How Far Along Are Waymo, MOIA, Tesla & Co. With Autonomous Driving? A Few Rules of Thumb for Classification

Car companies and robotaxi developers keep announcing that they are about to launch their own robotaxi service, that they are catching up with leading companies such as Waymo, and that this will catapult a region or an entire country to the forefront of this topic. Most media are reporting well, people are excited, but how much truth is really behind these announcements?

Not much! My observations in this sector, which for me personally began in October 2010 with a report in the New York Times in which the newspaper presented Google’s developments in autonomous cars, have taught me a few things. At that time, the Google prototypes had already driven 140,000 miles (224,000 kilometers) on public roads. On April 29, 2013, I saw the first Google vehicle with my own eyes on Highway 101 south of San Francisco near Menlo Park – and immediately made a video of it. The report and the first encounter are signals – as they say in Foresight, the art of forecasting the future – that show how far the development of a technology really is and where it could lead.

To date, I have captured almost all generations of Google and later Waymo vehicles in over 300 small videos – signals – like the one above. And another 1,800 such videos from over 70 different manufacturers. I should mention that I have lived in the San Francisco Bay Area, better known as Silicon Valley, since 2001. California had the most approved autonomous test vehicles and manufacturers in the world, so it’s the epicenter of autonomous driving. At its peak in 2019, the California DMV website listed more than 70 companies that had received a license in California to test autonomous development vehicles on public roads. Today, in mid-2025, there are “only” around 30 licensees.

Back then, I mainly filmed passing cars, but for the past three years I have also been able to record videos of rides in vehicles from around a dozen manufacturers, including Tesla FSD on my own Model 3. I took my first ride in a driverless Cruise in the summer of 2022, and my first ride in a Waymo six months later. And since then, I have been able to complete more than 280 such rides, taking around 500 guests (mostly from Europe) to give them this experience. So it took two and a half years from first reading about autonomous cars in October 2010, to the first encounter, to the first ride in another nine years.

These time periods are important because they give an initial indication of how much effort is required to develop autonomous driving and roughly how far along a manufacturer is with their efforts and how serious they are about it. It should be noted that manufacturers started from scratch back then. In the meantime, an entire ecosystem has developed that can provide technologies, data and experience, which drastically shortens the development process. Nevertheless, development is not a walk in the park and the development time is still long.

Rules of Thumb

Over the years, I have developed a few rules of thumb based on my long observation and the detection of small signals. By this I mean encounters with such test vehicles, how often I see them, where they drive around, whether with or without a driver and so on. My experiences are not limited to the SF Bay Area. I’ve also encountered and driven such vehicles in Los Angeles, Las Vegas and Brno.

By autonomous driving I mean L4/L5 here, by the way, whereby I do not differentiate whether it is legally permitted or not. For example, the Tesla FSD is legally an L2 system, functionally it is already in many parts or aims at an L4 system. In general, the SAE level system for automated/autonomous driving is problematic, because some systems, such as the Intelligent Park Pilot and the Intelligent Drive Pilot from Mercedes count as L3 or even L4 and have the legal approval for it, but should not actually be allowed to bear this title (here and here).

Also: I speak German (my native tongue), Englisch, French and Russian. So I scout media in multiple languages and are highly aware of what is reported in those languages on this topic.

Before we get started, I will focus on the rules of thumb and not analyze the individual companies. I will only use the MOIA example to show how this is applied. Tesla, on the other hand, requires other factors to be taken into account, on which I have already written articles in the past. On the one hand, there’s the repeatedly played up, but actually not so important factor of Tesla’s camera-only versus camera/lidar/radar combination for most others (article here). On the other hand, the essential part that Tesla is pursuing a much broader and therefore more protracted approach. Keyword: “boiling the ocean” (article here).

Rules of Thumb 1

Don’t trust the press releases, media reports or patriotic rallies.

