Opinion: I Watched the Tesla Roadster Fly and I Don't Believe It Yet

Overnight (Australian time) Tesla finally revealed the second-generation Roadster, nine years after the prototype rolled out of a Semi trailer. The event was livestreamed from SpaceX's rocket test site in McGregor, Texas, and the headline act wasn't the car's styling or its sub-one-second 0-60 claim. It was a white Roadster driving up a C-shaped ramp, stopping fully inverted at the top, then leaving the ramp, rolling in mid-air and settling to the ground under its own cold-gas thrusters. Later it jumped a row of parked Teslas, and closed the show with a flip from a standing start.

The crowd, by every account, went nuts. Elon Musk's reaction was a two-word "Well, it worked."

I watched it. My gut reaction was not amazement. It was: that doesn't move the way a 1.5-plus tonne object should move.

The physics that bother me

I want to be clear about the limits here. I'm not a rocket engineer. What I have is a long time spent around cars and a reasonable intuition for how mass behaves. And a few things in that footage didn't match the intuition.

• The ramp climb was slow. Engadget's own liveblog called the speed up the ramp "surprisingly slow". Tesla's explanation is a suction system that holds the car to the surface, which is plausible on its face. But a car that is clinging to a wall by suction has to generate a seal strong enough to counter its full weight at the inverted point, and then release cleanly, instantly, to leave the ramp. The transition from "stuck to the structure" to "free in the air" looked frictionless to me. That is the moment I'd expect to look messiest, and it looked the cleanest.
• The hover looked too stable. Cold-gas thrusters are blunt instruments. They're great for nudging a spacecraft in a vacuum. Balancing a car-shaped object with a high, shifting centre of mass in atmosphere, with wind, with the mass of the car itself, is a control problem that even purpose-built VTOL drones solve with visible wobble and correction. The Roadster's descent looked like a video game.
• Where was the violence? Thrust capable of arresting the fall of a car should kick up dust, debris, visible condensation plumes and a lot of noise. The pre-event reporting said spectators would be kept several hundred yards back specifically because the thrusters were loud enough to damage hearing. On the feed, the landing looked gentle and the ground around it looked undisturbed.

Again: each of these might have a perfectly good engineering explanation. Maybe the control system really is that good. Maybe the thrusters are far more capable than I think. But my instincts say the footage shows an object reacting to a lot less force than a real car of that mass would need.

This should have been a public event

Think about what Tesla was claiming to have built: a road car that drives up a wall, hangs upside down, flies, rolls and lands under its own power. If that's real, it is one of the great engineering party tricks of the decade. It's the kind of thing you put in front of a stadium. You invite the motoring press to stand at the fence. You let a hundred thousand phones film it from a hundred thousand angles, because every one of those clips is free advertising and every one of them is proof.

Instead we got the opposite. A closed rocket test site in rural Texas. A hand-picked audience of reservation holders and friendly influencers, held several hundred yards back. One broadcast, produced end to end by the company making the claim, with cuts and camera changes at exactly the moments I'd most want to watch uninterrupted. No one in the car. No independent media at the ramp. No post-stunt walk-up to the landing zone.

Everything about the staging reads as a production, not a demonstration. It's too polished and too closed for what it's supposed to be. A company confident its car can fly would want the world as close as possible. A company that needs the footage to look a certain way would want exactly this.

I'm not accusing anyone of anything. I'm saying the format chosen is the format you'd choose if you wanted control over what was seen, and that choice is itself a reason to ask for more.

The evidence problem

Here's the part I can't get past.

Everything the world has seen came off Tesla's own livestream, produced by Tesla, shot from Tesla's cameras, at a closed SpaceX facility, with the audience deliberately positioned a long way from the action. The car was remotely operated with "Starman" in the seat, which Musk was careful to say "is not a real human".

Tesla has spent years producing demonstration videos that were later shown to be staged or heavily edited, from the 2016 Autopilot clip onward. That's not a conspiracy theory, that's the public record. It means this company does not get the benefit of the doubt from me on a single-feed stunt.

So, Tesla: show me the evidence. What would change my mind is not complicated:

• Multiple independent angles. Phone footage from people in the crowd, shot at the same moment, that lines up with the broadcast.
• Footage of the ground after the landing. Scorch marks, blown dust, anything that says a lot of gas was pushed into the dirt.
• An uncut wide shot. One camera, no cuts, from the ramp approach to the car sitting still on the ground.
• A third party in the car's airspace. Any journalist or engineer who was close enough to feel it and is willing to say so on the record.
• Do it again, somewhere open. A public venue, daylight, press at the barrier. If the car can do it once at McGregor it can do it twice.

Some of that may already exist. If it does, I'll update this piece and say so plainly. Until then, "it looked incredible on the stream" is not the same thing as "it happened the way the stream showed".

Why this matters for a marketplace

MotorLoop exists because people are about to spend real money on cars, and the single most useful thing we can do is help them tell marketing from reality. Tesla opened Roadster reservations on the night. The suction-and-thruster version is listed at US$250,000, and the version that actually hovers is invite-only. People are being asked to put money down against a demonstration that nobody outside Tesla has been able to independently verify.

That alone is reason enough to hold the line. Extraordinary claims need ordinary evidence: a second camera.

Where I land

I'm not saying it's fake. I'm saying I'm not convinced it's real, and that's a different thing. The physics instinct says "that's not how a car behaves", the staging says "this was a production", and the sourcing says "you've only seen what Tesla wanted you to see". All three have to be answered before I'll call it the engineering moment everybody else seems to have decided it was.

Show me the angles. I'd honestly love to be wrong.