Chasing the Robotaxi: A Full Day Tailings Waymo Through San Francisco
Almost anywhere in the world, if you find yourself gridlocked in rush-hour traffic, you’re still surrounded by the same familiar sight: row after row of heads on shoulders, arms hooked to steering wheels, all piloting their own vehicles. But in a small handful of metro areas—Los Angeles, Phoenix, San Francisco, and Wuhan, China—you can pull up next to a taxi that has no one behind the wheel at all: a fully autonomous vehicle that picks up paying passengers, no human oversight required. If you call one of these cities home, you’ve likely stopped even doing a double-take at the sight. The future of transport feels split cleanly in two: one half that has never encountered it, and another that has already stopped noticing it’s there.
It’s easy to grow numb to tech hype cycles, after all. Autonomous vehicles, a staple of sci-fi convenience for decades, were the hottest ticket in tech around 2018. Then everything shifted: a test self-driving Uber killed a pedestrian jaywalking in Arizona, regulatory scrutiny ramped up dramatically, and dozens of self-driving projects folded as broader disillusionment with unproven tech set in. The entire autonomous driving industry started to look like it was on the ropes, if not completely irrelevant.
That slowdown, though, was just a rope-a-dope feint. Investment capital has kept pouring into the space. Since 2020, Alphabet’s subsidiary Waymo alone has secured more than $11 billion in funding. Over the past few years, hundreds of empty, all-electric Jaguar SUVs have begun roaming the streets of Waymo’s three U.S. operating markets, eventually picking up paying passengers and charging for rides. Around the same time, a similar fleet of white robotaxis operated by Baidu’s Apollo Go spread across Wuhan. The long-imagined future of driverless ride-hailing is finally here—quietly—and all indicators point to it expanding rapidly. Waymo is set to launch operations in two new U.S. cities, Austin, Texas, and Atlanta, Georgia, in early 2025. Right now, driverless ride-hailing is in varying early stages of testing in Las Vegas, Houston, Detroit, Seattle, and at least seven other Chinese cities, plus test programs in Japan, the United Arab Emirates, Germany, South Korea, the U.K., Sweden, and Singapore.
WIRED maintains a newsroom in San Francisco, one of Waymo’s first launch cities, and that gives us a unique perspective—for better and worse. Like most locals, the novelty of sharing the road with a fleet of driverless robots has long worn off for us. Even one of our 90-year-old family matriarchs, who took her first Waymo ride with us, felt instantly at ease with her sci-fi chauffeur. “There was no period of getting used to it,” she told us. (To be clear, WIRED as an outlet doesn’t have one shared mother—but this story has multiple authors, so we’re using a collective voice here, much like Alphabet frames its work: it doesn’t build a fleet of separate robotaxis, it develops one single, unified “Waymo driver.”)
To write the most grounded, useful report from this already-arrived future, we realized we needed to make driverless cars feel new and strange again. We wanted to dig past surface-level first impressions to uncover what years of Waymo operations have actually taught San Franciscans. So we started brainstorming.
San Francisco has two huge car-related claims to fame: it’s ground zero for the first mainstream rollout of self-driving taxis, and it was the backdrop for one of the most iconic car chases in cinematic history. So we settled on a plan: we’d chase a Waymo robotaxi across the city for a full workday, to get a fresh look at what this future actually looks like.
Here’s how it would work: we’d pile a small team of reporters into a traditional, human-driven hire car, then follow one Waymo wherever it went all day. We’d analyze its driving choices, how it interacts with other street users, and its overall approach to navigating a chaotic city. We’d interview any passengers that agreed to talk to us, and see the robotaxi through the eyes of a human driver—the exact job Waymo is designed to replace. We’d chase it for a minimum of six hours, or until we crashed in a fiery wreck. Whichever came first.
On the morning of our chase, we met in a parking lot outside WIRED’s San Francisco office to meet our chase car driver. If you can squint past the morning sun, let’s set the scene: we promised ourselves he’d look just like Steve McQueen, the race-car-driving lead of Bullitt, the iconic San Francisco car chase film.
Fine, he doesn’t look anything like Steve McQueen. If anything, he’s a dead ringer for Jean Reno, the stubbled French-Moroccan action star of Ronin, another classic car chase flick. Only our driver is friendlier, more talkative, obsessed with dad jokes, and far less of an ex-military assassin. His name is Gabe Ets-Hokin.
A third-generation San Franciscan, Gabe grew up playing with Nancy Pelosi’s kids and went to high school with Gavin Newsom; driving is in his blood the same way politics is in theirs. He’s worked as a taxi, Uber, and Lyft driver since 1995, and even helped organize a taxi worker strike in the late 1990s. For the past 20 years, he’s also written about driving, ride-hailing, and motorcycling for outlets across the country. And for what it’s worth, if you think our movie trope bit is silly: Gabe was a machine gunner with the U.S. Marines during the first Gulf War, so he is at least actually ex-military. He drove a gray Hyundai Ioniq 5 electric car (9/10, WIRED recommends), with his military service ribbons taped to the dashboard, and a 100-year-old ukulele sticking out of the center console.