Press reports usually only show an embellished, optimistic view of how good their own company is at autonomous driving. The media often take these reports uncritically because they themselves have little practical experience with this technology and are unable to assess how the technologies differ and how far the individual manufacturers have progressed. Many bloggers or posters on social media are often overcome with patriotic feelings, more than happy that their own domestic manufacturers would now also be at the forefront and that they would now catch up with or even overtake the leading foreign company (to date this is Google’s sister company Waymo).

Rules of Thumb 2

Question: Where do you encounter autonomous test vehicles?

Do you see them on the highway, in cities, in suburbs or in rural areas? My first encounters with them were on the highway for years. Testing is easier there because there is no oncoming traffic, pedestrians, intersections or traffic lights. Although everything takes place at higher speeds, test drives often begin – after the first thousands of kilometers on a test track – on the highway.

Only after that do you see first isolated, then more and more vehicles in suburbs or cities. This means that much more complex scenarios have now begun to be tested here. Cities and suburbs are much more chaotic and more complex to record. Incidentally, one of the most complex places to operate an autonomous car is in shopping center parking lots, because there are no longer any traffic rules there, or no one follows them, and cars wait, park in and out, and many, many pedestrians are out and about pushing shopping in front of them.

Note: Waymo put around 40 of the Google Firefly vehicles on the road between 2014 and 2017, which were then used exclusively in suburbs and cities. They were replaced by several hundred Fiat-Chrysler Pacifica minivans.

Rules of Thumb 3

Question: How often do you see the vehicles?

Do the encounters on public roads happen rarely or are they becoming more frequent? Are they always the same or, at best, a handful of vehicles with the same license plates or names (Cruise vehicles had names)? Or are they so numerous that, after initial enthusiasm, you no longer even notice them because they have become a normal part of the traffic scene?

Note: Frequent encounters over a longer period of time in urban areas indicate serious efforts and progress, but not always. The best example of this was Uber ATG, which drove like hell, ignoring blatant safety standards that ultimately led to a fatal accident and the end of Uber ATG.

Rules of Thumb 4

Question: Are there safety drivers in the vehicles?

Do you see these vehicles with a safety driver behind the wheel or not? Are there other people in the car who don’t look like they are performing technical tasks, such as operating a laptop on the passenger seat or hanging cables down somewhere? Do the other people appear to be passengers, whether they are employees, who are often the first to be required to use the vehicles for business and private purposes, or external, non-company persons?

Note: Waymo carried out the first driverless drive on public roads in 2015 in a Firefly. It was not until 2018/19 that the first driverless Fiat-Chrysler Pacifica minivans were spotted on public roads in Phoenix. It also took Cruise five years (from 2013 to 2018) for the first driverless Cruise to drive on public roads. Before that, dozens of vehicles with safety drivers had been on the road in San Francisco.

Rules of Thumb 5

Question: How many videos of encounters and experiences with autonomous cars have been published?

The more vehicles test drive on the roads, the more likely it is that they will attract the attention of other road users and be photographed or filmed by them. These videos often end up on social media and various forums dedicated to autonomous cars, such as on reddit or Facebook, share and discuss these videos very frequently.

These videos should not be confused with videos published by journalists or the manufacturer itself. These are highly curated and edited, and often only appear at one or more points in time with long breaks in between. Videos published by other road users appear continuously if the manufacturer is continuously testing, sometimes with more intense moments if that manufacturer has intensified efforts.

The popular argument that American companies are good at marketing and that German companies, for example, prefer to work quietly but then have great technology to show for it – in this specific case, a lot of testing on test tracks and not so much on public roads – is unfortunately not true at all. Just as you can’t develop a perfect airplane exclusively in a wind tunnel, but have to fly it in the real world, autonomous cars have to be tested and developed in the real world. There are too many marginal scenarios that could not be tested on the test track (or in the simulator), nor could they be foreseen. In order to develop good and safe autonomous cars, millions of kilometers have to be driven on public roads under various conditions.

