Daniel wrote to us with something that's been eating at him — and honestly, if you've been in Jerusalem lately, you know exactly what he's talking about. He says the city is total chaos right now. High-rise construction everywhere, and at the same time the municipality is digging up huge sections of the city for light-rail infrastructure and stations. All of it happening at once, seemingly without any coordination. And this has sharpened a conviction he's held for a while — that cities and private car ownership fundamentally cannot coexist. That the only livable city is one built around walking, public transport, and micro-mobility, all three together, not just picking one.
But — and this is where the prompt gets interesting — he says until we reach that point, the best we can hope for in places like Jerusalem that have already been wrecked by decades of bad planning is harm reduction. Symptom control. People need to get to work. So he's got two questions.
First — the history. When cars went from a rich person's toy to something everyone owned, and it happened faster than anyone expected, how did the first generation of traffic planners figure out how to design roads that had to handle usage patterns, volumes, pedestrians and cars all sharing the same small space without constant gridlock and collisions? How did they even begin?
Second — the AI question. He uses Waze, has a love-hate relationship with it, mostly hate because the detours it dreams up around here are... let's say creative. But he sees the core idea. If enough people use a navigation app, the app becomes part of the infrastructure — it can route different drivers different ways across the same roads. And he wonders: why is this left to a private company? Could the Waze model be mandated, centralized, and deployed by the government? He knows it sounds fringe, he admits that upfront. But he wants to know if anyone has seriously thought about how intelligent routing could work as a public utility — a stopgap while we figure out the bigger mess.
Eight million car trips a week. That's what moves through Jerusalem. In a city whose oldest neighborhoods were laid out for donkeys and foot traffic. The math doesn't work — it hasn't worked for decades — and Daniel's two questions are really one question viewed from different ends of a century. How did we build the system that's failing, and can we retrofit intelligence onto it without repeating the mistakes that got us here?
So let's start at the beginning. When does traffic engineering become a profession?
Nineteen thirty. That's the year the Institute of Transportation Engineers was founded. And the number that matters alongside it: in nineteen twenty there were about eight million cars on American roads. By nineteen thirty, twenty-three million. That's a tripling in a single decade. No city on earth had planned for that. They were laying asphalt over cow paths and calling it infrastructure.
A tripling. In ten years. And I want to sit with that number for a second, because it's easy to hear "eight million to twenty-three million" and nod along without really feeling what that meant on the ground. Imagine your street today. Now imagine three times as many cars on it, but the street itself hasn't changed — same width, same intersections, same everything. And this happened everywhere, all at once, with no warning.
And no regulatory framework. Driver's licenses weren't standardized. Traffic laws varied from town to town. In some places you drove on the left, in others on the right. There were no lane markings — those don't appear until the nineteen twenties. Stop signs weren't standardized until the nineteen thirties. The stop sign we know today — red octagon, white letters — that wasn't codified until nineteen thirty-five. Before that, stop signs were whatever the local government felt like putting up. Different shapes, different colors, sometimes just a painted board that said "STOP" in whatever font the sign painter had.
So you've got twenty-three million cars, no consistent rules, and roads that were literally built for horses. That's not a transportation system. That's a public health crisis waiting to happen.
And it was. Traffic fatalities spiked dramatically in the nineteen twenties. Pedestrian deaths were especially brutal — cars were moving faster than anything people were used to, and neither drivers nor pedestrians had developed the instincts for sharing space. Children were being killed at rates that sparked genuine moral panic. This is actually when the term "jaywalking" enters the American vocabulary, and it's worth understanding where it came from. "Jay" was slang for a rural person, a hick — someone who didn't know city ways. The auto industry literally invented and popularized the concept of jaywalking as a way to shift blame from drivers to pedestrians. Before cars took over the streets, pedestrians walked wherever they wanted. The street was public space. The idea that crossing outside a designated zone was a violation — that was manufactured by car companies to clear the road for their product.
Wait. The auto industry invented jaywalking?
Campaigns funded by automakers and dealer associations in the nineteen twenties. They pushed for laws that restricted pedestrians to crosswalks. They ran ads shaming people who walked into the street. They lobbied to redefine the street as a space for cars, not people. And it worked. By the nineteen thirties, the cultural norm had flipped. The street belonged to vehicles. If you got hit, it was your fault for being there.
That's a pretty effective reframing. And it lines up with what you said earlier about the first traffic lights optimizing for car throughput. The infrastructure and the culture shifted together.
