A color picker that stores no colors. The surface z = x·y, a saddle, sits inside a wireframe cube. A color is read off the surface by a lens, computed from where you are and which face you look through. Nothing is kept but the surface and the lens recipe. Drag to turn it. Click the surface to pick a color.
face: +Z (top)
Picked color (derived, not stored)
#000000
complement
your color·
hex·
rgba()·
hex + alpha·
alpha (from |z|)·
where (x, y)·
height z = x·y·
slope area·
through face·
The lens (the only knobs)
0
1.0
1.0
1.0
1.0
These reshape the reading. The colors and the alpha still come off the surface; the lens never stores one.
What is actually stored
the surfacez = x·y (a formula)
the lens recipe·
total stored·
colors reachable4,294,967,296 (32-bit RGBA)
colors stored0
vs a stored RGBA LUT·
The recipe: a color, read off the surface
This is the whole lens. Give it a place on the surface and a face; it returns a color. There is no table, no stored palette, no bytes of color anywhere. Open the source and check.
The six faces: the result, and its two slopes
A face is a direction you read the saddle from, and each surfaces a different real quantity. Coming from the top gives a different color than from the side because you are reading a different thing about the same point.
Top +Z · the result itself, the height z = x·y. What the two ingredients make.
Bottom −Z · the same result, flipped. Its opposite, the underside.
Right +X · the x-slope of the surface, which is exactly y: how much y is pushing the result.
Left −X · the same x-slope, opposite sign. y's pull the other way.
Front +Y · the y-slope of the surface, which is exactly x: how much x is pushing the result.
Back −Y · the same y-slope, opposite sign. x's pull the other way.
So the cube tells a small story: from the top and bottom you read what was made (the product and its opposite); from the sides you read why (each ingredient's contribution, since the slope in one direction is just the other ingredient). One surface, six honest readings.
Why exactly six? Because a dimension is a perpendicular direction, a 90-degree, independent way to look. The cube holds three of them and their opposites. Read the surface at any other angle and you are no longer on a face; you are reading a vector, a blend of these perpendicular views. The faces are the pure dimensions; everything between them is a mix.
The honest part. The colours are real and derived, not stored: every pixel and every pick is computed from z = x·y and the lens, with no colour table anywhere in this file. "Frequency as the area of the slope" is a model, a way of mapping geometry to colour, not a claim that the saddle is light. And there is no compute speedup; the point is that a few dozen bytes of definition stand in for the whole palette. It works because colour is derivable. Store nothing, read everything off the shape.
⨳ Powered by Butterflyfx Manifold · valuesai.online · dependency-free · a way of looking, not a language · free for early adopters