold. The technology will arrive whether or not we are ready for it. The only open question is whether each community shapes that arrival or is shaped by it, and that question will be answered, for better or worse, within the next few years.

Section AThe Robotaxi Is Already Here

For a decade, the autonomous vehicle lived comfortably in the future tense. That is no longer true. By late 2025, driverless robotaxis were completing more than 700,000 paid rides every week worldwide, with the United States alone accounting for over 450,000 and China for more than 250,000.1 Waymo operates a commercial, fully driverless fleet of more than 2,500 vehicles across the San Francisco Bay Area, Los Angeles, Phoenix, Austin, and Atlanta.2 These are not demonstrations on closed courses. They are paying customers, in mixed traffic, with no one in the driver’s seat.

Industry leaders expect the technology to reach large-scale commercial deployment around 2030, with broad availability across most major markets within three to seven years.3 Just as important, they expect robotaxis, rather than privately owned self-driving cars, to lead that first wave. The first autonomous future is therefore a fleet future, and a fleet is precisely the kind of thing a city can see, price, and regulate.

That timing is the heart of the matter. It gives cities a planning window of roughly four to seven years: long enough to act deliberately, short enough that delay is itself a decision. The travel and parking patterns a city permits to form in the first years of deployment will be slow and costly to unwind later, once operators, riders, and routes have hardened into habit.

There is a second, subtler reason to act now. The most effective tools for managing autonomous traffic — congestion pricing and curb pricing chief among them — have always been politically difficult, because they ask drivers who already use the road to pay for something that used to feel free. A charge applied to robotaxi fleets faces no such constituency today, for the simple reason that almost no one owns the vehicles yet. As one leading transportation scholar puts it, the rise of autonomous vehicles supplies both the opportunity and the imperative to put these tools in place: the opportunity, because no entrenched group of owners stands ready to fight them, and the imperative, because without them the technology strips away the last effective brake on urban driving.4 A rule written before deployment sets the terms of entry. The same rule written afterward is a tax on incumbents. The politics will never be easier than they are right now.

Section BTwo Futures, One Technology

The promise attached to robotaxis is genuinely appealing, and city leaders hear it constantly: fewer private cars, fewer parking lots, smoother and safer streets, and acres of asphalt handed back to people. Promoters show renderings of parking craters reborn as parks. Some of that promise is real and worth pursuing. None of it is automatic.

The uncomfortable truth is that the same technology can just as easily deliver the opposite. Left to run on purely commercial logic, autonomous fleets push in four congestion-increasing directions at once.5

First, they generate induced demand. When a trip becomes effortless and chauffeured, people take more trips, and many of those trips come straight out of transit, walking, and cycling. Analysis of today’s ride-hailing found that roughly half of its trips would otherwise have been made by transit, on foot, by bicycle, or not at all.6

Second, they drive empty. A vehicle with no need to park has every incentive to keep moving between fares. One detailed model of downtown San Francisco found that, because cruising can be made cheaper than parking, the effective cost of parking could fall by ninety percent — a mechanism Lever 1 (Reinventing the Curb) examines in detail.7

Third, they encourage sprawl. When the commute converts from dead time into time for work or rest, the friction that keeps people living near the center weakens, and travel distances stretch outward.8

Fourth, they cannibalize transit. Absent deliberate integration, a cheap and seamless robotaxi ride competes directly with the bus and the train. Ridership falls, fare revenue follows, service erodes, and still more riders abandon transit, a spiral that one model projects could cut transit ridership by as much as three-quarters in an unregulated scenario.9

None of this is destiny, and the evidence is genuinely contested. Other studies find that autonomous vehicles can smooth traffic flow and sharply cut collisions, and early insurance data show far fewer claims for driverless fleets than for human drivers. The point is not that robotaxis are inherently good or bad. The point is the one thing optimists and pessimists agree on: the outcome is not dictated by the technology. It is dictated by whether cities price the externalities and regulate the behavior. Autonomy does not solve the underlying problem of how scarce street space is shared. It amplifies whatever rules it finds in place.

So cities face a fork, and the same vehicle stands at the head of both roads. What decides the direction is policy, and the single most powerful instrument is one cities have leaned on for a century without always naming it: the price and design of the curb. In most downtowns, the cost and scarcity of parking is the most effective limit on private car travel there is. Robotaxis threaten to dissolve that limit by making parking optional. A city that reprices and redesigns its curb keeps its hand on the lever. A city that does nothing watches the lever snap off in its hand.

Section CDefinitions and Roles

Before turning to what cities should do, two questions need clear answers, because the rest of this paper depends on them: what, precisely, are we regulating, and who has the authority to regulate it? Both are murkier in public debate than they need to be, and both, once clarified, point to the same conclusion: the decisions that matter most belong to cities.

What a robotaxi is, and is not

The phrase “self-driving” has been stretched to cover two very different things, and the confusion is not harmless. At one end are consumer driver-assistance systems, the adaptive cruise control and lane-keeping features sold under names like “Autopilot” or “Full Self-Driving.” These are, in the language of the international standard, Level 2 systems: the car assists, but a human driver remains responsible for the vehicle at every moment and must be ready to take over instantly.10 At the other end are true robotaxis, Level 4 systems that drive themselves within a defined service area with no human responsible for the driving task, and frequently no one in the driver’s seat at all.

The difference is not academic. As the legal scholar Bryant Walker Smith argues, a vehicle marketed as self-driving should actually drive itself, with the company that builds and operates it, rather than a human occupant, answerable for how it behaves.11 That single distinction governs everything downstream. A Level 2 car that blocks a crosswalk or runs a light has a driver who can be cited; a Level 4 robotaxi does not, which is exactly why the enforcement tools built for human drivers come apart against it, as the next part will show. This paper is about Level 4 fleets: the genuinely driverless vehicles, operated by companies, that a city can regulate as a fleet rather than as ten thousand individual motorists. When we say robotaxi, that is what we mean.

Who governs the robotaxi

The second question is jurisdictional, and here the division of labor is clearer than the debate suggests. Three levels of government touch the robotaxi, and they touch different parts of it.

The federal government regulates the vehicle as a product. Through the National Highway Traffic Safety Administration, Washington sets the motor-vehicle safety standards a car must meet, grants the exemptions that let a vehicle ship without a steering wheel or pedals, and requires operators to report crashes. The federal autonomous-vehicle framework, updated in 2025, is explicit about this scope: it governs whether the vehicle is safe to build and sell.12 It does not tell a city how to manage its curb, where fleets may operate, or how its streets are designed.

States regulate the vehicle in operation: registration and insurance, the licensing or permitting of autonomous operators, and the rules of the road. Much of the deployment-permission regime that lets robotaxis carry passengers for money lives at this level.

Cities govern the street and the service that runs on it: the curb and its price, parking and loading, land use and zoning, street design, congestion pricing, and how all of it integrates with transit. This is the level at which the questions this paper cares about are actually decided.

Set out this way, the point becomes hard to miss. Washington decides whether a robotaxi is safe enough to sell. The city decides whether it makes the city better or worse. The levers that determine whether autonomous fleets relieve congestion or manufacture it, whether they feed transit or starve it, whether the curb becomes more valuable or simply more crowded, are overwhelmingly local. That is both the good news and the obligation, in a single fact: cities cannot wait for Washington to solve a problem Washington does not regulate.

What cities and states should regulate

If the vehicle’s safety is the federal concern, what remains for state and local regulators is the service and its effects. The agenda is the familiar one for any network industry, adapted to a fleet of autonomous cars. First, externalities: the congestion, curb occupancy, emissions, and empty mileage a fleet imposes on everyone else, which is the subject of the pricing levers in this paper. Second, competition: ensuring that a city does not wake up captured by a single operator, and that the market stays open to more than one network. Third, rider and public protection: fair and transparent pricing, accessibility for people with disabilities, sound data practices, and basic service quality.13 A useful feature of autonomous fleets is that many rules can be built in rather than enforced after the fact, because a vehicle can be programmed to honor a geofence, a speed limit, or a no-stopping zone in a way a hurried human driver cannot. We return to this agenda in detail in the final lever; the point here is simply that it exists, and that it belongs to the levels of government closest to the street.

With those two questions settled, what a robotaxi is and who governs it, the path forward comes into focus. The rest of this paper is about the local level, because that is where the decisive choices live. It is organized around six levers a city can begin pulling inside the current window: reinventing the curb, planning for where fleets rest and recharge, reforming building and zoning rules, protecting equity, integrating transit, and setting the market rules of the road. The first, and the most important, is the curb itself.

Section D · Lever 1Reinventing the Curb

If robotaxis have one inescapable physical fact, it is this: they begin and end every trip at the curb. The curb is where an autonomous fleet touches the city, and it is the one piece of that fleet’s world a city already owns outright. It is also the piece still governed by the oldest rules in the system. The meter logic of 1935 was built to do one thing, turn over a scarce space in front of a shop, and for ninety years it has done that job through a posted time limit and the threat of a ticket. That arrangement is about to meet a vehicle that cannot be ticketed, never tires, and has no particular reason to park at all. Reinventing the curb is the first lever, and the one on which the others depend.

The tool that no longer works

The dominant instrument for managing curb parking today is the posted time limit backed by citation enforcement, and it is a blunt tool even now. It tells a driver when to leave, but says nothing about what the occupancy is worth; it requires a steady supply of enforcement labor to mean anything, and it caps the value a high-demand space can return no matter how scarce that space becomes. Against an autonomous vehicle, it stops working altogether.

Consider how a robotaxi defeats the time limit without breaking a single rule. A two-hour space is free for two hours, so the vehicle simply leaves after two hours and re-parks, around the block or in the next open space, as many times as the day requires. A metered space cannot be enforced against a car with no driver to feed the meter, so the vehicle parks, declines to pay, and keeps a watchful electronic eye out for the enforcement officer who, on most blocks, comes by only occasionally. The entire apparatus assumes a human being who would rather not risk a citation. Remove the human, and the apparatus has nothing left to push against.14 Enforcement does not scale either: no city can hire ticket-writers fast enough to police a fleet that can re-park itself indefinitely. The time limit was a clever answer to the problem of 1935. It is no answer to the problem of 2030.

Pricing the curb by what it is worth

The replacement is to price the curb directly, by the value of the time a vehicle occupies it, rather than policing a clock. The mechanism is progressive pricing, also called escalating or graduated pricing: a low rate for the first hour or two, then a higher rate for each additional hour. The first hour might cost a dollar, the second two, the third three. Short stays stay cheap and plentiful, long stays grow progressively expensive, and turnover is produced by price rather than by a ticket on the windshield.