They had to. You can't sell twenty-three million cars if people are afraid to drive them, and you can't drive them safely if pedestrians are still treating the street like a public square. So the entire social contract around streets got rewritten in about fifteen years. And the engineers who were trying to make sense of all this — they weren't malicious. They were overwhelmed. They had a tripling vehicle fleet, a fatality crisis, and no established science for how to handle it.
So they reached for what they knew.
They reached for fluid dynamics. They started modeling traffic as a stream. Cars as molecules. You've got flow rate, density, speed. The "traffic stream" wasn't a metaphor to them, it was the working model. They literally borrowed equations from hydraulics. And if you model cars as water molecules, certain conclusions follow naturally. Water doesn't stop to chat with a neighbor. Water doesn't cross the street to buy bread. Water just flows, and your job is to keep it flowing.
Which explains a lot about how we ended up with eight-lane highways that still jam. Water doesn't change its mind about where it's going. Water doesn't double-park. And water definitely doesn't see a friend on the sidewalk and slam the brakes.
Right. The first electric traffic signal goes up in Cleveland, Ohio in nineteen fourteen. It's just a timer — red, green, red, green, no intelligence at all. But by the nineteen twenties they're experimenting with pressure plates in the road and inductive loops that can actually detect a vehicle and adjust the timing. That's the first responsive infrastructure. And it's clever, genuinely clever for its era, but it's doing exactly one thing: optimizing for car throughput. The pedestrian? An afterthought. Someone who interrupts the flow.
So from the very first traffic light, the assumption is baked in. The road belongs to cars. And if you're a pedestrian waiting to cross, the system sees you as a disruption. You're not a user of the road — you're a blockage in the pipe.
And then in nineteen forty-four, the Highway Capacity Manual is published. This is the foundational document. It gives planners a formal method for measuring road capacity and what they call "level of service." An A through F grading system, like a report card for roads. A means free-flowing traffic. F means gridlock. And the entire framework assumes the primary user is the automobile. Everything else — pedestrians, cyclists, transit — is friction.
Level of service. That phrase is still around, isn't it?
Still in use today. Still grading intersections. And still carrying the same baked-in assumption. Here's the thing, Corn — this is where the induced demand paradox was born, even if nobody named it yet. The manual tells you how many cars a road can handle. So you widen the road to handle more cars. And more cars show up. And your level of service drops again. And you widen again. The manual gave planners a tool that looked scientific, but it was measuring the wrong thing.
It's like treating a fever by buying a bigger thermometer.
That's... actually exactly right. The thermometer gets longer, the fever looks worse, so you buy an even longer thermometer. And the whole time you're congratulating yourself for being so rigorous and data-driven.
So the nineteen forties gave us the rulebook, and the rulebook said cars win. But there's a specific face to this story. Robert Moses.
The man who reshaped New York. Moses looked at a neighborhood and saw obstacles to vehicle speed. He ran the Cross-Bronx Expressway straight through established communities — hundreds of thousands of people displaced over his career. His philosophy was simple: the car is the future, and anything that slows the car down has to go. Short blocks, mixed-use streets, front stoops, corner stores — these were inefficiencies to be bulldozed.
And it's worth pointing out that Moses wasn't some rogue operator. He held multiple positions simultaneously — parks commissioner, head of the Triborough Bridge Authority, other roles — and he used that overlapping authority to make himself essentially unaccountable. No single person could overrule him because no single person had jurisdiction over everything he controlled. He was the coordination problem, deliberately.
And he wielded that power for decades. The Cross-Bronx Expressway alone displaced something like sixty thousand people. He chose the route specifically because it was cheaper to bulldoze working-class neighborhoods than to go around them. The human cost was a feature, not a bug — it saved money.
And that's when Jane Jacobs enters the picture.
Nineteen sixties, Greenwich Village. Moses wants to run a highway through Washington Square Park. Jacobs looks at the same streets and sees something completely different. She sees that short blocks disperse traffic naturally. That mixed-use neighborhoods mean people walk instead of drive. That "eyes on the street" — shopkeepers, neighbors, kids playing — that's the real safety system, not wider lanes and higher speed limits.
She famously organized the opposition and won. Moses didn't get his highway through the Village. But here's what I find striking about that story — she was fighting against the entire professional consensus. The planners, the engineers, the government agencies — they all agreed with Moses. She was a journalist with no formal training in urban planning, and she was telling the entire profession they had it backwards.
And she was right. Everything we now know about urban vitality — the research on street connectivity, on mixed-use zoning, on the relationship between pedestrian activity and safety — it all validates what Jacobs argued from observation. She didn't have models or equations. She just walked around and paid attention.
