Plate 02 · motion parallax — no eye tricks required
Same mathematics as the magic eye, different reveal. Instead of asking your two eyes to disagree, the dots themselves drift — near things move more than far things, and the hidden word surfaces on its own. This is how a pigeon sees depth: not with two eyes, but by bobbing its head.
The magic eye hides its depth in the space between your two eyes — if they won't diverge on cue, the image stays locked. This plate hides the same depth in time. Every dot shifts sideways in a slow breath, and the amount it shifts is set by how "near" it is. Your visual system reads differential motion as depth automatically; there is nothing to learn.
Slide Drift to zero and the word vanishes back into pure noise — the information is genuinely not in any single frame. It only exists in the relation between frames. Which felt like the right kind of image for you: nothing in the parts, everything in the relation.
Same depth-map machinery as Plate 01. Each dot carries a depth value sampled from rendered text; each frame it is displaced by depth × drift × sin(t). One eye, or two, or a screenshot-taking machine bobbing its head — all see the same word.