This inverts the way most meters behave today. Traditional metered pricing is frequently concave: discrete intervals and daily caps mean the effective hourly cost falls the longer a driver stays, handing a volume discount to exactly the all-day occupancy a busy block can least afford. Progressive pricing flips that curve to a convex profile that makes each added hour cost more than the last. It is not an exotic idea. The rising-block tariff is the ordinary structure of household water and electricity bills in jurisdictions around the world, and cities would simply be applying to the curb the same logic residents already accept on their utility statements.15

The evidence on progressive parking pricing is young but consistent. The peer-reviewed work finds that when a market holds both short- and long-stay drivers, progressive pricing raises turnover and yields more revenue than flat hourly pricing while efficiently segmenting demand.16 That same research delivers a caveat city leaders should absorb before adopting it: progressive pricing does not improve aggregate social welfare over flat pricing, because the additional revenue the city collects offsets the surplus drivers would otherwise keep.17 For this paper, that is not a flaw. It is the reason the tool fits. A city dismantling curb parking to build the pick-up, drop-off, and loading infrastructure an autonomous future demands needs precisely an instrument that converts the value of a scarcer curb into revenue to fund the transition, while keeping short-visit spaces available for the businesses that depend on them. Progressive pricing is a turnover-and-revenue instrument, and a period of reallocation is exactly when a city wants one.

It is already working

None of this is theoretical. Progressive curb pricing is in operation across cities of every size.

Table 1. Progressive curb pricing in practice.

City	Structure	Example rates	Signature feature
Omaha, NE	Citywide; three demand zones, hourly escalation, no time limits	$1/hr base, escalating to a daily cap	Revenue up 30–40%; no overstay tickets
Auckland, NZ	Citywide escalating on-street	Steps up sharply after the first two hours (central zone)	International flagship; publicly accepted
Pasadena, CA	Occupancy-based base plus duration multiplier	Roughly $1.25–$2.00/hr base, escalating after two hours	Strategic Plan urges replacing time limits
Sacramento, CA	Tiered “n+” zones; rate steps up after a threshold hour	$3.00/hr base (raised July 2025), escalating in higher tiers	Escalation layered on a raised base rate
Manitou Springs, CO	Small-town progressive rates	Escalating hourly	Proof the tool scales down (pop. ~5,000)

Omaha runs the most complete example in the United States, a citywide structure in place since 2023. Working with Walker Consultants, the city replaced its time-limit meters with three demand-graded zones, lowered base rates to a dollar an hour, and let the rate escalate hourly beyond each zone’s threshold. It abolished time restrictions and stopped ticketing drivers for overstaying; a vehicle may remain as long as it likes and simply pays more per hour to do so. A Park Omaha official reported parking revenue up thirty to forty percent after the change.18 Auckland is the international flagship, having taken progressive pricing citywide rather than confining it to the center, with central-zone rates that step up sharply after the first hours to signal that the curb is for short stays. Its reforms are now broadly accepted by the public and local business, evidence that the politics are winnable.19 Pasadena offers the most sophisticated American model, pairing an occupancy-based base rate with an escalating duration multiplier, and its Citywide Parking Strategic Plan formally recommends replacing time limits with an escalating rate model, the very swap this paper proposes.20 Sacramento layers the same escalating logic onto its meters, holding a base rate and stepping it up after a threshold hour across tiered “n+” zones, and neighboring West Sacramento has adopted the same approach in a stadium district.2122 Manitou Springs, a town of roughly five thousand, proves the tool scales down, having moved to progressive rates citing both Shoup and the Auckland precedent.23

Beyond the parked car: pricing the moving curb

Repricing parking, however, only binds a vehicle that intends to park. A robotaxi has a third option no human driver ever seriously entertained: it can decline to park at all. After dropping a passenger, an empty autonomous vehicle can drive to a distant free space, return to its home base, or simply circle the block until the next fare. Because an electric vehicle creeping through slow traffic burns very little energy, cruising can be made astonishingly cheap. Modeling of downtown San Francisco found that these strategies could cut the effective cost of parking by ninety percent, more than doubling vehicle travel into and within the core, and that cruising could undercut even a four-dollar curb at roughly fifty cents an hour.24

The lesson is that parking pricing alone, however well designed, leaves the curb’s most important constraint with a hole in the middle of it. The ability of an autonomous vehicle to cruise blurs the line between parking and traveling, and a price that applies only when the vehicle is stationary simply pushes it into motion. The remedy, proposed in the same research, is to charge for occupying the public right-of-way whether the vehicle is parked or moving: a time-based charge a fleet pays for its presence in a priced district, full stop. A vehicle that cruises to dodge a parking fee then pays anyway, and the incentive to manufacture congestion disappears.25

Cities can close the same loophole from the operator’s side by pricing empty and low-occupancy miles directly. A per-mile charge on zero-passenger trips, with a premium on solo rides, makes deadheading and aimless circling expensive for the fleet that orders them, and rewards pooling and prompt return to service instead.26 Taken together, a progressive parking rate, a right-of-way occupancy charge, and an empty-mile fee give a city a pricing surface that covers the curb in all three of its states: occupied, hunting, and idle.

From storage to flow

Pricing settles how much demand arrives at the curb. Design settles whether the city can absorb it. The two are halves of one lever, and the second has gone neglected for as long as the first.

For nearly a century the curb has been, above all, a place to store private cars. An autonomous future needs very little of that and a great deal of something the old curb was never shaped to do, which is to load and unload people and goods quickly, safely, and all day long. A robotaxi fleet does not generate demand for storage. It generates a relentless stream of short stops: a passenger in, a passenger out, a parcel handed off, then gone. As parking demand falls and progressive pricing manages what remains, the lanes that pricing frees should be converted to the work the fleet actually creates, which is passenger pick-up, drop-off, and loading. This is where curb revenue earns its keep, paying for the redesign of the lanes themselves.

A painted white zone and a “no parking, loading only” sign are not equal to the volume a fleet will bring. Pick-up and drop-off at fleet scale is not a sign problem. It is a design problem, and the most durable way to solve it is not to widen the public curb without end but to give the pick-up, the drop-off, and the loading a purpose-built room of their own. We call that room the mobility court.

The mobility court

The mobility court is what the curb becomes when it stops storing cars and starts moving people. The organizing idea is almost embarrassingly simple, and it can be read at a glance: cars at the edge, people through the middle. Vehicles travel a drive aisle along the outer edge and never leave it. People move down a protected interior, the pedestrian spine, and never step into traffic. Between the two sits the transfer zone, the few feet where a rider crosses from the spine to the open door of a waiting vehicle.

That one move, holding the vehicle realm and the people realm apart and letting them touch only at the door, is what makes the court safe enough to run at fleet volume and pleasant enough that a person would choose it over the chaos of an unmanaged curb. The asphalt that spent the human-driver century holding cars still is returned to its only honest purpose, which is moving people.

Buildable today, to the foot

The mobility court is not a rendering waiting on a breakthrough. It is a dimensioned plan a city could pour next year, drawn to the vehicles and the codes we already have.

The geometry is ordinary on purpose and calibrated to a mid-density commercial context with available setback and surface area. The mobility court takes a different physical form in a zero-lot-line urban neighborhood, at the ground floor of a mixed-use podium building, at a transit station entrance, or at the edge of a high-volume event venue. Each of those conditions produces a different configuration of drive aisle, pedestrian spine, transfer zone, and charging provision, each involving design decisions that pricing and zoning policy can incentivize but cannot specify — a point for conversation with urban planners, building designers, and parking consultants. Transfer stalls are sized at nine by sixteen feet for a quick passenger handoff and trunk stalls at nine by twenty for loading, with six-foot shared strips between them so that a door or a tailgate never opens into a moving aisle. The pedestrian spine is a true eight feet, wide enough to be a place rather than a leftover. The electric-vehicle stalls do double duty, charging private cars through the day and topping up autonomous vehicles in the off-peak hours when the court would otherwise sit idle. Nothing in the plan waits on a technology that has not shipped: it serves a human driver dropping a child this year and a driverless fleet the year it arrives.

That is the discipline behind the whole proposal, and it is worth naming: operations first, technology second. A design earns its place by working on an ordinary Tuesday, not by dazzling on a launch-day demonstration. The court passes that test because every dimension on it answers to a vehicle and a code that exist today.

Who builds these courts, and how a city rewards the developments that host them rather than mandating them by formula, belongs to the zoning lever, and we take it up there. The point here is narrower and comes first: the form exists, and it is buildable now.

The hard case: the surge

Every curb design is easy to love at a quiet office on a weekday morning. The honest test is the worst hour a city can throw at it, and for the curb, that hour is the end of a sold-out event, when twenty thousand people leave an arena at once, and all want to be somewhere else immediately.

The mobility court meets the surge with the principle it uses everywhere, pushed to its limit: people and cars never cross. The crowd leaves the venue via a wide main walkway that no vehicle ever uses, flowing toward transit and the streets beyond without a single point where a person waits for a car to pass. Only those who want a ride peel off, onto a short spur that leads to the boarding bays. The vehicles run a one-way pull-through: a car enters a single lane, takes on its rider at a bay, and pulls forward into the exit lane, never reversing and never circling back through the crowd. Between fares, the fleet waits and recharges in the very bays where it will board the next round of passengers. Movement reduces to a minimum: a charged vehicle slides out to the ready position, another rolls into the empty bay to plug in.

The pattern outlives the game. The same bays that boarded the post-event crowd do double duty on an average non-event night, serving as a charging and cleaning depot for a significant share of the regional fleet, in a parking lot that would otherwise sit empty. The owner of that lot, whether a city or a private operator, can monetize the access through direct agreements with the fleet operators. Empty, clean vehicles can rent the stadium lot through the midnight hours. It is the fastest and cheapest path to fleet-station infrastructure a city has, sparing the long and expensive build of purpose-built depots.

The busiest and most chaotic moment in a city’s mobility week is handled without people and cars ever sharing a surface. They meet at exactly one place, the door of the vehicle, which is the whole point. A pattern that holds against the post-event surge, and earns its keep through the off-night, will not be troubled by a Tuesday.

Priced by what it is worth and shaped with intent, the curb stops being a place where cars wait and becomes a place where people move. Pricing supplies the discipline, and the court supplies a destination for the demand that pricing sets loose. What remains is to win the argument with the residents who will pay for the change, and to make sure the bill they get is a fair one.

Winning the politics

The political objection is predictable, and it should be met head-on rather than waited out. When West Sacramento raised its tiered rates, a resident complained that the city was treating residents as an income stream.27 The answer has two parts. First, progressive pricing lowers the cost of the short visits that make up most curb trips and recovers money only from the long-dwell occupancy that crowds everyone else out; Omaha was able to pitch its reform, honestly, as a rate cut for most users. Second, acceptance follows visible reinvestment. Auckland directs its parking surplus to the transport authority rather than the general fund, and parking-benefit-district models return the money to the very streets that generate it. A city proposing to reclaim curb lanes for pick-up, loading, and transit can tell residents, truthfully, that progressive pricing is what pays for the improvements they will see. Survey work for the U.S. Department of Transportation finds meaningful driver willingness to accept progressive curb rates once the structure and its purpose are explained.28

The block that parks on the street

There is a harder version of the political problem, and it does not live at the meter. It lives on the residential block. Progressive pricing asks a driver to pay a little more for a long stay. Reclaiming a curb lane asks a household to give up the space it has parked in for twenty years. In the dense, mixed neighborhoods where this transition will be won or lost — apartments above shops, a car at every curb because the buildings were built without garages — the second ask is an order of magnitude larger than the first, and a city that treats them as the same problem will lose both.

