What the object does
CrankGPT runs a local model on a Jetson Orin Nano, measures the energy consumed while the answer is generated, and then hides the answer behind a lock until the visitor hand-cranks that same amount of energy into the system.
The hardware and software are inseparable here. The project uses local inference, a measured power path, and a physical unlock loop to make the cost of “just ask the model” legible as an embodied interaction.
Why it belongs in Projects
This is not just a demo. It is a technical object used as a cultural argument. It treats interface, hardware, and computation as material and asks users to feel an otherwise invisible infrastructural cost.
It is also one of the clearest examples in the archive of critical AI work that is built rather than merely described.