xanadu.science is a continuous laboratory -- an experimental cartography of interconnected research, rendered as a generative particle field that reorganizes with attention. Move the cursor; the network responds.
Following Ted Nelson's original Xanadu vision, we treat every citation as a bidirectional thread. Each node in the network maps a paper, a dataset, or a question; each edge marks a transclusion -- a piece of one document quietly held inside another.
The background you are watching is not decoration. It is a Verlet-damped field of 112 agents whose target positions are recomputed each time you scroll. The system models reading itself as a force.
Within a 150 px radius the nearest agents brighten and curve toward the pointer. This is local-illumination: a tiny pocket of bright nodes travels with you across the field as you read.
As you descend the page, the particle field re-seeds its cluster centers around the section currently in view. The knowledge network literally reorganizes around your attention -- an animated, embodied table of contents.
A single requestAnimationFrame loop integrates the particle field. Velocity is damped to 0.92 each tick; targets drift smoothly with eased interpolation toward cluster centroids.
// per-frame integration p.vx += (target.x - p.x) * 0.012; p.vy += (target.y - p.y) * 0.012; p.vx *= 0.92; p.vy *= 0.92; p.x += p.vx; p.y += p.vy;
On every frame the five particles closest to the cursor are scored and their glow radii are interpolated up to 15 px. Their connection edges rise to full opacity in 300 ms.
A barely-visible grid (rgba 100, 200, 255 / 0.03) is drawn first each frame. It is the only static element in the canvas -- a quiet spatial reference for the otherwise flowing field.
The laboratory is open at all hours. Submit a question, a citation, or a dataset. Replies arrive in plain text, by light.