The scenario to avoid is easy to picture, because versions of it have already played out. A city, persuaded by everything in this paper, walks into a neighborhood of three-story walk-ups and corner stores and announces that half the curb is being converted to loading and an autonomous drop-off court. The design is right and the long-run logic is right, and it lands on a family that owns two cars today, parks both on that street tonight, has no driveway, and has no robotaxi yet cheap and available enough to make the second car optional. The next lever will make the case that such a family may, in time, need one car instead of three. One day is not tonight. Tonight they need somewhere to put the car they still own, and a plan that cannot answer that is not a plan. It is a notice of eviction for their vehicle, and they will read it exactly that way.

Call it the transition trap, and note that it is dangerous precisely because the destination is worth reaching. The car-light neighborhood is a better place to live, and the residents who stand to gain most from it are often the ones with the least slack to survive a botched arrival. Lower car ownership is an outcome a neighborhood reaches, not a condition a city can impose from the front end by pulling out the parking and trusting people to adapt. Remove the supply before the alternative is real, and you do not pull the future forward. You manufacture the backlash that pushes it back.

The discipline that avoids the trap states easily and demands patience to follow: add before you subtract, and never subtract faster than the alternative arrives. The first thing a neighborhood should see is something it did not have before — a sheltered pick-up bay that makes the rideshare and the grocery run easier, guaranteed accessible service for neighbors who cannot drive, a managed loading zone that clears the double-parked delivery van, a mobility-court node at the multifamily building that earns the block a little revenue and gives its residents a better place to be picked up, dropped off, and charged. Only as those alternatives prove themselves, and as the households that want to shed a car begin to, should the stored-car supply contract to meet the demand that has actually fallen. The curb gives up its parking at the speed the neighborhood gives up its cars, and not one space faster.

In the densest residential urban fabric, walk-up apartments above corner stores, buildings built to the property line, no surface lots, no alleys, the physical response to the transition trap requires a different toolkit than the mobility court and the multifamily bay. The neighborhood-scale structured garage, the time-of-day curb zone, the residential permit parking framework amended to accommodate managed fleet access, and the district-level staging node serving a residential catchment rather than a single property are the physical instruments available in this context. How each of those instruments is designed, dimensioned, and integrated with the existing built fabric is the design problem that sits beyond this paper’s scope — an invitation to the design community, informed by the policy framework this paper establishes.

None of this is possible without the residents in the room, and in the room early — engaged before the first line is drawn, not consulted after the design is finished. The community that helped shape the plan will defend it; the community that had the plan done to it will fight it, and will be right to. On a residential block, engagement is not the public-relations step that follows the engineering. It is part of the engineering — the part that decides whether the rest survives contact with the people who live there. A city that brings the block real choices, this much parking against this much loading, this bay here rather than there, buys the one thing no ordinance can supply on its own: a constituency that owns the result.

The cautionary tale is recent, and it earns a study of its own. When San Francisco rebuilt Valencia Street around a center-running bike lane in 2023, it had a legitimate safety rationale, a real mobility goal, and — by a later assessment — no measurable economic harm to the businesses along it. What it did not have was a baseline parking and loading utilization study, the foundational step that would have established how the corridor actually functioned before a single space was removed, and given the city factual ground to stand on when opposition came. It was torn out inside eighteen months anyway, after an outrage that set cyclists against merchants against residents against commuters, each convinced the street had been remade for someone else at their expense. The intensity of the merchant opposition was driven as much by the loss of loading access as by parking, a distinction the public debate never made clearly. On a mixed-use corridor, loading is an operational necessity; its removal is felt immediately, on the first delivery window of the first morning. Parking loss is felt differently, and often later. A reclamation design that had protected loading function while repricing stored-car supply would have separated those two grievances before they merged into a single political force.

There is also a design lesson embedded in the outcome: the center-running configuration, placing the protected lane in the middle of the roadway rather than adjacent to the curb, maximized the displacement impact when a lower-impact curbside alternative existed from the beginning. The SFMTA ultimately replaced it with exactly that curbside design. The physical solution was available, it simply was not chosen first.

We take Valencia apart in a companion study, because how a city loses a good idea is as worth studying as the idea itself. The lesson it leaves this lever is short. The curb can be repriced by ordinance and reclaimed by permit, but it is only transformed with the consent of the people who live along it — and that consent is earned in the here and now. Valencia’s deeper lesson for the AV context is that the political sequence and the physical sequence must run together. The project that adds the mobility court, the managed loading zone, and the pick-up bay before removing a single stored-car space gives residents a visible gain before they absorb a visible loss. Valencia inverted that sequence. This paper’s “add before you subtract” approach is the correction, and Valencia is precisely the proof of what happens when it is not applied.29

Why the curb comes first

The curb is the first lever for a reason. It is the asset a city already controls, it is where every robotaxi trip begins and ends, and it is the instrument that funds everything downstream. A city that modernizes curb pricing before the fleets arrive will enter the transition with a calibrated revenue tool and a curb already operating on the same dynamically priced logic the fleets use. A city that waits will find itself managing ninety-year-old meters against a twenty-first-century fleet, discovering one empty circling car at a time that its most reliable tool for governing the street has quietly stopped working. Every lever that follows, where the cars rest and recharge, how buildings and zoning adapt, how equity and transit are protected, and what rules bind the market, assumes a curb a city can still price and still shape. That is the work of this first lever, and it can begin now.

Section E · Lever 2Where the Cars Sleep

The curb lever governs the robotaxi while it is working: arriving, loading, moving, and paying for the space it occupies. But a fleet is not always working, and a city is not only its moving traffic. Two quieter questions decide nearly as much as the curb does. Where does the fleet go when it is carrying no one, to wait, to recharge, and to be cleaned? And, on a longer horizon, how many cars will a city actually need? The answers are linked, and they point to an opportunity that the parking industry, of all industries, is best positioned to seize. The off-street facility, written off by some as a stranded asset of the car age, may turn out to be one of the most useful pieces of infrastructure in the autonomous city.

Fewer cars on the road, and fewer in the driveway

Start with the trajectory, because it is widely misread. The first and most immediate effect of the robotaxi is not on the private car at all. It is on the human ride-hailing driver. The robotaxi is, before it is anything else, a ride-hail trip without the person in the front seat, and that is where the early disruption lands. Uber’s own chief executive has said plainly that autonomous vehicles will take over the work of human drivers over roughly the coming decade, with hybrid networks of people and machines in the interim, and in Austin Uber reported that its Waymo vehicles were already busier than ninety-nine percent of all drivers.30 Forecasters expect robotaxis to capture a large share of ride-hailing trips as fleets scale, even as, in a wrinkle that surprises people, the overall market for on-demand rides keeps growing in the cities where robotaxis have launched.31 The structural point for a city is simple: the first wave converts a driver-based service into a fleet-based one, and a fleet is a thing a city can plan for.

The second effect is slower, larger, and far more consequential for how a city looks and feels. As driverless service matures and knits together with transit and micromobility, the calculation that leads a household to own two or three cars begins to change. The decisive insight is old and well documented: most households do not actually use their vehicles at the same time. Analyzing national travel data, researchers at the University of Michigan found that on an average day roughly eighty-four percent of households have no overlapping trips at all, and that a single shared self-driving car operating in a return-to-home mode could in principle cut average ownership from about 2.1 vehicles per household to 1.2, a forty-three percent reduction.32 Separate modeling of within-household sharing finds reductions on the order of twenty percent from that mechanism alone.33 The second car, and the third, exist largely to cover the rare hours of conflict. Remove the need to own a depreciating asset for those hours, and the math shifts.

Made concrete, the future looks something like this. A family of four that today keeps two or three vehicles finds that one is enough. On a weekday, one parent takes the household car to commute. The other works from home, or close enough to it to handle the day’s errands and meetings by some combination of micromobility, transit, and the occasional robotaxi. The children reach school and after-school activities the same way. The single owned vehicle, no longer duplicated two or three times over for the few hours of overlap it never really had, is freed for what families prize it for most: the weekend excursion, the trip the shared modes serve least well. None of this is automatic, and we will say so more than once. The same Michigan study that projects fewer cars also projects that each remaining car would travel far more, on the order of seventy-five percent more miles, and other researchers caution that without supporting policy the sheer convenience of automation can raise vehicle ownership and travel as easily as lower it.34 Fewer cars is a possible outcome, not a guaranteed one. It depends on the curb being priced, on transit and micromobility being good enough to lean on, and, above all, on the groundwork being laid now, before habits and infrastructure harden around the opposite result.

Less parking, but not no parking

If households shed cars and ride-hailing converts to fleets, the demand for parking falls, and falls sharply. A single shared autonomous vehicle can serve the trips of many private cars, by one well-known estimate roughly eleven of them, and studies of parking in a shared-autonomous future project steep reductions in the land a city must devote to storing idle vehicles.35 This is real, and it is the source of the genuine optimism about parking craters reborn as parks. But less is not none, and that distinction is the whole of this lever. A fleet sized to meet the morning and evening peaks has surplus vehicles during the midday lull and overnight, and those vehicles have to wait somewhere. The demand does not vanish. It changes shape, migrating from thousands of scattered private spaces, each holding one commuter’s car for nine hours, to a smaller number of concentrated facilities where fleet vehicles stage, recharge, and are serviced. The question is not whether robotaxis need off-street space. It is who provides it, where, and on what terms.

From parking box to fleet hub

Here is where the parking industry’s supposed liability becomes its asset. Robotaxis already sleep somewhere, and that somewhere is the off-street facility. Waymo’s largest operational depot, in the Bayview district of San Francisco, services hundreds of vehicles that pull in to park, recharge, and be cleaned and maintained by human crews before heading back out to the street.36 An entire fleet-services industry has formed around this need in a remarkably short time. Avis manages depot operations, charging, and maintenance for Waymo in Dallas; Moove does the same in Phoenix and Miami; Lyft is building a purpose-built autonomous-vehicle facility for the Nashville fleet through its Flexdrive subsidiary; Uber provides cleaning and charging in Austin and Atlanta in exchange for a share of revenue; and specialist operators such as Terawatt run dedicated fleet-charging sites near Los Angeles.37 Charging is the binding constraint. Unlike a private electric car that a person plugs in overnight, a driverless fleet cannot rely on anyone to handle the cable, so it depends on dedicated charging depots that double as operational hubs, with maintenance, cleaning, and eventually automated charging under one roof.38

For an owner of structured parking near a city center, this is a second act. The box built to store commuters’ cars for eight hours a day is, with modest adaptation, well suited to stage, charge, clean, and service a fleet that needs to stay close to its riders to avoid long empty trips back and forth. As private-car parking demand recedes, fleet operations can backfill the space, and the facility that once collected hourly parking fees becomes a hub that keeps the autonomous network running. The same buildings can absorb the pick-up and drop-off that the curb lever wants to pull off the public street, charge and store shared micromobility, and stage the delivery robots discussed below. The drop zone is only the beginning of what an off-street facility can offer a fleet.

