z = x·y is a surface. Over the square where x and y each run from −1 to +1, the height z at every point is the two coordinates multiplied. Seen straight down the z axis it flattens into this square, and here the color is set by that height. Points with the same z lie on the same curve and share a color, so the rainbow bands you see are the contours of the saddle.
Sound and light are both waves, measured in frequency. A musical octave is simply a doubling of frequency. The visible spectrum, from deep red to violet, is almost exactly one octave of light: violet sits at roughly twice the frequency of red. So a pitch and a color can be the very same frequency, many octaves apart. Take the pitch at a point and fold it upward by doublings until it reaches the octave your eyes can see, and you get its natural color. We are not assigning colors to notes by taste. We are reading where each one already sits on the frequency line.
As you move and the pitch climbs through its registers, the color completes one full red to violet sweep per octave. The register you hear and the spectrum you see are the same octave, counted once in sound and once in light. That is why the bands repeat: each loop of the rainbow is one octave of the tone.
z = x·y is a function, so the same x and y always give the same z, the same frequency, the same color, and the same tone. Nothing here is random and nothing is remembered. Move back to a point and it returns exactly what it returned before. That reproducibility is the whole point. A result you can reproduce is a result you can audit, trust, and build on.
It is a substrate demo: one deterministic field, read out as frequency, then as color and sound at once. It does not make a computer faster or a model smarter, and it does not claim to. What it buys is determinism and addressability. Every state has an exact coordinate, and the same coordinate always returns the same thing. That is the floor you want under an AI system when the goal is to take the unpredictability out.