One prompt, six titles, and nobody in the end credits

On X, klöss gave Claude Opus 5.5 a job: make a 90-second motion graphics piece with only one goal—land the job of his dreams. He posted the prompts verbatim, revising them across three versions.

The first version was posted on October 30, with the art direction set in Neo-brutalist Swiss editorial design: 1-bit dithering, pixel noise, modular grids, ultra-bold neo-grotesque type, monospaced technical annotations, and Y2K digital interfaces. In the frame, a dot matrix scatters outward from the center; the top-left reads "00 / SIGNAL," and the bottom-right has a line reading "PIXELS DECIDED 8,420,617 / 2,073,600."

The next day he reposted it with a different visual language: Y3K liquid chrome and mercury surfaces, oil-slick iridescent reflections, inflatable 3D blob type, wireframe holographic UI, caustic light refraction, glossy product rendering, extreme negative space, graded in HDR with highlights exceeding paper white. Delivery specs were listed as 1080p 10-bit HDR HEVC, Rec.2020.

In a reply, he said he had remixed the prompt for a third version:

All three prompts have the same structure, and that's the point.

It starts by laying out the identities: motion designer, animator, artist, sound engineer, video editor, creative director—six roles. Then comes a long list of art keywords that locks down materials, lighting, type, and composition. Next is sound: explicitly no synth tones or cheap generated audio, only organic, textured sounds—poured water and mercury drips, glass resonance, struck metal bowls, slow breathing, fabric friction. Finally, delivery parameters.

On the surface, this thing is called a prompt, but it reads more like a brief. The difference is that a brief spells out what it doesn't want. It says no synth sound and lists five specific material sounds—this is the working language of a sound recordist. It says HDR, 10-bit, HEVC, Rec.2020; most clients can't explain the difference between these terms.

Someone in the comments did the math. Cheftoshi said that six job titles were written into one prompt, and there isn't a single person in the end credits. Luis Junior wrote in Portuguese that what comes out of one prompt used to be the work of an entire team—motion designer, editor, sound effects, art director. People who say AI isn't taking jobs probably have never quoted a price for a 90-second video.

There were also those who weren't buying it. Tina said that 90 seconds is where adjectives start charging interest, and the first thing to give it away will be shot continuity. Petra noticed something else: whenever you ask Claude to add sound effects to a video, it always sneaks in bells and chimes—this has already become a new AI signature.

These comments are more interesting than the video itself. The first two are about cost; the last two are about the tells.

The model really can generate 90 seconds in one shot, and the image quality, materials, and sense of light are above commercial standard. But 90 seconds doesn't equal a film. Every shot being beautiful doesn't mean there is breathing room, rhythm, or narrative progression between shots. This is a problem long prompts cannot currently solve: they can dictate what every frame looks like, but not the relationship between frames.

The same goes for the sound effects. Even though the prompt explicitly specifies concrete material sounds, the model still falls back on habitual conventions at certain points. Bells and chimes are the high-frequency answer for a premium feel in the training data; it remembered the answer but not the reason.

So the barrier to this piece isn't on the model's side. The person who can write this prompt already knows what materials, lighting, sound, and delivery specs a 90-second motion graphics piece should have. He had the film in his head first; only then could he write this checklist. What the model does is execute what's in his head, much faster.

What the model amplifies is judgment. The execution part was already being compressed.

There's also a small detail. klöss himself said in a reply, "one shotting this is nuts." One shot. Three versions of the prompt, three one-shots. A year ago, that claim would have been treated as bragging.

If someone actually used this piece to apply for a job, the interviewer would probably ask: Did you make this? There is still no standard answer to that question.