The binding constraint on that second act is not floor area; it is electrical capacity. A robotaxi does not charge the way a commuter’s car does, plugged in overnight and idle for twenty hours a day. It runs continuous revenue cycles and returns to the depot for short, high-power charges many times a day, which turns a fleet hub into a serious electrical load. A core depot of roughly two dozen fast-charging bays, each drawing 150 to 350 kilowatts, can pull on the order of four to eight megawatts at peak, and a depot cycling hundreds of vehicles a day can need several megawatts of firm capacity, the kind of service a single garage rarely has on hand.39 Three implications follow for any operator weighing the conversion. First, the limiting factor is the utility interconnection: securing that much power often requires an upgrade that is expensive and slow to schedule, so the grid conversation has to start years before the fleet arrives, not after. Second, the load has to be managed, through on-site battery storage and smart scheduling that charges off-peak and shaves the peak, or the demand charges alone will erode the economics. Third, because no driver is present to handle the cable, the charging itself has to become hands-free, through robotic connection systems built for vehicles that arrive, dock, charge, and leave without anyone touching them.40 The garage that lines up power, storage, and automated charging early is the one that can host a fleet at all. The one that treats charging as a tenant-improvement afterthought will find the asset stranded for the one use the market is actually offering it.

Two cautions keep this opportunity from curdling into its opposite. First, fleet facilities are not invisible. Neighbors of existing depots have complained about noise and light, and the city of Santa Monica went so far as to sue to enforce a depot’s operating curfew.41 These hubs belong in places zoned and designed for them, in industrial corridors and adapted structures, not dropped onto residential blocks by default. Second, and more important for policy, a city should resist the temptation to conjure cheap fleet parking into existence by subsidizing it on the periphery. Cheap, distant, publicly supported parking simply lowers the cost of keeping vehicles in service and rewards the empty deadhead miles between the fringe lot and the core, undercutting the very curb pricing the first lever installs.42 The principle is the one running through this entire paper: price the space, and let the fleet operator internalize the cost of resting its vehicles rather than socializing it onto the public. Provide for the idle fleet, by all means. Do not pay to warehouse it.

When the court isn’t busy

The stadium case at the end of the previous lever is the most visible version of a more general idea, and the general idea is the one this lever should actually rest on. Almost every mobility court the previous lever proposed sits idle for some predictable share of the day. A grocery store’s court is busy at lunch and on weekend mornings and quiet after the store closes for the night. A multifamily building’s pick-up zone is full between seven and nine in the morning and again between five and seven in the evening, and underused for the long hours in between. An office tower has the opposite curve: heavy at the morning and evening peaks, dead through the weekend. Each of those properties already owns a small piece of the very infrastructure a fleet needs to stage between fares. None of it is currently asked to earn a second shift.

The proposition is straightforward enough to draw on a napkin. Instead of building a few large, centralized depots on the urban periphery, with the entitlement fights and the deadhead miles and the noise complaints that follow, a city can let its existing mobility courts host the fleet at the hours those courts are not otherwise busy. A robotaxi finishes a fare in a residential neighborhood at eleven at night, pulls into the multifamily building’s pick-up bay around the corner, plugs into a charger, and waits for the next call. A vehicle that ends its shift downtown at six in the evening parks in the grocery store’s court after the store closes at ten and stays the night. Each property owner collects a staging fee that helps underwrite the cost of building the court in the first place. Each vehicle saves the deadhead trip to a distant depot. The fleet’s resting infrastructure is distributed across the city the same way its demand is, and that is what reduces the empty miles the rest of this paper has been working to avoid.

Three things have to be true for the distributed model to work, and all three are within reach. The first is that the connection between vehicle and charger has to happen without a person. We noted this in the depot discussion above, and the same automated-charging systems being developed for centralized fleet hubs serve a distributed network just as well, with the additional virtue that a court hosting a handful of stalls is a cheaper place to pilot the technology than a depot hosting a hundred. The second is that cleaning and routine service still need hands, and a sensible model uses the workforce already on site at properties large enough to keep one: the stadium’s event-day staff, the multifamily building’s maintenance crew, the grocery store’s overnight stockers. A fleet operator pays them for the work, and the property gains a small revenue line and a few hours of work for people who are there anyway. This is the lever that begins the workforce transition Lever 4 (Who Gets to Ride) will return to in detail, not as a future problem to plan for but as a present opportunity to seize. The third is that the dispatch software has to know how to share. A staged vehicle is preparing for service, not stored, and the moment a paying fare requires the bay it is sitting in, the staged vehicle moves. Active service always preempts staged dwell, with no exceptions. That is a software rule, not a building rule, and it is the same kind of rule Lever 6 (Who Owns the Network) will be asking a city to require of any fleet that operates on its streets.

The reframing this enables is worth naming, because it changes the cost case for the entire previous lever. A grocery owner asked to pay for a mobility court out of nothing but the morning pick-up demand may not pencil the project. The same owner offered the staging revenue from an overnight fleet may. The court that Lever 1 (Reinventing the Curb) wants poured does not have to justify itself on a single shift of use any more than a parking garage does. It earns its keep on a second shift, on a third, and on a midnight one, the same way the stadium does and for the same reason.

There is a useful word for what these courts then become, and it is not depot. A depot is a destination; a distributed mobility court is a node, identical in form to the courts that handle active pick-up, distinguished from them only by the hour at which the vehicles inside are sleeping rather than working. The fleet does not need a separate, purpose-built network of resting places. It needs the courts the city is already building for everyone else, available at the hours they would otherwise sit empty, governed by software that knows the difference between a parked car and a staged one.

When the cars are owned again

The paper has been about the fleet era, because the fleet era is what is arriving first and what cities have the leverage to shape. But the fleet era is not the final destination. The price of a Level 4 autonomous vehicle will eventually fall to a level that individual households can afford, and when it does, some of them will buy one. We are not predicting, as some optimists do, the end of the privately owned vehicle, even when the technology becomes accessible to private buyers. We are predicting that the transition will take a long time, on the order of decades. The cost of an autonomous vehicle will fall slowly, and the human-driven cars already on the road will turn over slowly. During that long transition, the infrastructure this paper has been recommending serves the next era as cleanly as it serves the current one.

Consider what a private AV owner actually wants. The naive scenario, the one that shows up in the renderings, has the car drop the owner at work in the morning, drive itself home, sit in the driveway, and return at five to pick the owner up again. The math falls apart on contact with the road. That pattern doubles the number of trips on every block in the city. A private AV that does it adds congestion at exactly the hours the network can least afford it, and a city in which a significant share of commuters do it has gone backwards on every metric this paper cares about. No serious analysis of private AV ownership lands on this model, and no honest one should.

The model that does work is a market for retrieval. A private AV owner who does not want to pay to park downtown does not need the car to drive home; the car needs to wait somewhere closer than home and farther than the office curb, at a price the owner is willing to pay for the retrieval time involved. The willingness to pay sorts itself naturally. An owner who values a fast retrieval pays more for a stall in the same building or one nearby, and the car is at the door within a minute of being summoned. An owner who is willing to wait longer pays less and sends the car to the top floor of a structure several blocks away, or to an edge-of-core garage near a transit station where land is cheaper, accepting a ten- or fifteen-minute retrieval in exchange for a meaningful price difference. The infrastructure the city needs to support that marketplace is, almost exactly, the infrastructure this lever has been describing: a distributed network of adapted garages and mobility-court nodes, priced by location and quality of service, governed by the same dispatch software that runs the fleet during the fleet era. The owners of those facilities are not picking which era they serve. They are serving both.

This is the deepest reason to build the infrastructure now. The fleet era is the immediate opportunity, and it justifies the investment on its own terms. The private AV era is the second act that justifies it again, on a longer time horizon, with a different customer paying the bill. The mobility court and the adapted garage, opened earlier in this lever and extended in the subsection above, have a third customer waiting in the wings: the private AV owner who needs the same building to keep a car between trips. Three customers, one piece of infrastructure, three different shifts of revenue across decades. The asset earns through every era. That is what it means to design for change itself rather than for any single prediction about which technology wins.

Designing the curb for everything else that moves

The household scenario above has a load-bearing assumption hidden inside it: that the family members who give up a car have something good to switch to. Transit is part of that, and the robotaxi is part of it, but a large part is micromobility, which has quietly become serious transportation rather than a weekend toy. Riders took a record 157 million trips on shared bikes and scooters across the United States and Canada in 2023, and those trips increasingly serve daily life. By one analysis of large systems, thirty-nine percent of riders use shared micromobility to run errands, thirty-four percent to get to work, and sixteen percent to reach school.43 These are not joyrides. They are the lunch run, the grocery trip, and the commute’s last mile.

That observation is the seed of a concept we will develop fully in a separate paper, and introduce here only because it is inseparable from the redesigned curb: the dedicated micromobility lane. We mean something different from the conventional painted bike lane, which has too often become a political flashpoint, resented by drivers as space taken from cars and used, in practice, as much for recreation as for getting anywhere. We mean purpose-built lanes for people who are actually going somewhere: commuters on bikes, e-bikes, scooters, and electric wheelchairs, moving with a speed and reliability the sidewalk cannot offer and the traffic lane will not permit. Designed well, such a lane is also the natural last-mile connector for the household that parks once. A commuter from beyond the core can leave a car at a mobility hub on the city’s edge, an idea rooted in the older logic of park-and-ride, and then move through the city all day by micromobility, for meetings and errands and lunch, returning to the car only in the evening, rather than repositioning two tons of metal for every stop.44

There is a second, less obvious tenant for that lane. A fleet of sidewalk delivery robots is emerging fast and, at the moment, colonizing the sidewalk. Serve Robotics has built some two thousand machines, Coco speaks of putting ten thousand on the streets, and the category leader has passed eight million deliveries across more than a hundred service areas. They block pedestrians, trouble wheelchair users, roll unpredictably through crosswalks, and occasionally do real damage, as when one recently plowed through a Los Angeles bus shelter; the national association of city transportation officials has warned that uncoordinated deployment could flood sidewalks and that the devices should be tightly restricted.45 The lesson cities learned too late with ride-hailing, that reactive policy invites congestion and conflict, applies here with time still on the clock. A micromobility lane gives these robots a place to be that is neither the pedestrian’s sidewalk nor the car’s travel lane, and converts a looming nuisance into a managed delivery network. Get this right, and the same strip of pavement carries commuters, shoppers, and parcels, and the community is better for all three. We take up the design in full elsewhere; the point for this lever is that the off-street hub and the redesigned curb must be planned together, with the lane in the drawing from the start.

Lay the groundwork now

The thread tying all of this together is timing. Almost everything in this lever is cheaper, easier, and less contentious to arrange before the fleets arrive at scale than after. Concretely, a city should do four things. Zone for fleet staging, charging, and service hubs, sited to minimize empty mileage and to spare residential streets the noise and light that depots generate, and zone equally for the distributed staging that lets ordinary mobility courts host the fleet at off-hours. Treat its off-street parking stock as convertible infrastructure rather than a fixed liability, encouraging owners to adapt structures into fleet and mobility hubs as private-car demand recedes, a theme we carry into the next lever on buildings and zoning. Decline to subsidize peripheral fleet parking, and instead price the right-of-way so operators internalize the cost of resting their vehicles. And reserve, in every curb redesign, the space for the micromobility and delivery lane before the sidewalk is overrun. Each of these is a decision a city can make now, with the leverage that comes from acting before a private operator’s sunk investment quietly sets the terms.