And Jerusalem is basically a Jacobs city that got Moses treatment.
That's the tragedy of it. You've got neighborhoods built centuries ago — narrow streets, tight corners, pedestrian scale by design. And then you layer on modern car volumes and light-rail construction and high-rises, all without the kind of coordinated planning that Jacobs argued for. Daniel's frustration about everything happening at once? That's not an accident. That's what happens when you apply Moses-style, top-down infrastructure thinking to a city whose bones are Jacobs-scale.
The construction isn't coordinated because the agencies doing it don't talk to each other, and they're all optimizing for their own project's throughput. Sound familiar?
The same flaw. The same single-variable optimization. So that's the first part of Daniel's question — the history. Traffic engineering was born reactive, built on a car-first model, and the tools we still use today were designed for a world that never anticipated pedestrians and cars having to coexist. The nineteen fourteen traffic light, the nineteen forty-four manual, the Moses playbook — they all optimize for vehicle flow. And when you optimize for one thing, you get really good at that one thing, and everything else degrades around it.
Which brings us to the second question, because it's basically the same question with better tools. If the old model optimized for cars individually, Waze optimizes for drivers individually — but it's got data those early planners couldn't have dreamed of. So could we flip the model? Optimize for the whole system instead?
Let's talk about what Waze actually is under the hood. It's a crowdsourced, real-time data pipeline. Every phone running Waze is reporting speed, location, and route back to the mothership. The app builds a live model of traffic conditions and then routes each driver to minimize their individual travel time. That's the key word — individual. Waze doesn't care if shaving two minutes off your commute sends you through a residential street where kids are playing. It's solving for you, not for the street.
And Daniel's experience — the weird detours through strange places — that's the algorithm finding paths the road designers never intended as through-routes. Rat-running, it's called.
Rat-running. And here's what most people don't know. Waze already shares its data with governments. The Connected Citizens Program launched in twenty fourteen, and it now has over thirteen hundred government partners worldwide. Cities get anonymized traffic data — where the jams are, how long they last, what routes people are taking. So the infrastructure for government access to this kind of intelligence already exists. Waze built it voluntarily.
Wait. Thirteen hundred government partners? So the data pipeline is already half-public.
Already there. The question Daniel's asking is whether we can go further — whether the routing decisions themselves could be made by a public agency instead of a private app. And this isn't as fringe as he thinks. Singapore has been doing centralized traffic demand management since nineteen ninety-eight.
The Electronic Road Pricing system.
Gantries over every major road. Variable pricing that changes with congestion levels. If you drive into the central business district at peak hour, you pay more. The system doesn't route you — but it shapes your behavior through price signals. And it works. Traffic in the restricted zone dropped by about fifteen percent almost immediately after launch, and it's stayed down.
So Singapore proves you can centralize the demand-management layer. But they're using pricing, not routing. Different tool. And I'm curious — how does that actually feel on the ground? If I'm a driver in Singapore, what's my experience of this system?
You have an in-vehicle unit on your dashboard — it's basically a transponder with a stored-value card, like a prepaid toll pass. Every time you pass under a gantry, the system deducts the current rate. The rate changes throughout the day based on real-time traffic speeds. You hear a beep, the money comes out, and you keep driving. It's frictionless in the moment, but the price signal accumulates over time. You start thinking, "Do I really need to drive through the central zone at eight forty-five, or can I leave at nine-thirty when the rate drops?"
So it's behavioral conditioning through small, repeated transactions. You don't get a lecture about congestion. You just get a beep and a slightly lighter wallet.
And it works because it's consistent and predictable. The rates are published. The gantries are visible. There's no ambiguity about what you'll pay. Compare that to Waze, where the routing logic is a black box — you don't know why it sent you down a particular street, and you can't predict whether it'll do it again tomorrow. Singapore's system is transparent in a way that algorithmic routing currently isn't.
Singapore proves you can do centralized demand management. London's congestion charge does something similar. But the model that's closer to what Daniel's imagining is the UK's National Traffic Information Service.
Right. It's a government-run data platform that collects real-time traffic information and provides it to navigation providers — Google Maps, Apple Maps, Waze, all of them. The government isn't doing the routing, but it's providing the authoritative data layer that routing decisions are built on. It's a nudge, not a mandate.
There's a spectrum. On one end, Waze does everything privately and governments just get a data feed. On the other end, a fully centralized system where the government tells every car where to go. And the UK is somewhere in the middle.
