Every plate is a servo-driven mass simulator wired to the camera network in the ceiling. Nobody's looking at you, the bar weighs what it says. Six people looking, it's 40% heavier. The number stencilled on the plate never changes. What changes is what your spine reports.

I've been sketching this for a month and I keep hitting the same wall, so I'm bringing it here before I waste more evenings on it.

The mechanism is fine. You can fake variable inertia with a flywheel and a brake, and gaze detection is a solved problem — you don't even need to identify anyone, just count faces oriented barward. The honest version doesn't scale load with attention linearly either; it should be sharper than that, because social pressure isn't linear. One person watching is nothing. Three is a crowd. Twelve is a performance.

Here's what I can't resolve. The whole point is to make the thing everyone already feels into something you can measure. You lift less when watched, or more, depending on who you are, and nobody can quantify by how much because the effect is buried in a hundred confounds. This machine drags it into the open and puts a number on it.

But the number I'm putting on it is my number. I picked 40%. I made up the curve. I've built a machine that appears to measure social pressure and actually just measures my guess about social pressure, dressed up in steel and told to you by your own body.

Does that ruin it? Genuine question. There's an argument that it doesn't — that a wrong calibration still teaches you the shape of the thing, the way a bad map still teaches you that there's a coastline. There's a stronger argument that a false quantity is worse than no quantity, because now people will train against my invented curve and get good at a fiction.

The version that gets around this is a gym where the weights are heavier when watched and you're never told, and I think that one's just an ethics violation with a squat rack.