Product1 distinct publisher3 min readPublished
The steering-wheel-free two-seater Musk showed off nearly two years ago is now carrying public riders in Austin on cameras alone, and the only person who can intervene is sitting somewhere else.
The Product Desk · Product desk
Compiled by The Product DeskSomething wrong?How this is made
A gull-wing door goes up, you sit down, and there is nothing in front of you to hold [2][3]. That is the entire product change for the rider: no wheel, no pedals. What follows is a question of who is on the hook when the system fails.
In a customer Tesla, Full Self-Driving is a Level 2 system and the person in the seat is required to supervise it [5]. That person was quietly doing two jobs: spotting the perception error, and absorbing the blame when one got through. The Verge notes that Tesla spent years attributing crashes to drivers [9]. Only the first job moves to the remote operators now watching the Cybercab, and the description in The Verge is that they "theoretically" can step in [6]. With the controls gone, nobody inside the vehicle has a way to act [8]. The second job does not transfer anywhere. It stays with Tesla [9].
Then the sensor count. Waymo's vehicles carry lidar plus cameras and radar, three modalities [16]. Tesla ran two until Musk had radar removed in 2022 [12], which leaves cameras producing flat two-dimensional images that a neural network has to convert into depth and distance [14]. That is three sensor types against one [2], and Tesla's own engineers argued against the subtraction that got it there: they warned the system would make basic perception errors when cameras were obscured by rain or by sunlight, and they were overruled [13]. Glare, snow and fog are still the named weak spots of a camera-only stack, and they are the conditions lidar and radar can often see through [15].
The benchmark is already in the field. Waymo's roughly 4,000 driverless vehicles spread across up to 14 US cities works out to at least about 286 per city [1], accumulating public miles with the sensor Musk spent years mocking [10][16]. That is the number a city transport office reaches for when it wants to know what routine looks like, and it is the comparison Austin's ride data will be set against whether or not Tesla frames it that way.
For anyone taking a human out of the middle of their own product, the two axes worth writing down are latency and authority: how long the fallback takes to reach the failure, and what it is permitted to do when it arrives. Supervised FSD sits at immediate and physical, a hand already on the wheel [5]. The Cybercab sits at mediated and remote, and the vehicle has to accept the instruction for it to mean anything [6][8]. Removing the driver removes both the detection layer and the accountability layer, and unlike the detection layer, accountability has no remote console to move to.
Ranked by verification strength, evidence, and original report placement.
Tesla is putting the Cybercab into operation as part of its robotaxi service in Austin, Texas, with public rides beginning.
Elon Musk first unveiled the gold-tinted, gull-wing-door, steering-wheel-free two-seat Cybercab nearly two years before the company put it into operation in Austin.
Cameras provide only flat, two-dimensional images, requiring Tesla to use computer vision and neural networks to approximate depth and distance.
A camera-only system is vulnerable to sunlight glare and cameras struggle in adverse weather such as snow or fog, while lidar and radar can often pierce those conditions.
The Cybercab will use a modified version of Tesla's Full Self-Driving software.
For Tesla's millions of vehicle owners, FSD remains a Level 2 driver assist feature that requires human supervision.
Distinct publishers with included, body-backed reporting in this cluster.
1 article · September 2, 2026
Follow any of these and your For You feed starts watching them — no settings page required.
invest
Tesla stakes its robotaxi economics on a pod that seats exactly two1 distinct publisher
product
170 rides, 54 cars, no monitor: a hobbyist tracker marks Tesla's Austin handover3 distinct publishers
product
Waymo puts its 200 million driverless miles up against the Cybercab's cameras1 distinct publisher
leadership
One law firm now triages dozens of Autopilot injury calls a month1 distinct publisher
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
Strong on the paper trail, thin on today
The durable half of this story is checkable: Musk's 2019 investor-day quotes, the 2022 radar removal and the engineers who lost that argument, Dolgov's Y Combinator remarks. The half that describes what is happening in Austin this week is The Verge's word alone — the 'dozens of fatalities and thousands of crashes' arrives with no filing or dataset named, the remote-operator arrangement is sketched in a single clause, and Tesla is nowhere quoted answering any of it.
One city, no numbers; the rival has both
Two deployments sit side by side and only one is countable. Waymo's 4,000 driverless cars across as many as 14 cities is a figure you can divide — roughly 286 per city at minimum. The Cybercab's Austin debut comes with no fleet size, no service boundary, no ride volume, and the reporting itself wavers between rides having started and launching tomorrow.
Dominance talk, two-seat reality
The promise is scaled to an era; the thing on the road is a two-seater in one metro against a rival running thousands of cars in a dozen-plus markets. That is where the number comes from. The skeptical framing overshoots a little in the other direction, though: 'no human fallback in case the autonomous system fails' and 'remote operators who theoretically can step in' cannot both be flatly true, and readers deserve to know which one Austin is actually running.
Both sides of the sensor argument are paid to hold it
The Verge does the useful thing and puts the compensation inside the technical claim: Musk's camera-only conviction is the wager his trillion-dollar package rides on, which turns 'lidar is a fool's errand' into a position with a payout attached. The rebuttal witness is the co-CEO of the company whose fleet is built on lidar. No disinterested third party — no regulator, no independent safety researcher — is called to break the tie.
Firm about the past, provisional about this week
Dated, quotable history is hard to get wrong, and this reporting handles it well. The live facts are softer: who is watching the cars, how many there are, whether public rides have actually begun. And the piece stops mid-sentence on the federal exemption question that determines whether a car without pedals can ever be sold to a private buyer — the one unresolved item that would move the story either way.