It's MOIA Time.
Plus, Waymo-hating personal injury lawyers, "sleepers", and why the China speed debate is misleading.
This is the Ride AI newsletter: The most comprehensive weekly digest of news and intelligence at the intersection of technology and transportation.
Early Access for Ride AI 2027 Now Available.
$750 Early Access tickets for Ride AI 2027 are now available until July 31st. Starting tomorrow, pricing climbs month by month toward $1,500 general admission.
If you’re considering joining us next April at SF JAZZ, please apply for early access today. Tickets won’t ever be at this price again.
Top Story
MOIA has launched passenger test operations for autonomous ID. Buzz shuttles in Hamburg, letting pre-registered residents book self-driving rides through the MOIA app for the first time. The launch is part of ALIKE, a Hamburg pilot funded by Germany’s Federal Ministry for Transport, and covers parts of Winterhude, Barmbek, and Wandsbek to start. Rides are free during the pilot and run with a trained safety operator on board, with up to ten autonomous vehicles deployed at a time, a smaller operation in scale for now.
The ID. Buzz vehicles run on Mobileye Drive, Mobileye’s self-driving stack built for robotaxi and shuttle operations. MOIA now has more than 100 ID. Buzz vehicles running the Mobileye Drive stack on public roads (for testing) across seven cities in North America and Europe, with Volkswagen building the vehicles and integrators like MOIA working with regional partners to deploy them.
This seems to be a very pivotal time for autonomous efforts inside the Volkswagen Group. Cariad, VW’s software unit, ended its Automated Driving Alliance with Bosch earlier this month and has been shopping for a new partner to get to SAE Level 3 faster, with Mobileye supposedly the leading internal candidate since VW already uses its tech for highway assist and now for MOIA. Now, all eyes are on Hamburg. It’s the first time riders outside a controlled test have actually booked a driverless ID. Buzz, and it’s also the first time we get to see Mobileye’s technology deployed at a meaningful scale for public consumers.
Note: an earlier version of this piece could be read to imply that Mobileye was running the ID. Buzz fleet themselves. The wording has been updated to clarify MOIA, Mobileye, and Volkswagen’s relationship.
Domestic News
Tesla has confirmed Starlink V5 is built directly into the Cybercab, per a cutaway diagram the company posted showing the terminal integrated alongside the FSD computer and the rest of the sensor suite. The system’s inclusion has long been rumored along with its placement due to the opaque rear panel where a rear window would have gone, but this is the first official confirmation we’ve gotten.
Tesla has also enabled direct airport terminal service to San Francisco International Airport as of July 20th. As of this writing, a single way trip seems to cost around $30, and riders can pick which terminal they want to be dropped off at directly from the app. The service is, however, still technically classified as a limo service in California, not an autonomous vehicle service, and is still supervised with a driver. It’s also still almost impossible to hail a vehicle due to persistent high service demand for the area’s 40 or so vehicles. Nonetheless, very cool.
San Francisco mayor Daniel Lurie has asked California’s Department of Transportation to impose tougher statewide rules on robotaxi operators after a Fourth of July gridlock event left dozens of Waymo vehicles stranded across the city. He wants operators to prove they can clear cars from active lanes in real time, share live operations data with local agencies, and handle a sudden surge of riders during a crisis, pointing to both the July 4th fireworks gridlock and December’s citywide blackout as evidence the current voluntary approach isn’t enough. Waymo now runs in 11 cities and completes over 500,000 paid rides a week, which Lurie says is all the more reason for greater coordination.
Zoox has issued a software recall across its 105-vehicle fleet after one of its robotaxis got confused navigating a smoke-filled emergency scene in June, braking hard and stopping until a teleoperator reversed it clear.
I actually also experienced a hard braking event due to steam from a manhole during a Zoox ride, also in June, also brought on because an emergency vehicle came up behind us flashing sirens and horns, making the Zoox pull over. After the firetruck had passed, the car continued to refuse to move due to the steam from the manhole. It wasn’t until the vehicle executed an extremely long reversing procedure (no doubt prompted by a remote operator) did it then resume driving. At the time, I’d left feedback like I always do and thought nothing of it. Now I’m wondering if this incident is connected…
The recall lands a week after NHTSA warned AV developers that interfering with first responders will have consequences. It’s Zoox’s fourth software recall to date.
Emergency responders have fielded nearly 100 calls about robotaxi passengers falling asleep in Waymos and failing to wake up when they reach their destination, according to new Bloomberg reporting on the messier side of driverless rides. It feels like every time every time an operator launches a public service, they always need to learn that the hardest part to deal with might not be the technology, but the human element. In one case, a 30-year-old fell asleep during a six-minute ride home from a birthday party in Los Angeles. His trip had ended nearly an hour before firefighters opened the door to check on him, after a remote Waymo assistant tried waking him over the speaker and then called 911.
Austin-Travis County EMS commander Roger Patterson says his department has coined a term for it, “sleepers,” and logged 99 such calls in Waymo’s first nine months operating there. Only about 3% require hospital transport, but they still tie up ambulance crews. San Francisco reportedly logged 250 similar incidents in 2025.
And it’s not just naps. A passenger gave birth in the back of a Waymo in San Francisco this past December, at least the second Waymo delivery on record after one in Phoenix. Cleanup crews report finding blood, vomit, and worse. When Cruise was still operating, riders reported performing sexual acts in the car.
