A founder's note
Why autonomy.
Roughly nine in ten road accidents are caused by human error. The machines do not have to be perfect to be better, they only have to be better than us. That is the bet this publication is built on.
A useful frame
Tim Urban's post is a good reminder that autonomy is not only a car feature. It is part of a wider shift in how machines take over work humans are not especially good at.
About 90 to 95 percent of traffic accidents are caused by human error: distraction, fatigue, aggression, alcohol, a single moment of inattention. Autonomous systems remove those weaknesses by design. Tesla, Waymo and Zoox have every commercial reason to make their systems safe. A serious incident damages the brand, invites lawsuits, and erodes public trust. Safety is not a side concern for these companies, it is the precondition for the whole product existing.
The real argument is human vs machine
People driving themselves are, statistically, far more dangerous than well-designed autonomous systems. The interesting question is philosophical: do we want a system that causes one accident per hundred kilometres, or do we keep humans who cause five?
On top of that there are technical differences worth respecting. Tesla bets on a camera-only stack, closer in spirit to the human eye. Waymo and Zoox combine lidar with cameras for redundancy. Both approaches have trade-offs in reliability, cost, and performance across conditions. We do not yet know which one wins, and the engineering honest answer is that we will not know for a few more years.
We are not autonomy engineers. That is exactly why the conversation needs nuance and a little humility about what "safe enough" actually means, and about how to accelerate good systems without raising unnecessary barriers. Regulation that is too slow or too strict can hold back genuinely better technology for years, while the companies building it already have every reason to get it right.
The levels, briefly
The industry uses the SAE J3016 scale, zero to five.
- Level 0–2. Human drives, the car assists. Tesla FSD Supervised lives here.
- Level 3. The car drives under defined conditions, but the human must remain ready.
- Level 4. Fully autonomous inside an operational design domain.
- Level 5. Fully autonomous in any conditions a human can drive in.
The players I keep watching
Tesla FSD Supervised was recently approved by the RDW in the Netherlands. Officially Level 2+. Cameras and neural networks, supervised driving, with the stated ambition to graduate to full autonomy.
Waymo already runs a Level 4 robotaxi service in multiple US cities, no driver behind the wheel, booked through an app. Lidar plus cameras plus radar. Years of operational miles.
Zoox, the Amazon subsidiary, is building a purpose-designed robotaxi with no steering wheel and no pedals. Bidirectional, facing-seat interior. In US testing.
What "autonomous" actually means
FSD Supervised (Level 2+) requires a human ready to take over at any moment. FSD Unsupervised, or ADS, means the car drives itself inside a defined operational design domain with no driver required. Today, real autonomy lives mostly in the robotaxi form factor. Tesla uses modified Model Y vehicles with extra camera cleaning and a Starlink tracker for the early CyberCab routes already running in Dallas. A small Early Access rollout of FSD Unsupervised on consumer Teslas is most likely a late-2026 or early-2027 event, with a clear ODD.
NHTSA data, not vibes
The best public source on real autonomous-system performance is the NHTSA. The distinction matters: ADS data (vehicles driving without a human, like Waymo or FSD Unsupervised) is reported separately from ADAS data (which includes FSD Supervised). Since June 2025, full incident reports are available. The raw CSV and the official Standing General Order dashboard are both public. If we are going to argue about autonomy in 2026, we should argue from data, not from anecdote.
What I want from this
Driving should not be a right. It should be the safest possible way to get from A to B, and that is also a kind of freedom. The Netherlands is not a bad place to drive, but the Nordics are better. And the technology to do better than any of us already exists, in prototype form, on public roads, today.
This publication exists to cover that transition with the seriousness it deserves. No retyped press releases, no marketing varnish, no cheerleading and no panic. Just the regulatory moves, the engineering choices, and the data, written in plain language by people willing to put their name on it.
Original Dutch thread: Autonome voertuigen: op weg naar 0 verkeersdoden on Tweakers Gathering of Tweakers, mobility forum.