The chase kicked off right on schedule: one of our team hailed a Waymo a few blocks away, rode it to the edge of our parking lot meeting spot, then jumped out to join the rest of us in the chase car. “You know what you have to say, right?” Gabe said from the driver’s seat as we scrambled to buckle our seatbelts.
We blinked.
“C’mon!” he said. “Haven’t you ever seen an old movie? When you jump into the cab, you yell ‘Follow that car!’”
But our target Waymo just sat there, idle, for two long, awkward minutes. That gave us plenty of time to stare at our quarry: its shape is a little like a cartoon shark with a bunch of spinning sensors stuck to its roof and body, and it was “staring” right back at us through 29 cameras and five lidar sensors, mapping every detail of our car and team.
“It looks shy,” Gabe said.
“It’s ashamed,” we joked back. “It knows we’re tricking it.”
Then, at 10:42 a.m., the Waymo finally pulled out. We yelled, just like the old movies: “Follow that car!”
Less than a minute later, Gabe sighed. “I’m not used to driving this slow.”
Before we go further, let’s get one core truth out of the way: riding in a self-driving car, especially for the first time, is an undeniably cool experience. It starts out feeling like a trip to an amusement park: the empty “gondola” glides up to the curb, you step in, and close the door. Then it becomes the exact opposite of an amusement park ride. No sharp turns, no sudden jolts, no loud noise. Just you, soft black leather seats, a calm default computer voice, and for now, a steering wheel that turns gently back and forth on its own, like a ghost at the wheel.
Social media platforms from Reddit to Substack to X are full of glowing write-ups of first Waymo rides, most of which favor the driverless experience over what many call the “messy” experience of riding with a human Uber or Lyft driver. (Of course, social media is also full of clips of confused Waymos stuck in intersections, and even videos of people vandalizing driverless SUVs.) The glowing reviews all tend to hit the same two key notes, which our 90-year-old matriarch echoed after her first ride with us.
“I felt safe, like there was a really responsible driver in charge that knew all the rules and would follow them,” she said. “You don’t have to take a chance like you do with a human driver.” And even though she knew the car was full of sensors and microphones, she didn’t feel like anyone was listening in or intruding on our conversation. “Talking in the car felt way more private,” she said. “There wasn’t another person there part of it.”
In our chase car, Gabe is front and center, which is a win for us: he’s easily the most interesting person in the vehicle. He can rattle off a bit of local history for almost every block we pass: that’s the public housing where O.J. Simpson grew up. That’s the public library where the founder of the Silk Road dark web marketplace was arrested. That’s the Safeway where former mayor Ed Lee suffered his fatal heart attack.
So far, Gabe feels pretty confident going up against the robot that could replace him. From his past experiences sharing the road with Waymos, he’s gotten the impression that they’re slow, plodding, and follow traffic rules to a fault. Our first target, license plate 53516F3, hasn’t done anything to challenge that stereotype yet. “It drives better than my mom, but she’s 85,” Gabe jokes. He nods at the speedometer: “Look at that. Exactly the speed limit.”
“Would you rather have more of these on the road than human drivers, just from a driving style perspective?” we ask, as our Waymo pulls away from a stop with the calm, steady grace of someone doing morning tai chi.
Gabe actually shudders at the thought. “That would slow traffic down so much,” he says.
“But the whole point is that it would save more lives, right?” we push back.
“I don’t know, would it really cut deaths that much?” Gabe says. “Fatal crashes on surface streets are pretty rare already.” Then he makes a bigger point about the future of cities: “We could actually eliminate traffic deaths entirely if we wanted to. What would actually cut car and pedestrian deaths is moving cities away from car dependence—and autonomous vehicles do the exact opposite of that, by definition. If this is the future of urban transport,” he says, nodding at the spinning lidar on the back of 53516F3, “then we have to keep building cities around huge streets, massive parking lots, traffic lights, lanes, all this infrastructure that takes up so much space and benefits so few people.”
We’re caught off guard for a second. “So you’re in favor of moving away from car-centric cities?” we ask.
“Absolutely!” Gabe says as we roll south. (It turns out Gabe is even more of a rarity than an ex-military assassin: he’s an ex-military urban planning nerd.) Cities built around far fewer private cars would be better for everyone, he says, and even better for ride-hailing in the long run: “If you don’t own a car yourself, you’re way more likely to take an Uber or Lyft.”
Or a Waymo, we could add. But we’re too busy wondering what’s coming next. We start to overthink: when our Waymo picks up its first passenger, will they notice they’re being followed by a car full of reporters taking notes and holding video cameras? When we run over to interview them at their stop, will they call us stalkers? Will we trigger some algorithm that alerts Waymo that we’re tracking one of its vehicles?
The more we watch the street, the more we realize our worries are pointless. San Francisco has so many tourists that a Waymo draws gawkers just like a parade float does. They’re also pointless because our first target 53516F3 never picks up a passenger: less than 30 minutes after we started, it turns left into a fenced parking lot across from a discount grocery store. There are no EV chargers in sight, but the lot is packed with idle Waymos, some parked, some pulling in and out, every single one still spinning its lidar sensors. The whole scene looks like a writhing hive of robot cars. (On a nearby rooftop, there’s a 40-foot wire sculpture of a reclining naked woman left over from Burning Man; it looks like a nudist at a picnic that’s just been overrun by robot ants.)
