Root Relay is a local Three J S learning diagram. Each genuine Launch click increments a browser-stored count, adds one box and one connector to a fixed central mesh. Starting empty, three clicks give three boxes, three connectors and four meshes including the root. Count bars at eighty display pixels per object have widths two hundred forty, two hundred forty and three hundred twenty. No wallet transaction, chain deployment, consensus or actual Root benefit is measured by this code. The supplied claim is shown as a claim, and the mechanism is explicitly unknown. New box number k uses angle one point seven times k, x equal three cosine angle and y equal two sine angle. The first box at k zero sits at three, zero. At k one it is about minus point three eight seven, one point nine eight three. At k two it is about minus two point nine zero zero, minus point five one one. The drawing uses a common eighty pixels per source coordinate unit and connects each box directly to the origin. The radius is elliptical, not a measured network distance, latency or transfer rate. Lines show a source visual connection, not evidence of an economic mechanism. The first Launch raises the displayed mastery to at least thirty-five percent. More Launch clicks leave it at thirty-five. Test a claim immediately sets it to one hundred, without scoring an answer or checking a mechanism. Bars at four pixels per displayed percentage point have widths one hundred forty and four hundred; that sixty-five-point difference is an interface state change, not measured learning. Reset removes all diagram boxes and connectors and stores zeros. Reload rebuilds the stored boxes and score, but the explanatory sentence returns to its original source claim. Export is a small three-field local JSON record.