A fleet at rest is still a fleet in the city, and where it rests is a choice a city makes or forfeits. Planned well, the idle robotaxi waits in an adapted garage that earns its keep charging and cleaning vehicles off the public street, or in a grocery store’s mobility court at midnight, or at a multifamily building’s pick-up bay between fares, and the deeper prize comes into reach: a city where a family of four genuinely needs one car instead of three, and where the space no longer spent storing the other two becomes housing, commerce, or simply room to move. Planned badly, or not at all, the fleet fills the cheapest space it can find, which is the street, while the surplus parking sits empty and stranded and the curb clogs. The difference is groundwork, and the time to lay it is now. The next lever turns to the buildings themselves, and to the zoning and design rules that will decide whether the off-street opportunity described here is allowed to happen.

Section F · Lever 3Where Every Trip Ends

The first lever priced the curb. The second found the fleet a place to sleep and gave the parking garage a second life. Both rest on something most cities have not yet built: a code that tells a robotaxi where it is allowed to stop. Every trip ends somewhere, at a door, a lobby, a loading bay, and the rules that govern that arrival, the zoning ordinance, the parking requirement, the building standard, are the quietest and most powerful lever a city holds. They were written for a world of private cars that arrive once and park all day. They are nearly silent on a world of vehicles that pull up, let a passenger out, and leave again. Rewriting them is the work of this lever, and forward-thinking zoning is the key. Encouragingly, the rewrite has already begun.

From parking minimums to passenger zones

For the better part of a century, the central command of the zoning code was the parking minimum: so many spaces per apartment, per thousand square feet of office, per restaurant seat. The requirement was rarely studied and almost never questioned, and it quietly shaped the American city, mandating asphalt, inflating the cost of housing, and pushing buildings apart to make room for cars. The economics of the parking minimum were laid bare years ago, and cities have finally started to act. Buffalo became the first major city to abolish minimums in 2017; Minneapolis did so citywide in 2021; Austin became the largest American city to follow in 2023; California eliminated them near transit statewide that same year; and in 2025 Chicago and Denver ended mandates near transit. More than two hundred cities have now reformed or are reforming their requirements, and the results point in one direction: more housing, built for less, on land that no longer has to be paved for cars that may never come.46

Repeal, though, is only half the rewrite. Removing the obligation to build parking does not by itself tell the robotaxi where to stop, and a curb left entirely unzoned is its own kind of failure. The other half is to replace the old command with one fitted to the new trip: the passenger loading zone. Chandler, Arizona, did precisely this in 2018, when it became the first city in the country to rewrite its zoning code in anticipation of autonomous vehicles. The ordinance lets a developer cut required parking by up to forty percent in exchange for designated pick-up and drop-off zones, ten percent for each, and sets standards for them: roughly fifty feet from the entrance, shaded, furnished with seating, and accessible. The logic Chandler’s planners named is the logic of this entire paper: automation lowers the demand for parking and raises the demand for a place to be dropped off. So trade one for the other, in the code, on purpose.47

Chandler is another cautionary tale, and it is better sounded here than buried. The 2018 reform made a bet — that automation would arrive fast enough to make the traded-away parking unnecessary — and it did not arrive that fast. Six years later, in 2024, the city revisited the framework and amended it (Ordinance 5075), for a reason that anchors a companion study on Chandler: the city’s own planning record shows the parking a developer was permitted to shed did not vanish so much as relocate, landing on the surrounding residential streets, the same displacement the equity lever will name and the curb lever’s residential case turns on. The reform was not wrong; it was early, and being early it learned in public what happens when the supply side of the code outruns the demand it is betting on — the shortfall finds the nearest residential curb. We take Chandler apart in full in that companion, because the pioneer’s second lesson is as useful as its first.48

A different curb for every use

The temptation, having discovered the passenger loading zone, is to write a single standard for it and stamp it across the whole city. That would be a mistake. A drop-off is not a generic event; it is shaped by whatever lies behind the door. Consider one ordinary afternoon trip, from the office to the grocery store to home, and watch how the right answer changes at each stop.

At the office, in the dense urban core, the robotaxi should not stop on the public street at all. The volume is high, the sidewalks are full, and no one is carrying anything heavier than a laptop. The drop-off belongs off-street, inside a parking structure repurposed as an arrival hall, which is exactly the relief the curb lever was straining to provide. At the grocery store, the trip ends with bags, and the need inverts: the pick-up and drop-off belongs in the surface lot at the store entrance, with room to idle while a cart is unloaded and a trunk is filled. A fifty-foot strip sized for an office tower is not a grocery loading zone. At the apartment building, the need shifts once more. Residents arrive at every hour with groceries, luggage, strollers, and the parcels and delivery robots of the previous lever, and the building needs its own dedicated bay, on-site and off the public street wherever possible, sized for dwell rather than for a quick handoff. Three destinations, three rules, one trip.

Each of those three rules implies a different physical design at the building edge: the in-structure arrival for the dense office core requires a different floor plate, clear height, and circulation geometry than the surface-level grocery bay or the residential pick-up zone sized for dwell rather than throughput. The zoning code can require each use type to provide the appropriate configuration; specifying what that configuration should be, its dimensions, its relationship to the sidewalk, its accessibility standards, its charging provision is a design problem that sits downstream of the policy this lever establishes — and an opportunity for dialogue with the design community.

This is where zoning earns its keep. The general problem, that the curb is a finite resource contested by transit, loading, dining, parking, and passenger pick-up all at once, has been mapped in detail by the national association of city transportation officials, whose Blueprint for Autonomous Urbanism also names the failure mode plainly: where the curb is left unmanaged, operators simply compete for it, and it clutters.49 The remedy is not to ban the drop-off but to assign it its place by land use, before the clutter arrives, and increasingly to manage it dynamically, with pricing and real-time data that let a single stretch of curb be a bus zone at rush hour, a loading zone at midday, and a pick-up zone at night.50

Buildings that change their minds

Two threads from the previous lever come due here, and the popular answer to both is the wrong one. As private-car parking demand softens and the off-street facility takes on new work as a fleet and mobility hub, the reflex of the future-minded developer is to reach for adaptive reuse: pour flat floors and tall ceilings so that, thirty years from now, the garage might be converted into apartments. It is a tidy idea, and mostly a distraction. Designing a new structure today around a speculative conversion three decades out is both costly and shortsighted, and it answers a question almost no operator is actually asking. The pressing question is not what the garage becomes in 2055. It is whether the garage built in 2026 still works, and still earns, in 2034, when the electric and autonomous transition this paper describes is well underway. Those are different problems with different solutions, and conflating them is how a brand-new facility ends up functionally obsolete inside a decade.

Reframed correctly, future-proofing is about operational and technological readiness over the next five to ten years, not adaptive reuse over thirty. We are not predicting the end of parking. We are describing its evolution, from a static container for cars into a dynamic mobility hub, and the design task is to build in the optionality to move with a transition whose direction is clear but whose timing is not. The hedge against an uncertain future is not a clever prediction; it is flexibility, and a handful of decisions, cheap at construction and expensive to retrofit, preserve it. A digital backbone of conduit, fiber, and electrical capacity sized for the ten-year load with real headroom, because adding it later costs several times what building it in costs. A long-span, modular frame that thins the interior columns, which yields the flexible floor plates a changing use needs and, just as important, gives autonomous vehicles the clear, gentle circulation their sensors require. Generous sightlines and clear heights where cameras and sensors will live, because license-plate recognition today and vehicle-to-everything systems tomorrow are only as good as what they can see. A ground floor designed not as the cheapest parking but as a mobility node, flat and high-clearance, configured for ride-hail staging, last-mile delivery, fleet charging, and micromobility, which pulls disruptive traffic off the public curb and earns more per space than a row of self-park stalls. Conduit and electrical provision for a future in which thirty to fifty percent of spaces may need to charge, without betting the design on any one adoption curve or any single season’s mandate. And wayfinding that can address machines as well as people, so the building is not forced into a second, costly retrofit the day AV operations become routine.

None of this assumes the dramatic version of the future. The near-term reality is a mixed fleet, human-driven cars and autonomous ones sharing the same ramps, and a facility that can serve both without compromising either. Nor is it only a new-construction problem. Most of the parking that will be operating in 2035 already stands today, so for the existing stock the work is a disciplined retrofit: audit the foundational systems first, the electrical service, the conduit, the network backbone, and the access-and-revenue platform, then phase the upgrades so that foundation precedes operations and operations precede revenue, rather than bolting on visible features the structure cannot actually support. Throughout, the platform choices should favor open standards and software-as-a-service over the proprietary, walled-garden systems that have stranded a generation of garages.51

The zoning implication follows directly, and it is the opposite of the convert-it-to-condos reflex. A forward-thinking code should not freeze today’s parking footprint, but neither should it lean on speculative residential conversion as its theory of the future. It should permit and reward the evolution this lever actually describes: let a parking structure operate as a charging hub, a fleet depot, a delivery node, and an intermodal mobility hub; clear the regulatory friction that keeps a garage classified as nothing but car storage; and reward, in new structured parking, the readiness investments, the digital backbone, the structural clearance, the sightlines, the charging capacity, that are trivial to build in and ruinous to add later. The building does not need to be designed to stop being a garage. It needs to be designed to become a far more useful one.

The design decisions that accomplish this vary significantly by building typology. For a standalone structured garage, the critical choices are structural: column spacing, floor plate depth, ramp configuration, made at the moment of construction and expensive to revisit. For a podium or integrated parking structure within a mixed-use building, the critical choices are mechanical and electrical: service capacity, conduit routing, transformer sizing, utility interconnection, embedded in the building’s base systems and equally expensive to retrofit. For an existing structure being adapted rather than built new, the audit sequence and the upgrade phasing determine whether the conversion is viable at all. Each of these typologies requires a different design strategy, and the design guidance for each belongs squarely in the design-community dialogue.

The stadium that works on a Tuesday

The hardest arrival problem in any city is the event venue. A stadium empties tens of thousands of people into the surrounding streets in roughly half an hour, by every mode at once, and the choreography of that crush, mass pick-up and drop-off, a flood of bikes and scooters, parking filled to capacity, is a discipline unto itself. Venues are, in the planner’s term, destination hubs, alongside airports and malls, and event-day operations already lean on designated rideshare zones and special-event transit.52 But here is the part the old code misses: the stadium’s vast parking sits empty most of the week. The shared-parking insight is not new, yet the autonomous era sharpens its payoff. A venue designed for the technology, with the curb geometry to absorb a micromobility surge and the bays and chargers to serve an AV fleet, can do far more than host Sunday’s game. On an ordinary Tuesday it becomes the neighborhood’s commuter parking, the fleet’s staging and charging ground between peaks, and a mobility hub where a commuter parks once at the edge of the core and takes a scooter or an e-bike the last half-mile to the office. After dark, the same bays revert to the overnight charging and cleaning role described in Lever 1 (Reinventing the Curb) — a fourth shift of revenue from operators whose alternative is breaking ground on a depot they do not have entitlements for yet. The same structure serves the Sunday crowd, the Tuesday commuter, and the midnight fleet, but only if it is zoned and designed for micromobility and AV taxis from the start, rather than for the tailgate alone. The stadium is the most visible version of a pattern Lever 2 (Where the Cars Sleep) generalizes, in which every well-designed mobility court doubles as a node in a distributed fleet-staging network.