We pulled over to wait for 53516F3 to come back out, and Gabe pulled out his ukulele to pass the time. After 10 minutes, we gave up on our original plan of following one single vehicle all day. We picked a random new Waymo just leaving the lot: license plate 40757F3. We took off after it, holding a steady 25 mph down Bryant Street, passing so close to our office that a coworker tracking our location snapped a photo of us through the window. Twenty minutes later, 40757F3 led us straight to another Waymo parking lot just a few blocks from WIRED’s office.
We picked up the trail of the first robotaxi that left that lot: 78889X3. We had high hopes at first, then we noticed the yellow “low battery” light glowing on its rooftop dome. “If this one heads to another lot, I’m going to scream,” we said.
Sure enough, 78889X3 led us three miles south to a charging and maintenance lot tucked in an industrial zone under an elevated freeway. As we waited outside the fence for our next target, a local man wandered over and offered us a sip of his vodka. We declined, and our mood dipped along with our energy. When a new Waymo, 53499F3, pulled out of the gate, we said flatly: “Follow that car.”
It’s now about 11:30 a.m. Late morning is the slowest part of the day for ride-hailing, Gabe says; his Uber and Lyft apps have shown almost no demand since we started our chase. All this empty driving between parking lots got us thinking more about Gabe’s argument over the future of cities. People already complain that most cars on the road today are 4,000-pound vehicles carrying just one person. Will Waymo make congestion worse by filling streets with 5,000-pound fully empty robot cars?
When we asked Waymo’s chief product officer Saswat Panigrahi about all this empty driving between lots, he calls it “fleet rebalancing.” During slow periods, he explains, cars automatically move to positions where they’ll be ready to meet expected peak demand later in the day, and they also time their trips to charging stations for when demand is low. As ridership picks up later in the day, we’ll see far fewer empty Waymos on the road.
When we asked the bigger question of how Waymo will reshape cities, Panigrahi was eager to talk, and at times he sounded like an evangelist from the mid-2010s rideshare revolution. Back then, Uber’s brash CEO Travis Kalanick loved to rail against the waste and inefficiency of car culture. “Around 20 percent to 30 percent of our land is taken up just storing these hunks of metal that we drive around in for 4 percent of the day,” he said back in 2016. Today in 2024, Panigrahi echoes that frustration: “In cities, there’s more space reserved for cars than for humans.”
Kalanick promised Uber would kill private car ownership and end urban congestion. Panigrahi isn’t quite that hyperbolic, but he does argue Waymo will create a world where “fewer metal hunks take up valuable urban real estate.” Gabe, of course, predicts the opposite: autonomous vehicles will only make congestion worse. So who’s right?
According to Adam Millard-Ball, an urban planning professor who studies autonomous vehicles, the past 15 years of Uber and Lyft give us a pretty good preview of how robotaxis will affect traffic. Whether they have a driver or not, both are on-demand private taxi services powered by algorithmic matching. And research shows Uber and Lyft actually put more cars on city streets, partly because thousands of new drivers bought new cars to work for the platforms. All that extra traffic led to, you guessed it, worse congestion.
Of course, no one is buying new cars to drive for Waymo, but Millard-Ball argues driverless cars could still lead to even more gridlock, mostly because cities don’t charge drivers for using busy roads. Right now, driving through a crowded downtown is free; the main cost that pushes car owners to take other transit is high parking prices. Think about the $1,000 parking spots that popped up for Taylor Swift’s Eras Tour, and the corresponding spike in public transit use in every city she visited. The problem is, robotaxis don’t need to park downtown. And they could make ride-hailing cheaper overall (we’ll get back to Waymo’s economics later), since there’s no human driver like Gabe to pay. That adds up to a simple recipe: cheap, easy, widely available private cars on every road. That’s a recipe for constant, endless traffic.
What about self-driving buses? Waymo says its “Waymo driver” software could eventually pilot trucks, delivery vans, large shuttles, any kind of vehicle. The problem is, it’s hard to cut labor costs with self-driving buses; when dozens of lives are riding in one vehicle, cities will almost certainly require a human supervisor to be on board to manage passengers and emergencies.
In other words, technology alone won’t magically fix cities, Millard-Ball says; you need smart policy too. If cities started charging people to drive on downtown streets, not just to park there, that would discourage unnecessary car trips and push more people onto climate-friendly buses, trains, and bikes, with ride-hailing only for occasional trips. Ironically, Uber and Lyft actually support this idea. In New York City, the companies joined with urban planners to lobby for a congestion pricing program that charges drivers to enter lower Manhattan. After years of political delays and wrangling, the groundbreaking program will finally launch this coming January: most drivers will pay $9 to enter lower Manhattan, and the program is expected to raise $500 million a year for public transit.
The arrival of a major new technology like robotaxis, which needs city approval