I think the middle is where the realistic near-term answer lives. A fully mandated system — the government routes every vehicle — faces two enormous problems. One is technical, one is political.
Start with technical.
The networking side is actually solvable. Modern cellular networks — 5G especially — can handle the latency and bandwidth for real-time routing instructions. The harder problem is compliance. How do you enforce routing without mandatory hardware in every vehicle? Singapore's ERP uses physical gantries and in-vehicle units. That's expensive to deploy. You could do it through smartphone apps, but then you're back to the Waze problem — you can't compel everyone to use the app.
If even ten percent of drivers opt out, your system model is incomplete.
Incomplete and potentially worse than nothing, because the routed drivers are now making decisions based on a model that's missing data. Garbage in, garbage out. And there's a subtler problem too — if the system routes compliant drivers away from congested areas, the non-compliant drivers get clearer roads. You've accidentally created an incentive to opt out.
The free-rider problem. If everyone else follows the routing instructions, the roads are empty enough that I can ignore them and get where I'm going even faster.
The moment enough people realize that, the system unravels. So you can't just make it optional and hope for the best. You need either a mandate or a set of incentives strong enough that opting out feels like a penalty.
And the political problem?
Privacy. A centralized routing system would know every vehicle's origin, destination, and real-time location. That's a surveillance infrastructure that makes a lot of people very nervous. And the objection isn't unreasonable. We're talking about a government database that tracks where every car in the country goes, at what time, on what route. Even if you trust the current government, you have to think about what a future government might do with that data.
But — and this is where I think Daniel's framing is useful — the private apps already have this data. Google knows where I am. Waze knows where I'm going. The data exists. The question isn't whether it's collected. It's who holds it and under what rules.
That's the governance argument. A public agency has democratic oversight — in theory. A private company has profit incentives and a terms-of-service document nobody reads. I'm not sure which is more reassuring, honestly. But the privacy objection to centralized routing often assumes the alternative is privacy, when the alternative is just corporate surveillance instead of government surveillance.
If we're picking our surveillance, maybe pick the one with elections.
Maybe. The stronger argument for the hybrid model is that it avoids the compliance problem entirely. Here's how it could work. The government builds a routing API — a public, transparent system that calculates optimal routes based on system-wide metrics. Total vehicle-hours, emissions, equity across neighborhoods. Private apps — Waze, Google Maps, Apple Maps — are required to query this API when routing drivers through congested zones. The app still handles the user interface, still provides turn-by-turn directions, but the route logic comes from the public layer.
If a driver refuses to use any app at all?
Then they're not in the system, and that's a problem. But you can incentivize compliance. Priority routing for drivers who opt in. Congestion pricing discounts. Access to bus lanes during peak hours. You don't have to mandate participation if the benefits of participating are strong enough.
The carrot instead of the stick. And the carrot is "you'll get there faster than the people who didn't opt in."
Which is basically the Waze pitch already. Waze users get better routes than non-users. The difference is that the optimization target changes from individual travel time to system-wide efficiency. And that's the real paradigm shift Daniel's question points toward.
Let me see if I can pull this together. The history of traffic engineering is a history of optimizing for one thing — car throughput. The nineteen fourteen traffic light, the nineteen forty-four manual, the Moses approach — they all assume the car is the only user that matters. Waze is the logical endpoint of that thinking, just with better data. It optimizes for one driver at a time. And the result, in a city like Jerusalem, is rat-running through narrow streets that were never meant to be through-routes, because the algorithm is solving the wrong problem.
The alternative — centralized, government-run routing — isn't technically impossible. The data infrastructure exists. The networking can handle it. Singapore and London have proven you can manage demand at scale. The UK model shows you can provide a public data layer without mandating a single app. But the political hurdles are real, and the privacy questions are legitimate even if they're sometimes selectively applied.
The hybrid model seems like the most plausible near-term path. Government provides the routing intelligence as a public utility. Private apps consume it. Drivers get incentives to opt in. Nobody gets forced, but the system gets smarter.
That's the stopgap Daniel was asking about. It's not the ideal city. The ideal city doesn't need algorithmic routing because it's designed for walking and transit and micro-mobility from the start. But Jerusalem isn't getting bulldozed and rebuilt. So you work with what you've got. You layer intelligence on top of the infrastructure you inherited, and you try to make it hurt less.
There's something almost clinical about that framing. Harm reduction. Symptom control. Those are Daniel's words, and they're exactly right.
