During the first week of December 2025, Tesla's humanoid robot, christened Optimus, showed two very different faces. The first was that of a triumphant machine. The second was that of a machine collapsing in a heap. The most interesting thing is not the fall itself: it is that both images were produced at very nearly the same moment, and that they may well be telling the same story about stage management.
Monday, the robot that runs
On 3 December 2025, Tesla released a video showing Optimus running smoothly across a laboratory floor, both feet leaving the ground, according to the specialist site Interesting Engineering. Running, for a two-legged robot, is no small feat: it means managing balance with every stride, that instant when no foot is touching the ground at all.
Elon Musk, Tesla's boss, shared the clip himself on the social network X with a terse caption: Running robot, once again according to the same source.
The same video, Interesting Engineering reports, makes a point of the robot's improved balance and gait. In short, the picture of progress under control, filmed in conditions of its own choosing.
A few days later, the robot that falls
Cut to a different scene. During a public demonstration in Miami on 8 December 2025, Optimus was handing out bottles of water to those present... and then took a tumble, again according to Interesting Engineering. So far, so unremarkable: a robot falling over in public can happen, and there is nothing scandalous about it.
Except that one detail set the whole thing alight. As it fell, the robot reportedly brought its hands up to its face, in a gesture some observers read as that of a human teleoperator pulling off a virtual-reality headset, according to the same source. Once the word was uttered, the doubt set in: what if, behind Optimus, a human was piloting the machine from a distance?
So what exactly is teleoperation?
A word of explanation, to keep things straight. We talk about teleoperation when a person controls the robot remotely, rather like a high-tech puppet, often wearing a virtual-reality headset and sensors that relay their movements to the machine. An autonomous robot, by contrast, decides on its own movements thanks to its artificial intelligence. The difference is enormous: a teleoperated robot is impressive, but it proves nothing about whether it can fend for itself.
The awkward bit: Tesla had already sworn the opposite
And this is where the two images meet. Because it is not the first time the question has come up. According to Interesting Engineering, Elon Musk had already brushed aside earlier accusations of remote control over Optimus, insisting that a kung-fu demonstration was a matter of artificial intelligence and not teleoperation.
So we have, on one side, a polished clip in which the robot runs all by itself; on the other, a live fall in which a gesture looks, for all the world, exactly like that of a human pilot. And in the middle, an earlier denial from the boss. Interesting Engineering points out, in fact, that the 3 December running video, which trumpets better balance, sits in direct contrast with the footage of the Miami fall.
Why this matters to you
Let us hold on to the underlying point, beyond the anecdote. A video of a robot running, filmed in a laboratory, is a piece of communication: you choose the setting, the angle, the moment. A fall in public, by contrast, slips out of your control. When one and the same gesture, that of pulling off a headset, blurs the line between an autonomous robot and a remote-controlled one, the question is no longer whether Optimus fell. It is which parts of everything Tesla shows us actually work on their own.
To date, none of this rests on a verifiable, independent demonstration that clearly separates what is autonomous from what is piloted. It is precisely this murkiness that turns an ordinary fall into a real question. Until the proof arrives, the right instinct stays the same in the face of any spectacular robot video: ask who, or what, is really holding the controls.