Write the code now

The connecting thread is that the code is the lever, and a rule is far cheaper to write than a building is to retrofit. A city serious about the autonomous transition should do five things in its zoning and building ordinances. Replace blanket parking minimums with use-specific passenger-loading standards, so the requirement matches the trip rather than a formula from the 1960s. Zone pick-up and drop-off by land use: off-street and in-structure for the dense-core office, generous surface bays at the grocery and big-box entrance, dedicated on-site bays for multi-family housing, and choreographed multi-modal access for venues. Reward evolve-ready design in new structured parking, the digital backbone, structural clearance, sightlines, and charging capacity that let a facility take on charging, fleet staging, and mobility-hub roles as the transition unfolds, and that are cheap to build in and ruinous to retrofit. Zone explicitly for shared and dual use, letting stadiums, garages, and lots serve event crowds, weekday commuters, fleet staging, and micromobility in turn, and permitting ordinary mobility courts to host distributed AV staging at the hours they are not otherwise busy. And manage the curb dynamically, with the pricing and data of the first lever, so a single stretch of it can change its job by the hour. None of this requires a technology the city does not already have. Each requires only the will to write the rule before the concrete is poured.

A robotaxi will stop wherever a city allows it to. Write nothing, and it stops in the travel lane, at the bus stop, across the crosswalk, wherever the software finds a gap, and the worst version of the autonomous city arrives not by malice but by default. Write the code with foresight, and every trip ends where it should: the office arrival handled inside the garage, the grocery run in a bay built for bags, the ride home at the building’s own door, and the same well-drawn structure serving the Sunday crowd and the Tuesday commuter alike. The buildings, for the most part, do not need to be reinvented. The rules do.

And the rules, once written, produce a city that still has to be designed, at the building edge, at the curb, and at every node in between. The architect specifying the ground floor of the next parking structure, the urban designer drawing the mobility court into the streetscape, the parking consultant advising on a residential curb reclamation sequence, each is writing a version of the code in the language of dimensions, materials, and phasing rather than ordinances and operating agreements. That design work is improved with the approach described in this policy framework. The next lever turns from the physical city to the people moving through it, and to the question of whether the city this code builds will serve everyone, or only those who can afford the fare.

Section G · Lever 4Who Gets to Ride

The first three levers built a city that moves. Priced curbs, fleet depots, evolve-ready garages, passenger zones written into the code: each one assumes a rider who can summon a car, pay the fare, and step into a vehicle that arrives. That assumption is doing quiet work. It is the difference between a transportation system and a transportation system for some people. Every lever so far has been about the machine. This one is about who the machine leaves at the curb, and whether the city writing this code decides that question on purpose or by neglect.

The honest version of the autonomous city is not the one in the renderings, where a clean white vehicle glides to a sunlit corner and a young professional steps in. It is the one that has to answer harder questions. What does the ride cost the person who has no other way to get to work. Whether it comes at all to the neighborhood the fleet has decided is unprofitable. Whether the door opens for a passenger who uses a wheelchair. Whether the worker who drove the last decade of ride-hail trips has anything to do when the driving stops. These are not soft questions to be appended after the engineering is done. They are the engineering, and the same code that zones a curb can zone an injustice just as easily.

The fare and the territory

Lever 1, Reinventing the Curb, made the case for pricing the curb, and the case holds: a resource that everyone wants and no one is charged for will be wasted. But pricing has a shadow, and it is time to name it. When the curb is metered, the trip is metered, and the ride itself is metered, the city becomes a place you pay to move through, and the people with the least money make the most of their decisions at that margin. A pricing regime that is efficient in the aggregate can still fall hardest on the household that has no slack in its budget to absorb it.

The evidence here is not speculative. In a congestion-pricing experiment in the Seattle region, the lowest-income drivers paid an average of roughly seven percent of their weekly income in charges, several times the share borne by wealthier drivers, who could more easily shift their trips or simply absorb the cost.53 The lesson is not that pricing is wrong. It is that pricing without a rebate is regressive by default, and that the fix is already known: return a meaningful share of the revenue to the people the charge falls hardest on, through credits, discounts, or direct transfers tied to income. The curb revenue this paper has been spending on redesign has a second claim on it, and it is this one.

The same regressive pattern shows up in parking, not just in tolls, and the modeling is explicit about where the burden lands. In an agent-based simulation of shared autonomous vehicles in Atlanta, Zhang and Guhathakurta found that once downtown parking is priced, parking demand does not simply shrink; it relocates from the priced core to adjacent low-income neighborhoods, which inherit the stored vehicles the center has pushed out.54 A later study of the same mechanism projects that residential-area parking demand could roughly double as the core sheds it. Pricing the curb without managing where the displaced demand goes therefore does more than fall hardest on low-income drivers at the meter; it can physically move the parking burden onto low-income blocks. That is an argument for pairing curb pricing with the territorial conditions below, and for directing some of the revenue to the neighborhoods that absorb the spillover.

Territory is the other half of the problem, and automation sharpens it. A human driver, for all the friction of the old system, would eventually take the trip to the far neighborhood because the driver lived somewhere and had to get home. A fleet has no home. It stages where the demand is densest and the trips are most profitable, and it can decline an entire district as a matter of routing logic, with no driver to overrule it. The result is a map that quietly sorts itself into served and underserved, not by anyone’s stated intention but by the optimization running underneath. If the city wants the fleet to serve the whole city, that has to be a condition of operating in the profitable part of it, not a hope.

The door that does not open

Here the paper meets the one place where its governing principle does not hold, and we would rather say so plainly than paper over it.

Everywhere else in these pages, the Parking Consultants Council position has been to reward desired behavior rather than mandate it. Incentives are the right tool when you want more of something the market will supply once the economics tilt: evolve-ready garages, passenger zones, dual-use venues. Give the market a reason and it will build them. Accessibility is different in kind, not degree. It is not a behavior to be encouraged. It is a floor beneath which service is not service, and floors are precisely the thing markets underbuild, because the economics of the marginal accessible vehicle never close on their own.

The current autonomous fleets make the point. The largest robotaxi operator in the country runs vehicles, the Jaguar I-PACE and newer models, that are not wheelchair accessible, and where it offers a wheelchair-accessible option at all, those rides are provided by human drivers in separate vehicles dispatched through a contractor, not by the autonomous fleet itself.55 The autonomous service, in other words, is the part that does not open its door to a wheelchair user. Worse, the operators have sought the same state-level preemption that ride-hail companies won, which stripped local authorities of the power to require accessible service.56 The pattern is clear: left to incentives and preemption, the driverless fleet will route around the rider who is most expensive to serve.

There is a working model for the alternative, and it is not theoretical. California funds accessible on-demand service through a flat fee of ten cents on every ride-hail trip that begins in the state, pooled and directed to wheelchair-accessible vehicle service through its Access for All program. Since 2019 the program has put roughly ninety million dollars into accessible service and supported on the order of a quarter-million wheelchair-accessible trips across twenty-two counties.57 It is a mandate dressed as a fee, and it works precisely because it does not pretend the market will volunteer. A city or state serious about the autonomous transition should write the same floor into the code that licenses the fleet: a per-trip access fee, a minimum share of the fleet that is wheelchair accessible, and a service standard that the autonomous vehicles themselves, not a separate contractor’s vans, are expected to meet over time.

This is not a retreat from the market principle. It is its boundary. We reward almost everything and mandate only the floor, and accessibility is the floor.

The driver at the end of the shift

The autonomous fleet does not only change who rides. It changes who works, and the paper would be dishonest to plan the curb and ignore the person standing on it.

Ride-hail created a large, flexible, low-barrier source of income, work that required no degree and no prior experience and that a great many people used as a bridge, a supplement, or a sole support. Automation removes the driver from exactly that trip. The displacement will not be instant, and the honest forecasts do not claim it will be: one widely cited projection has ride-hail trips in the United States growing to roughly fifteen billion a year by 2040, with about half driven autonomously and half still by people.58 Even the chief executive of the largest ride-hail platform has called the eventual displacement of drivers “a big, big societal question,” placing it on a ten-to-fifteen-year horizon rather than denying it.59 A transition measured in years rather than months is a gift, because it is time to plan, and the worst response would be to waste it pretending the question is not there.

Planning for it does not belong entirely to the parking and curb code, and we will not pretend a zoning ordinance can retrain a workforce. But the mobility hub this paper has been designing is, among other things, a place where new work concentrates: fleet charging and cleaning, vehicle staging and maintenance, remote operation and customer support, the micromobility and delivery operations that share the building. The distributed staging model of Lever 2 (Where the Cars Sleep) multiplies those points of work across every grocery, multifamily, and venue that hosts a court, and concentrates them in places that already employ a workforce. A city that wants the autonomous transition to add jobs as well as subtract them can use the same hubs and the same incentives to favor operators who hire and train locally, and can treat the displaced ride-hail workforce as the first candidates for the roles the new system creates. The point is narrow and it is the recurring point of this paper: the outcome is a choice, and the time to make it is before the fleet arrives, not after the drivers are gone.

Who owns the map

The autonomous city runs on data. Every priced curb, every dynamic zone, every dispatch decision is a record of where people go and when, and that record is valuable, both as a tool for running the system and as an asset someone will want to own. Two risks follow, and they are quieter than the others because they do not show up at any single curb.

The first is that dynamic pricing, left to optimize for revenue or throughput alone, will rediscover the regressive pattern in a new form, charging the most at exactly the times and places that the people with the least flexibility have to travel. The remedy is the same as the fare remedy above, applied to the algorithm: equity has to be an explicit objective in how curb and congestion prices are set, not an afterthought to be patched once the burden is visible. The second is ownership. If the movement data the public’s streets generate is captured entirely by private operators behind proprietary, walled-garden systems, the city loses both the ability to govern its own curb and the leverage to hold the fleet to the standards above. The same preference this paper urged in the zoning lever, open standards over proprietary lock-in, is an equity position as much as a technical one: the city that can see its own data is the city that can govern who its system serves.

Write the code now

The connecting thread of every lever has held: the code is the lever, and a rule is far cheaper to write than an injustice is to undo. A city serious about an autonomous transition that serves everyone, and not only those who can afford the fare, should do five things in the ordinances and operating agreements that license the fleet.

Return the revenue. Pair every curb and congestion charge with an income-based rebate or credit, funded from the charge itself, so that the pricing that disciplines demand does not fall hardest on the households with the least slack.

Condition the territory. Make service to the whole city, including the neighborhoods a fleet would otherwise skip, a condition of operating in the profitable core, rather than a courtesy the city hopes for.

Mandate the floor. Write accessibility into the licensing code directly, with a per-trip access fee on the model of California’s Access for All, a minimum accessible share of the autonomous fleet, and a standard that the autonomous vehicles themselves meet the need over time, not a separate tier of contracted vans.

Plan the transition. Use the mobility hubs and the incentives already in this paper, including the distributed staging nodes of Lever 2 (Where the Cars Sleep), to favor operators who hire and train locally, and treat the displaced ride-hail workforce as the first candidates for the work the new system creates.

Keep the map public. Require open standards and city access to the movement data generated on public streets, and make equity an explicit objective in how dynamic prices are set, so the algorithm cannot quietly become regressive.