I spent decades in medicine before this podcast, and the parallel is uncomfortable but real. When a patient shows up with a condition that can't be cured — only managed — you don't throw up your hands and say "come back when you're healthy." You manage it. You reduce the harm. You control the symptoms. Traffic in Jerusalem is a chronic condition. The cure is rebuilding the city, and that's not happening. So you manage.
The question is whether we manage it with tools designed a century ago for a different problem, or with tools that actually see the whole system.
That's where I think Daniel's instinct is right, even if he calls it fringe. The Waze model — crowdsourced data, dynamic routing, real-time adaptation — that's powerful. The mistake is letting it optimize for the wrong thing. Flip the objective function from individual speed to system health, and you've got something that looks a lot like public infrastructure.
Although — let me push on this a little. Even the hybrid model assumes we can define "system health" in a way that's fair. Who decides which neighborhoods get routed through and which get protected? The same agencies that are currently digging up Jerusalem without coordinating with each other?
That's the governance problem, and it's harder than the technology. But it's also not new. Cities already make these decisions — they just make them with static tools. Speed bumps, one-way streets, no-turn signs. Those are routing decisions. They're just carved in asphalt instead of calculated in real time. A dynamic system could actually be fairer, because you can adjust. A residential street that's a rat-run at eight AM could be protected during school drop-off hours and available as an overflow route at noon.
The dynamism cuts both ways. It could be more responsive to community needs, not just driver needs. But that also means the routing algorithm is making value judgments every few minutes — this street is a through-route now, a protected zone an hour from now. And someone has to program those thresholds.
Someone has to be accountable when they get it wrong. That's the part that worries me. Right now, if your street gets turned into a rat-run, you can at least identify the problem — too many cars, not enough enforcement, whatever. You can go to a city council meeting and complain. If the routing is done by an algorithm, who do you complain to? The API? The optimization function?
"Your honor, I'd like to file a complaint against the objective function."
This isn't a hypothetical problem. We've already seen it with Waze. Residents in Los Angeles, in London, in towns all over the world have complained about app-routed traffic flooding their streets. And Waze's response is basically "we're just showing drivers the fastest route." There's no mechanism for a neighborhood to say "not us, not at this hour." The algorithm doesn't have a hearing process.
The governance layer isn't just about who writes the code. It's about who hears the appeals.
That's the piece nobody has built yet. Singapore's ERP has a governance structure — the Land Transport Authority sets the rates, publishes them, and can be held accountable through normal political channels. But algorithmic routing is different. The decisions are made continuously, by software, at a scale that no human committee could review. You'd need a whole new kind of oversight — maybe something like an algorithmic auditor, a public office that reviews routing patterns for equity impacts and has the authority to adjust the parameters.
That sounds expensive and slow and bureaucratic. Which, to be fair, is also a reasonable description of how we build roads now.
It is. And the alternative — letting a private app optimize for whatever it wants — is fast and cheap and completely unaccountable. Pick your poison.
If you take one thing from this episode, I think it's that the tools we use to manage traffic were built for a world that no longer exists — and the new tools we're building are still optimizing for the wrong variable. The nineteen forty-four manual optimized for cars. Waze optimizes for individual drivers. Neither one optimizes for a city.
The one thing I'd add: the shift from individual to system-level optimization isn't a technical problem anymore. The data exists, the pipelines exist, the networking can handle it. The hard part is governance — who decides what "optimal" means, and who trusts them to decide. That's a political question dressed up as a routing algorithm.
Which leaves us with an open question that's bigger than traffic. If we could design a city from scratch today, would private cars even be part of it? Jerusalem's eight million weekly car trips in a city built for pedestrian scale — that's not a traffic problem. That's a design failure. And as autonomous vehicles become common, the line between private and public routing is going to blur whether we plan for it or not. Every self-driving car will be a node in a centralized system. The only question is who controls the routing logic.
That question is coming faster than most people realize. Autonomous vehicles don't work without centralized coordination — you can't have a bunch of independent self-driving cars making individual routing decisions and expect the system to be safe or efficient. They'll need to talk to each other, to the infrastructure, to some kind of routing authority. The architecture of that system is being designed right now, mostly by private companies, mostly without public input. Daniel's question about whether routing should be a public utility isn't fringe at all. It's just early.
If you want to go deeper on this, the Highway Capacity Manual is surprisingly readable for a government document — it's a fascinating look at how the assumptions got baked in. And if you've got thoughts on centralized routing, send them in. We might do a follow-up. Our producer Hilbert Flumingtop, as always, keeps this show running.
We'll be back soon.
