Premium
Attention Arc Weave
Each new token on a sliding row throws arcs back to earlier ones, sized by attention weight, so a few strong links rise over a haze. Faster streams pile the fans into a dense weave.
Also known as: attention visualization, transformer attention arcs, token attention graph.
Premium
Unlock the standalone HTML and GLSL
Everything you see above is live. Basic and Pro include every premium piece, with the standalone HTML and GLSL ready to copy, plus Motions in Claude Code on Pro.
Customizing this shader
- Speed:
uTpsis the token rate anduTokensthe running count. The standalone runner adds a natural wobble to a fixed rate, the way real streams speed up and slow down. - Color: set
data-coloron the canvas, or changeuColorin the Shadertoy version. Currently previewing#8b7bff. - Size: style the canvas with CSS. It renders at up to twice your screen's pixel density and stops drawing while it is off screen.
- Reduced motion: the runner draws a single still frame for people who asked their OS for less motion.