None of this requires a technology the city does not already have. Each requires only the will to write the rule before the fleet writes it for them. A robotaxi will serve whomever the code tells it to serve, and decline whomever the code lets it decline. Write nothing, and the autonomous city will sort its residents into those it finds worth the trip and those it does not, not out of malice but out of optimization. Write the code with intent, and the same system that prices the curb and reshapes the garage can be the one that finally widens the door. The next lever turns from who gets to ride to what they ride toward, and to the network that ties the whole city together: transit.

Section H · Lever 5What They Ride Toward

The last lever asked who gets to ride. This one asks what they ride toward. A trip is not just an origin and a fare; it is a movement through a network, and the most important question about the autonomous vehicle is not how it behaves at the curb but what it does to the network that has carried cities for a century: mass transit. A bus moving forty people down a corridor and a robotaxi moving one are not doing the same job at different scales. They are doing different jobs, and a city that confuses the two will buy a great deal of the wrong one.

This is the lever where the paper’s optimism has to earn itself. It would be easy, and wrong, to present the robotaxi as the natural heir to the bus, the upgrade that finally makes transit obsolete. It would be just as easy, and just as wrong, to cast it as the bus’s assassin. The truth is that the autonomous fleet is neither, on its own. It is a powerful instrument that will do whatever the surrounding network is built to make it do, and the deciding factor, once again, is not the technology but the code and the design around it. Build nothing, and the fleet competes with transit. Build the connections, and it feeds it.

Two networks, one trip

Start with what each mode is actually good at, because the whole argument rests on the difference. High-capacity transit, a subway, a light-rail line, a bus on a dedicated lane, moves enormous numbers of people along a fixed corridor at a cost per rider that nothing else approaches. Its weakness is the same as its strength: it runs where the rails and the route run, and most people do not begin or end their day at a station. The autonomous fleet is the mirror image. It is hopeless as a way to move forty thousand people down one corridor at rush hour, the geometry simply does not work, but it is unmatched at the thing transit does worst, which is the flexible, low-density, door-to-anywhere trip, including the half-mile between the front door and the platform.

That half-mile has a name in the planning literature, the first-mile and last-mile problem, and it is the single largest leak in the transit bucket. A train that is a fifteen-minute walk from home in the rain is a train most people will drive past. Solve the connection and the corridor fills; leave it unsolved and the corridor empties, not because the train is slow but because reaching it is a chore. This is precisely the gap the autonomous fleet is built to close, and it is why the honest framing of robotaxi and transit is not competitor but feeder: the fleet at the edges, the trunk line down the middle, each doing the job it is actually good at.

The mobility hub from the earlier levers is where the two networks physically meet. We have described it at length already, so we will not belabor it here; the point for this lever is narrower. The hub is the transfer, the place where a rider arrives by autonomous vehicle, scooter, or bike and steps onto high-capacity transit without crossing a parking lot or waiting in the weather. A hub that is not anchored to a transit line is a nicer curb. A hub that is anchored to one is the joint that holds the whole network together.

The fork in the road

None of this is automatic, and the same fleet can pull the network in either direction depending on what a city does. This is the synthesis at the center of the lever, and it deserves to be stated plainly.

Left to optimize for its own revenue alone, the fleet has every incentive to compete with transit rather than feed it. The dense, profitable corridor at rush hour, the very trip the subway serves best, is also the trip a robotaxi operator would most like to capture, and a fleet that skims those riders off the trunk line does real damage: it adds vehicles to the most congested streets at the most congested hour, raises vehicle miles traveled, and pulls fare revenue out of the system that carries everyone else, the classic warning of the analysts who study this. The corridor that loses its peak riders cannot cross-subsidize its coverage routes, and the network frays from the profitable center outward.

The other fork is already being built, and it is not hypothetical. In September 2025, Waymo and Via announced a partnership to fold autonomous vehicles directly into public-transit networks, with the first deployment in Chandler, Arizona, integrating Waymo’s service into the city’s existing Chandler Flex microtransit, where Via’s routing software treats the autonomous vehicles as part of the public fleet, dispatched to aggregate riders and connect them to the wider system.60 It is worth pausing on where that first deployment landed: Chandler, the same city whose zoning rewrite anchored the buildings lever, is now also the first to make an autonomous vehicle an official part of public transit. The same city that wrote the code early for the curb is writing it early for the network. That is not a coincidence; it is what writing the code early looks like.

The Chandler model is one shape the integration can take. There are others already in the field: San Francisco has run a program letting transit riders earn credit toward autonomous trips, and well before any of this, transit agencies used federally funded pilots, the Dallas-area first-mile and last-mile demonstration among them, to subsidize on-demand rides that delivered people to fixed-route transit rather than around it.61 The tools differ, but the principle is constant: the autonomous trip is structured to end at the trunk line, not to replace it.

What the network needs from the code

If the fleet will do whatever the network is built to make it do, then the city’s job is to build the network so that feeding transit is the path of least resistance, for the operator as much as the rider. The recurring discipline of this paper applies here with particular force, because the integration is cheap to design in at the start and ruinously hard to retrofit once a fleet has learned to compete. A city serious about a network that holds together should do four things.

Anchor the hubs to transit. Site and zone the mobility hubs of the earlier levers at transit stations and along trunk corridors, so the natural end of an autonomous trip is a platform, not a parallel road. The hub is the physical incentive to feed rather than compete, and its location is a choice the city makes in the code.

Price the trip as part of a network, not against it. Use the curb and congestion pricing of the first lever to make the short feeder trip to the station cheap and the long peak-hour trip down the transit corridor expensive, so the fleet is steered toward the work transit cannot do and away from the work it does best. A trip that ends at a train should cost less than one that races the train.

Integrate the fare and the trip. Require that autonomous operators serving the city expose the open interfaces that let a rider plan, book, and pay for the feeder ride and the transit ride as one journey, on the model of the agency-integrated services already in the field. A trip the rider has to assemble from two apps and two fares is a trip that defaults back to the car.

Make feeding transit a condition, not a hope. Where a city grants an operator access to its streets and curbs, it can require participation in the transit network, service to the stations and the coverage areas, the way Chandler and others have begun to, rather than leaving the operator free to cherry-pick the corridors that compete. The permit is the leverage, and the time to use it is before the routes harden.

None of these requires a technology that does not exist; the autonomous fleets are already running, and the integration software is already dispatching them into public systems. Each requires only that the city decide, while it still holds the leverage, that the fleet’s job is to widen the mouth of the transit network rather than to drain it. A robotaxi pointed at a train station strengthens the whole city’s mobility. The same robotaxi pointed at the train’s own riders weakens it. The vehicle does not know the difference. The code does, and the code is still ours to write. The next lever steps back from any single mode to the structure of the market itself, and to the question of who owns the network we have been building, and on what terms.

Section I · Lever 6Who Owns the Network

The last lever followed a single trip to the edge of the transit network. This one steps back from any one trip, any one mode, and asks the question every earlier lever has quietly assumed away: who owns the system that runs all of it, and on what terms. A priced curb, a fleet depot, an evolve-ready garage, a passenger zone, a feeder route to a train: each is a piece of a network, and a network has an owner. The question is whether that owner is the public that paved the streets or a private operator that learned to run them, and that question does not answer itself. It is answered, like all the others, by the code the city writes before the market writes it instead.

This is the lever that decides whether the first five matter. A city can price its curbs perfectly, zone its passenger pick-ups, and stitch its hubs to transit, and still wake up to find that the company dispatching every vehicle on its streets owns the data, the rider, and the map, and treats the city as a tenant on its own pavement. The technology does not produce that outcome. The market structure does, and the market structure is a choice.

The shape the market wants to take

Begin with what the autonomous mobility market is already doing, because the trend is not subtle. In December 2024, General Motors shut down Cruise, which had been one of two companies running commercial robotaxis in American cities, and for a stretch afterward a single firm, Waymo, was effectively the only one operating at scale.62 By the latter half of 2025 that firm was running on the order of 250,000 paid trips a week across five markets with a fleet in the low thousands, and was operating, planning, or testing in roughly two dozen markets at home and abroad. Competitors have since entered, but it is worth naming who they are: not a crowd of independents, but Tesla, which launched its branded service in Austin in 2025, and Amazon, whose Zoox unit builds a purpose-made driverless vehicle with no steering wheel.63 In China the dominant operator is Baidu. The new entrants to the most local of businesses, moving people down a city’s own streets, are among the largest technology companies on earth.

That is not an accident of the moment; it is the gravity of the business. Autonomous mobility rewards scale the way few industries do. The vehicles are capital-intensive, the mapping and the machine learning cost a fortune to build and almost nothing to copy across the next city, and the same handful of firms that can afford the fleets also own the cloud, the maps, and the artificial intelligence underneath them. Markets shaped like that do not tend toward many competitors politely sharing a street. They tend toward one or a few, and they tend to stay that way.

The flywheel

The reason the few become fewer is worth stating plainly, because it is the engine of the whole problem. An autonomous fleet runs on data, and data compounds. Every mile a fleet drives trains the model that drives the next mile; the better the model, the better the service; the better the service, the more riders, and the more riders, the more miles and more data. This is a flywheel, and flywheels reward whoever is already spinning fastest. The leader’s advantage is not a head start that competitors can close; it is a gap that widens with every trip.

For a rider, a dominant operator can mean a smooth and cheap ride, at least until the competition thins. For a city, the flywheel produces something more troubling than a big company. It produces an information asymmetry that hardens into dependence. The operator accumulates a complete, real-time, proprietary picture of how the city moves, every origin, every destination, every curb dwell, every detour, while the city that owns the streets sees only what the operator chooses to show it. We raised this in the equity lever as a question of who owns the map. Here it returns as a question of who owns the market, and the two are the same question wearing different clothes. The data is the moat, the moat is the market power, and the market power is exercised on public ground.

Sovereign on paper

Here is the failure mode, and it is not the cartoon of a wicked monopolist gouging riders, though that can happen too. The quieter and more likely failure is dependence. A city that has let a single operator’s proprietary system become the way its streets actually function is sovereign only on paper. It owns the curb, the lane, and the right-of-way in law, and it cannot audit what happens on them in fact. It cannot verify the operator’s safety claims against data it is not allowed to see. It cannot enforce the priced curb of the first lever, or the passenger zones of the zoning lever, or the transit-feeding conditions of the transit lever, if the only record of compliance lives on the operator’s servers in the operator’s format. And it cannot switch to a competitor or bring in a second operator, because the city’s entire way of managing its streets has been built around one vendor’s walled garden, and walls are expensive to climb back out of.

This is the outcome the paper has been working to prevent from the first page, expressed at the level of the whole system. The constraint was never the technology. It was always whether the city wrote the rules while it still had the leverage to write them. On market structure the leverage is real but perishable: a city negotiating with an operator that wants access to a lucrative new market has enormous power to set terms, and a city dependent on an operator it can no longer live without has almost none. The window is open now. It does not stay open.

The interface, not the fleet

The temptation, having described a monopoly problem, is to reach for a monopoly remedy: if the network tends toward a single owner, let the public be that owner, and run the fleet as a utility. That is not this paper’s recommendation, and not the council’s position. Public ownership of the vehicles answers a market problem with an operating burden cities are poorly suited to carry, and it abandons the discipline that has run through every lever, that the city’s job is to set the terms and let the market build to them, not to become the builder.