Tesla charges riders $50 to $150 for messes depending on severity, and Waymo charges $50 to $100, although I haven’t heard of anybody actually getting charged… yet. Have you? Let me know in the comments.
Trial lawyers have been quietly lobbying against autonomous vehicle legislation for nearly a decade, according to economist Alex Tabarrok. Tabarrok posits that the trial lawyers’ lobby keeps fighting AV bills because mandatory insurance, like the UK’s model where insurers pay victims directly and then pursue manufacturers, would cut litigation, thereby dramatically cutting into a large part of trial lawyers’ income.
And they’re not wrong. Over 220 million miles, Waymo reports 94% fewer serious-injury crashes and 93% fewer pedestrian injuries than matched human driving, yet a large majority of people still don’t consider AVs as safer. Neurosurgeon Jonathan Slotkin’s essay in Noema argues the data alone won’t fix that gap. Most people rate themselves an above-average driver, so a machine’s mistake, even a Waymo safely stopped at a dark intersection during December’s SF blackout, seems scarier than the roughly 115 Americans who die in ordinary crashes on a typical day.
Move Fast, Ship Cars
In a lot of ways, China’s playbook for bringing anything to market, EVs included, is simple. Ship something good enough now, fix it later, prioritize Time to Market above all else. I’ve been seeing a lot of discussion about this recently, and it is absolutely true. Chinese EV makers now take about 24 months to get a new model from concept to showroom, less than half the 40-to-50 month cycle most legacy automakers still run, per McKinsey. Take Xiaomi. From zero automotive experience to shipping its SU7 sedan in roughly three years, then 100,000 units built in 230 days, the company is the poster child of Chinese speed.
Autonomous driving gets the same “good enough” treatment. One driverless delivery truck spotted on the road in China is just a standard flatbed chassis with the cab sawed off and a sensor-and-compute pod bolted on the front. The company clearly eschewed a custom-designed vehicle platform with years-long redesign cycles, in favor of the fastest possible path from truck to robot. It’s not elegant, but it sure helped them ship fast.
That’s not to say there aren’t some trade-offs. The other side of the Xiaomi story that Western media never tells is that the veneer of quality is often surface level. Reports abound on Chinese social media about build quality, user experience details, creaks and rattles, and fit and finish problems that leave a lot to be desired, alongside marketing that tends to advertise features that never existed in the first place. Fewer layers between engineers and whoever signs off, along with less physical prototyping in favor of virtual validation, means that the products might ship imperfectly and get fixed later, or might not even ship in its advertised form. It trades certainty for speed, which might not necessarily be what you want for self driving.
However, none of this means Chinese AV hardware and software is inherently less safe, or that Waymo’s caution is wasted effort, just that the two industries are optimizing for different things. I think this is extremely important to recognize in any conversation comparing China’s developments against that of Western countries. China is optimizing for iteration speed and market share, while most of the West is optimizing for a safety case it can defend to regulators and skeptical city halls one incident at a time. So the next time somebody starts talking about how any one country is “winning,” remember: It’s nuanced.
International News
Xpeng has completed European localization testing of its second-generation Vision-Language-Action driving model in Germany, clearing the last milestone toward bringing its smart-driving stack from China to global markets in 2027. VLA 2.0 has been live in China for about six months, and the company’s smart-driving business head Liu Xianming says assisted-driving mileage share has crossed 50%, the first time Xpeng has seen its system surpass human driving at scale. The same model, trained entirely on Chinese roads, is handling German arterial roads, narrow alleys, and tight turns with what Xpeng describes as “almost no additional local data training”, while also picking up European road signs and right-of-way rules.
Xpeng completed the local acceptance test in Germany on July 14, two days ahead of Xpeng’s global launch of the Mona L03 in Munich, badged the XPENG L03 outside China. It’s the company’s first simultaneous global launch in its 12-year history, planned for 64 markets. Top trims run Xpeng’s in-house Turing chip at up to 1,500 TOPS and support the full VLA 2.0 stack. Xpeng has also pushed the same model to its L4 robotaxi fleet, which started internal testing this month.
VinFast has begun pilot testing “Level 2++” driver assistance on its VF 8 and VF 9 SUVs in Hanoi, built with Israeli AI firm Autobrains under a partnership the two companies announced earlier this year. The pilot runs alongside a second, longer-term concept called Robo-Car, which drops lidar and HD mapping entirely in favor of seven production-grade cameras and a compact chip processing roughly 20 trillion operations a second. Both systems lean on Autobrains’ Agentic AI paired with what the company calls Air-to-Road localization, fusing live camera feeds with satellite imagery instead of pre-built maps.
VinFast’s current EVs ship with basic L2 ADAS (adaptive cruise, lane centering, automatic emergency braking), while Robo-Car is still confined to controlled zones in Hanoi. Chaotic, mixed-traffic Hanoi is a tougher proving ground than most structured U.S. test cities, so success there could translate better to other emerging markets than another few million miles in Phoenix or San Francisco.
Alright, that’s it from me… until next week. If you enjoy this newsletter, share it with your friend, colleague, or boss. Thank you for reading; Sophia out!
So much excellent content here (again). I could comment (again) but your words are a better read. Appreciate your work.