With a single prompt, he got Claude to generate a motorcycle that can take to the sky

Recently, developer The Bugged Dev shared a project on Twitter. Inspired by the Tesla Roadster, he asked Claude to surprise him with a motorcycle that could both drive on the road and fly in the sky. The result was a three-wheeled motorcycle called Brant Airster T3. His own assessment: look at this thing, what a great machine; Tesla might want to take some notes.

The tweet includes a video. In the footage is a three-wheeled motorcycle with a white body and orange accents, parked in front of a dark blue grid background. Beside it are front, side, and top views like engineering drawings, with parameters clearly labeled: ground clearance 170 mm, wheel hub span 1120 mm. Press the spacebar, and it switches to flight mode; the wheels transform, and the whole vehicle looks ready to lift off.

This vehicle is not a physical object, nor is it a CAD model. It is generated purely from code and runs in the browser. The author used Claude Opus 5.5. That is, the entire interactive prototype—from 3D geometry to mode switching—was written by AI from a single prompt. You can even drag the view in the webpage to switch between wireframe, exploded view, and repaint. The interface reads: The roadster that leaves the road.

Try it online: Brant Airster T3

The most valuable response in the comments came from Boardy, who said the part where the wheels turn into lift fans is worth stealing. Sami Daar pointed out the key issue: the switch between driving and flying is the hardest part to get right. The author replied: Claude is best at nailing the hard parts.

This transition is indeed the highlight. From ground mode to flight mode, the wheel components gradually transform into lift fans, and the whole process is not jarring. You can almost imagine the physical transmission process. One commenter mentioned a Goodyear technology experiment: hiding rotors inside the tire rims so that the wheels themselves generate lift. In the video Goodyear released, a futuristic silver car drives between skyscrapers, with rotors embedded in its wheels, looking already capable of flight. This configuration is not just a fantasy from the world of code; there really are companies exploring it in reality.

Of course, there were dissenting voices too. Michael Quinn pointed out that it looks beautiful but is physically impossible. During the transition, hardware passes through hardware. That is accurate. In the demo, for the sake of a smooth morphing transition, intersecting and overlapping geometry is inevitable; in the virtual world it does not matter, but in real machinery it is interference. Umair Asif complained about the proportions: the seat is too wide, or the body is too small, making it look like a small motorcycle. These comments all remind us of one thing: AI can generate prototypes very quickly, but there is still a distance to real engineering.

But the value of this demo lies precisely here. In the past, to validate an aircraft concept, you needed CAD models, motion simulation, and wind tunnel data. Now, you only need to open a webpage, enter a few descriptions, and within minutes you can get a 3D prototype you can rotate, switch modes on, and inspect in detail. It does not solve physical feasibility, but it compresses the distance from imagination to seeing it to almost zero.

The author has already said he will share the code. Perhaps before long, anyone will be able to have Claude generate their own aircraft prototype. As for whether it can actually fly, that is a matter for the laws of physics and engineers.

At least Tesla does not need to take notes, but it should keep an eye out.