The opposite temptation, to trust that competition will sort itself out, we have just shown to be wishful. The flywheel does not self-correct.

There is a third path, and it is not theoretical; cities are already on it. The city does not need to own the fleet. It needs to own the interface. If the data, the digital policy, and the access to the public right-of-way all run through an open, common specification that any operator must speak as a condition of doing business, then the city keeps the one thing that preserves its sovereignty: the ability to see its own streets, to set the rules in a format every operator obeys, and to swap one operator for another, or run several at once, without rebuilding everything around a new vendor. The fleet can be private. The interface must be public.

The working model is the Mobility Data Specification. It was created by the Los Angeles Department of Transportation in 2018 and handed in 2019 to the Open Mobility Foundation, a nonprofit governed by a board of city and county transportation officials, precisely so that no single company or city would own it. It is an open API that lets a public agency require data sharing, express digital policy, and manage the right-of-way for shared vehicles operating on its streets, and it has been adopted by well over a hundred agencies worldwide, from Los Angeles and San Francisco to Seattle, Denver, Detroit, and the New York City Taxi and Limousine Commission.64 It began with scooters and bikes, the first shared fleets to crowd the curb, but its latest version was built to extend to exactly the modes this paper is about: passenger services, including taxis and ride-hail, alongside carshare and delivery robots.65 The robotaxi is the next vehicle to roll onto the public right-of-way at scale. The standard for governing it on the city’s terms already exists.

This is the open-standards thread that has run quietly through the paper, surfacing in the zoning lever and again in the question of who owns the map, pulled now into the open as the central structural choice. Open standards over proprietary lock-in is not a technical preference. It is the difference between a city that governs its streets and a city that rents them.

Write the code now

The recurring discipline of this paper holds at the level of the market as firmly as at the level of the curb: a rule is far cheaper to write before the network hardens than to impose after one operator has become the network. Setting the terms of access to the public right-of-way is not a mandate on what the market must build; it is the city stating the conditions for using something the public already owns, which is the most basic form of the lever this paper has been describing. A city serious about owning its mobility network, rather than renting it back from the firm that captured it, should do four things in the licenses and operating agreements that let an operator onto its streets.

Require the open standard. Make participation in an open data specification, the Mobility Data Specification or its successor, a condition of the license to operate on public streets, so that the city can see, in a format it controls, how its own right-of-way is being used.

Own the interface, not the cars. Keep the data schema, the curb and policy APIs, and the terms of access public and common, so that any operator can connect and none can make the city dependent on a single proprietary system. Let the fleets compete; do not let the platform enclose.

Refuse the exclusive deal. Do not grant any one operator an exclusive franchise on the public right-of-way, however attractive the offer. Structure access so that multiple operators, and the city’s own transit, can interoperate, because the moment one company is the only way the streets work, the city has traded its leverage away.

Treat the movement data as a public asset. The record of how a city moves, generated on public streets, should be available to the public agency that owns those streets for planning and enforcement, with rider privacy protected, rather than captured solely as a private moat.

None of this requires the city to build a fleet, write an algorithm, or own a vehicle. It requires only that the city insist, while it still holds the leverage, on keeping the keys to its own streets. The robotaxi will run on whatever terms the city sets at the moment it grants access, and on no better terms afterward. Set them well, and the autonomous network becomes infrastructure the city governs in the public interest, the way it governs the roads themselves. Set nothing, and the city will hand the operating system of its own streets to whichever firm spins the flywheel fastest, and spend the next generation asking permission to see what is happening on its own pavement.

That is the last of the levers, and it returns us to where the paper began. None of the six is a forecast, and none waits on a breakthrough. Each is a choice already available to any city willing to make it, and each comes down to the same proposition: the constraint was never the technology, it was the imagination and the will to write the rules in time. What remains is to put the levers in order, to say what a city should do first and what can follow, and to answer the question the whole paper has been building toward, which is not whether the autonomous city is coming, but who will steer it. That is the work of the final section.

Section 10A Sequence, Not a Wish List

Six levers are a lot to hand a city all at once, and a paper that ended by demanding all of them tomorrow would be no more useful than one that demanded nothing. The levers are not a menu to pick from according to taste, and they are not a single bill to be passed in one session. They have an order, and the order is not the one in which we presented them. We arranged the levers to build an argument. A city should arrange them to do the most good with the leverage it has, in the time it has, and that is a different sequence.

Two facts set the order. The first is that some of these moves pay off today, with the streets exactly as they are, no robotaxi required, so a city can act now and capture the benefit whether the autonomous fleet arrives in three years or ten. The second is that some of these rules grow more expensive to write with every month a city waits, because the concrete sets, the buildings rise, and the operators consolidate, and a rule that costs nothing to write on a blank page costs a fortune to impose on a finished one. The sequence follows from those two facts. It is a matter of cost and leverage, not of the technology’s timetable, which no one can predict anyway.

Start with the curb. The first lever, pricing and reclaiming the curb, is the foundation, and it should be the first dollar and the first ordinance. It needs no autonomous vehicle to justify it: the curb is congested, underpriced, and mismanaged today, and pricing it disciplines the traffic a city already has. It also funds everything downstream, the redesign, the rebates, the hubs, from revenue the city is currently leaving on the table. A city that does only one thing from this paper should price its curb, because every other lever spends what the curb earns, and assumes the principle the curb establishes: that the most valuable few feet in the city are a public asset, priced like one. One distinction inside that first move is worth carrying forward: pricing the curb needs no one’s permission, but reclaiming it does. A block can be repriced by ordinance; pulling the parking off a residential street has to be earned, with the residents in the room and a real alternative in place before the spaces come out. On the blocks where people live, the curb is reclaimed at the speed the neighborhood can absorb, and not faster.

Some readers will object that cities have no parking revenue to spare, that they depend on it, and that repricing the curb threatens a stream the budget already counts on. The fiscal objection deserves its own answer, and it gets one in the companion analysis, The Parking Revenue Cliff That Isn’t One: Who Pays for the Curb After the Meter in the Age of Autonomy (PCC-WP-2026-02, in preparation), which works through the public-finance mechanics in detail: why the revenue cities fear losing is smaller and more volatile than they assume, and how the same levers described here read as a replacement stack rather than a funeral.

Write the rules that get more expensive every day you wait. Two levers share the unforgiving property that they are nearly free to write now and ruinous to retrofit later: the zoning of where every trip ends, and the market structure of who owns the network. Stop mandating parking and start rewarding evolve-ready buildings and passenger zones before the next construction cycle pours its foundations, because a garage built wrong is wrong for fifty years. Require the open data standard and refuse the exclusive franchise before a single operator becomes the network, because leverage a city gives away at the first contract does not come back at the second. These are the rules to write while the leverage is still in the city’s hands, which is to say now, while the fleet still wants something the city controls.

Steer the fleet as it scales. Two more levers are responsive to deployment, and they mature as the technology does: where the cars sleep, and what they ride toward. As the fleet grows from a pilot to a presence, a city decides where the vehicles stage and charge, both in adapted central hubs and across the distributed network of mobility courts the previous levers describe, so they do not idle on the curb it just priced, and it structures the integration so the fleet feeds transit instead of bleeding it. The moves themselves scale with the rollout, but the conditions that govern them belong in the first operating agreement a city signs, not the fifth, because the agreement is the leverage and the first one sets the template for all that follow.

Thread equity through all of it. The equity lever is not the last step in the sequence. It is the test applied to every other step. The rebate that rides with the curb price, the accessibility floor that is this paper’s one honest mandate, the service-territory condition on the fleet, the public claim on the movement data: each of these lives inside another lever, not in a phase of its own after the rest are built. A city that prices the curb and waits to think about who can afford it, or designs the network and waits to ask whom it serves, will find that equity does not bolt on after the fact. It has to be machined in from the first cut.

None of this waits on a breakthrough, and none of it requires a tool the city does not already hold. The only variable is whether a city acts while it still sets the terms, or waits until the terms are set for it. Which leaves one question, the one this paper has been circling since its title page.

Section 11Who Steers

The subtitle of this paper makes a claim and implicitly asks a question. The claim is that robotaxis are already reshaping our streets, and by now that is not in dispute: the vehicles are running, by the hundreds of thousands of trips a week, in a growing list of cities, owned by some of the largest companies on earth. The question is who steers the policy, and that remains genuinely open. It is the only thing about the autonomous city that still is.

For a decade the debate about driverless vehicles was a debate about arrival. Would the technology work, and when. That debate is over, settled not by argument but by the cars now turning onto real streets, and it has left a great many cities still bracing for a question that has already been answered while the one that matters goes unasked. The useful question was never whether the autonomous vehicle is coming. It is who writes the rules it runs on. The answer resides in Who Owns the Network, and here the cities have been told a story about themselves that is worse than wrong, it is disarming: the story that the autonomous future is something that happens to them, a wave to be braced for, a disruption to be survived, a thing done to the city by companies and engineers somewhere else. That story is the single most expensive idea in municipal government today, because a city that believes it will wait, and a city that waits will get the version of the future that the people not waiting decided to build.

The argument of this paper is the opposite, and it is simple. The autonomous city is not something that happens to a city. It is something a city writes, in the most ordinary instruments it already owns: the curb regulation, the zoning code, the operating agreement, the price of a parking space, the conditions on a permit. These are not glamorous tools. They are the tools that have always determined how a city moves, quietly, while the attention went to the vehicles. The robotaxi did not change that. It raised the stakes of it. The same code that has always shaped the street will now shape a street that runs itself, and it will do so faster, more completely, and with far less friction than before. Technology is an amplifier. It will take whatever rules it finds and enforce them at scale, the good ones and the foolish ones alike, which is exactly why the rules are worth getting right before the amplifier is fully on.

That is what it means to reinvent gridlock rather than to automate it. Gridlock was never only about cars; it was about a century of rules that priced the curb at zero, mandated parking nobody needed, walled the data, and let the street be governed by default. A city that ports those rules into the autonomous era unchanged will not escape gridlock. It will get gridlock that runs more efficiently, congestion optimized to the decimal, inequity executed by algorithm, a worse outcome reached more smoothly. The promise of the autonomous transition is not that the technology will fix the old mistakes. It is that the transition forces the rules open for the first time in fifty years, and a city willing to rewrite them can build something better than what it had. The window where that is possible is open now, while the fleet still needs what the city controls, and it narrows with every contract signed and every street ceded.

So, who steers. Not the technology, which has no preferences and will serve whatever rules it is given. Not the operators alone, though they are writing their version of the code every day a city declines to write its own, and theirs is a perfectly rational code that happens to optimize for them and not for the people on the sidewalk. The honest answer is that the city steers, if it chooses to, using the lever it has held the whole time and too rarely thought to pull. The constraint was never the technology, and it was never the money, and it was never the timing. It was the imagination to see the curb as a lever and the will to use it before someone else did. That is a constraint a city can lift in a single council session, on a single afternoon, with the tools already on the table.

The cars are coming whether or not the city is ready. What the city decides, in the time it still has the leverage to decide it, is whose city they are coming to. That decision is not made by the engineers, and it is not made by the market, and it is certainly not made by waiting. It is made by the people willing to pick up the lever and steer. We would encourage them to begin.

ApparatusEndnotes

65 notes. Click any reference number in the text to read it in the margin without losing your place.