This has also had an impact on my blog, where I have been reporting on such encounters since December 2015. Through my global readership, I keep receiving photos and videos of encounters that my readers have had in their home countries and send them to me. For me, this is also a signal of how intensive (or mostly less intensive) the development activities of local companies are.

Note: on LinkedIn, for example, several testimonials from other LinkedIn users about their first Waymo ride are flushed into my timeline every day. I also see Zoox again and again, as well as the Tesla FSD, which always leads to controversial discussions. Occasionally, I see ride reports from other manufacturers such as Pony.AI, Kodiak or Aurora and Chinese manufacturers such as Baidu/Apollo, WeRide or AutoX.

Rules of Thumb 6

Question: Who is developing the self-driving technology?

If you look at Waymo, Zoox or Tesla, you can see that the main components are being developed under the leadership of these companies.

	Vehicle	Sensors	Self Driving Software	MaaS
Waymo	No	Yes	Yes	Yes
Zoox	Yes	No	Yes	Yes
Tesla	Yes	No	Yes	Yes
Cruise	Yes	No	Yes	Yes

Waymo develops both the sensors and the software that steers the vehicles through traffic. Waymo has also developed its own software in the background for simulations and the operation of a robotaxi service.

Although Zoox has not developed the sensors, it has developed its own vehicle and the self-driving software.

Tesla has developed everything except the sensors.

Cruise received the vehicle platform from its parent company GM, but also developed its own (Cruise Origin), as well as the software.

These examples show that none of the companies leave the development of the central element – the self-driving software – to others, but instead take the lead in developing and integrating it themselves. So as soon as a company outsources the development of this component, any claims about the state of the technology and timetables should be viewed with caution.

Rules of Thumb 7

Question: How is the company set up and financed?

I pay attention to the parent company – Alphabet Group with Waymo, Amazon with Zoox – and their experience with and willingness and ability to finance the development of long-term software projects. Also whether these companies are run by founders or by managers. Managers are no more stupid than entrepreneurs, but they are evaluated differently and their incentives are different from those of entrepreneurs and founders.

The founders of both Waymo and Amazon still have important roles, either as board members or as supervisory board members. Both are very experienced in the development of large software projects and have the financial strength to finance them for years.

This was no longer the case with Cruise. There, the parent company General Motor is run by a management board for a few years (usually four years), which also had a different business model in mind – keyword: selling cars to private individuals, but not operating a robotaxi service. For GM, which was already struggling with the transition from internal combustion vehicles to electric cars, the cost of $10 billion was no longer sustainable after 10 years. An incident was then the catalyst for GM to discontinue Cruise Automation.

Rules of Thumb 8

Question: How well does the company communicate about the development?

The public wants to know how things are progressing. To this end, companies should also provide regular updates that address various topics. A prime example is Waymo’s Waypoint blog. There, the company publishes data on collisions, safety aspects, where new test vehicles are driving in order to map the region in terms of navigation technology, reports from passengers, technical details about the vehicles, the simulators and so on. We should also find plenty of videos there, as well as on their social media accounts. How active are the companies on LinkedIn, Twitter, Bluesky, Facebook? What do they share there? Waymo is exemplary here too.

A company that tries to prevent the publication of data on security aspects is already suspicious. If nothing happens on social media channels for a long time, then there is probably no real progress to report.

Rule of Thumb 9

Question: Does the company already have a driverless commercial robotaxi service in operation and is it testing in other regions?

If the answer is yes, then it’s just a question of scaling up – and that’s hard enough. But this company has not only mastered the technology, it has also overcome the legal hurdles, set up infrastructure such as operations centers and depots, and secured the support of various interest groups (such as associations for the blind, women’s groups, business Associations, etc.).

Rule of Thumb 10

Question: Can I take a ride myself?

How easy do the manufacturers make it to ride along in such an autonomous car? Do you have to go through the press office, talk to people you know at these companies, or do they already have a waiting list or are they even open to the public? If you are constantly turned away or told that there are currently no options, then there is probably nothing that would allow the company to make a reasonably good impression.

