Mutually Distinct Targets

Target Optimization

“The result: every member of every set is mutually, measurably unlike every other member.”

An associative target set lives or dies on one property: no member of the set can be mistakable for any other. The viewer's sketch must discriminate — if two candidate targets share a silhouette, a texture, a theme, or even a general visual rhythm, the session was compromised before the first line was drawn. Most practice systems address this with a human eyeballing a folder of photographs once. We address it the way it deserves.

Every image in our pool is evaluated several independent ways. A modern vision model reads it roughly the way a person does — what it depicts, its overall gestalt, the semantic neighbourhood it lives in. Classical computer-vision passes read what the model glosses over: edge structure, silhouette geometry, the distribution of shape and mass. Each evaluation is distilled into a descriptive vector embedding, so that any two images in the pool can be compared for similarity — not by filename or by category label, but by what they actually look like, measured along every dimension that matters for remote-viewing discrimination.

Set construction is then an optimization, not a lottery. When a session needs a pair, a triplet, a quad, or more, the system assembles candidate sets and scores each one by its weakest link — the single most similar pairing anywhere inside it — and only a set whose worst pairing still clears a calibrated distinctness threshold survives. The distances themselves are normalized against the statistics of the entire pool, so a set that qualifies is distinct in absolute, population-calibrated terms, not merely relative ones. The result: every member of every set is mutually, measurably unlike every other member. This is machinery that until recently lived exclusively inside data-science teams. Here it runs for anyone, in milliseconds.