The Job Claw Machine
Brian Stever
AI Job Claw · 2026 · p5.js, generative animation, labor data
Abstract. AI Job Claw turns labor-market anxiety into a retro arcade machine. The prizes are jobs. The claw patrols automatically, or the viewer can take control and drop it themselves. JOLTS job-openings data controls how crowded the machine feels, which makes the piece part toy, part chart, and part accusation. It is technically playable, in the way a bad dream is technically a narrative.
1.The Piece
Interactive sketch
AI Job Claw
Move the pointer over the canvas to control the claw. Click or press space to drop it.
2.The Metaphor
The sketch is built like a little arcade cabinet. Inside it, jobs are drawn as tiny anxious characters: Writer, Coder, Analyst, Teacher, Driver, Artist, and others. The machine keeps running whether the viewer participates or not.
That is where the metaphor gets sticky. AI-driven productivity is usually sold as abstract efficiency, with all the sharp edges hidden behind a nice dashboard. A claw machine makes the abstraction embarrassingly physical: something gets picked up, removed, and counted as progress.
3.Data Layer
The number of characters in the machine is mapped from JOLTS total nonfarm job openings. In the class sketch, the values are hardcoded from 2022 through early 2026 so the animation can run without a backend.
It is not an economic model, and it should not be allowed near a central bank. It is a visual argument built on a real signal: when job openings shrink, the crowded prize pit thins out. The machine does not explain the economy. It makes the trend feel less bloodless.
4.Character Behavior
The first version felt too clean. The characters existed, but they did not seem to care. Then came the small behavioral variables: panic, bravery, gaze direction, blinking, cowering, and reaction to nearby removals.
That changed the whole piece. The jobs stopped feeling like labels and started feeling like little simulated bodies under a machine. It is a tiny bit silly and much more uncomfortable, which is precisely the tonal pocket where the joke starts looking nervous.
5.Reflection
The sketch can be played wrong, which is usually a good sign. Leave the claw alone and the automatic patrol continues. Take control and the viewer becomes part of the system. Either way, the machine keeps doing what machines with incentives tend to do.
That feels close to how AI adoption often lands in daily life. Nobody has to be a cartoon villain for the outcome to feel brutal. Sometimes the most unsettling thing is a cheerful interface asking you to play along.