If you can then go for a ride – with or without an attendant from the company – the question arises as to the frequency of interventions by a safety driver, the driving experience of the autonomous system itself, and how often a teleoperator or teleassistant needs to intervene in situations in a driverless car. Also, how pleasant is the vehicle to drive? Is it hesitant, brisk, confident, does it drive very comfortably or more jerkily and does it resemble a wild ride?

Analysis Using the Example of MOIA

Last week, the VW subsidiary MOIA presented the jointly developed robotaxi, the ID.Buzz AD, to the public. This vehicle is set to go into production from the end of 2026, with a production capacity of 10,000 units per year.

MOIA aims to offer transportation companies and transport service providers such as Uber a complete package that they can then seamlessly incorporate into their services. In other words, a passenger vehicle with all the necessary sensors (ID.Buzz AD), the self-driving technology software and a whole stack of Mobility-as-a-Service.

I have already described the challenge for MOIA here, but the rules of thumb I have listed make it easier to understand, why I have some serious questions about the claims and schedule. The vehicle with sensors won’t help public transport companies if the self-driving software isn’t functional – and that in the long term. This is because there are hardly any public videos of autonomous driving on public roads, even from the partner Mobileye. The integration test between ID.Buzz AD and self-driving software requires massive testing on public roads, and with passengers. The fact that other road users have not yet published such videos of either MOIA or Mobileye, and that these vehicles are therefore not present and have not driven hundreds of thousands of kilometers in chaotic traffic in German cities or Los Angeles, shows that they are still a long way from being ready.

What will hopefully be included in 2026 are test drives in fleet operation with safety drivers. And these journeys will take place in this form for some time. It is likely to be 2-3 years before the first MOIAs are actually driverless. So we are talking more about a reasonable use, i.e. without drivers, for Uber or public transport providers from 2029/30. And MOIA probably doesn’t have that under control. Whether the software is ready depends on Mobileye, provided they don’t jump ship. In view of the difficulties in which Mobileye’s parent company Intel finds itself, I don’t expect the necessary progress from them. Mobileye seems to be or become a walking dead like Intel. And Volkswagen will have to save money at the latest with a change of CEO and the probably unavoidable purchase of Rivian. It stands to reason that the red pencil will be applied to MOIA. VW has already dropped Argo, or GM Cruise.

Potential business customers of VW/MOIA must expect to provide drivers for these cars in the long term, and drivers are in short supply. Everyone is desperately looking for them. And all this only if VW does not draw the line.

Note: my intention is not to badmouth MOIA or others. Quite the opposite: I wish them success and that their years of hard work will pay off. But behind all the euphoria in the media and the blogosphere, I see things differently here based on my rules of thumb and would like to call for honesty. Somehow, certain aspects and information are missing in this reporting that make the announcements seem feasible to me. What I see and what is said somehow doesn’t fit together at the back and front.

From my years of observing the scene, I don’t see VW/MOIA soon being able to present a ready-to-use and usable solution for public transport companies or transportation service providers in view of these facts. The fear is that, as with the tests with the “autonomous shuttles” from providers such as Navya, Easy Mile or Olli, they will once again burn their fingers if they believe the assurances that a turnkey solution could be ready by 2027.

Conclusion

With these rules of thumb, I try to evaluate how far along a manufacturer really is and when you can reasonably expect a robotaxi service to be launched. Of course, these time frames have become shorter over the years. Companies like Waymo or Cruise had to start from scratch. In the meantime, an entire technology and data infrastructure with experienced specialists has developed, which even smaller start-ups and manufacturers can rely on.

However, it is possible to make at least a rough assessment of how far the manufacturers’ claims about autonomous driving really are from the reality on the roads. A manufacturer that does not already have large numbers of autonomous cars with safety drivers in test operation on public roads will barely be able to start a driverless, commercial operation of robotaxis (or private cars) before 2030.

This article was also published in German.