Motion World · 1 of 6

One rule, grown slowly, never the same twice.

Arctic Snowflake is an authored experiment in order, variation and fragile accumulation. A single deterministic growth rule, driven by a visible seed, builds a luminous crystalline form against a cold dark field. Your pointer biases which way the crystal reaches. GA made it to practise seeded state, patient rendering and interfaces that reward slow attention.

The current seed is printed on the stage. Same seed, same size, same crystal.

Poster: generated still from the world’s own growth rule.

A still crystal grown from one rule: sixfold symmetry, related arms, none identical, holding a fragile shape against the dark.

Move to influence the direction of growth. Reset grows a new seed. Nothing here tracks you.

Technical note

How it works

The crystal grows on a 2D canvas from a seeded pseudo-random generator (mulberry32). The seed is printed in the corner of the stage: the same seed, at the same size, grows the same crystal. Each growth step extends one active tip; with a small probability the tip forks. Every segment is drawn with twelvefold symmetry, six rotations mirrored, which is what makes one simple rule read as a snowflake.

Your pointer does not steer individual tips and it never draws. It warms the growth: alignment with the pointer direction slightly increases step length and branching probability, so the whole form leans toward your attention without becoming a drawing tool.

Performance choices

  • Growth is capped at about 3,200 base segments on large screens and 1,500 on small; active tips never exceed 42.
  • Settled segments accumulate on an offscreen canvas, so each frame draws only new growth.
  • The render loop pauses when the stage scrolls offscreen, and pixel ratio is capped at two.
  • Reset chooses a new seed and grows again from a single point.

Accessibility and comfort

With prefers-reduced-motion set, the canvas never starts and the static interpretation above stands in for the piece. Growth is slow and low contrast by design; there is no flashing, and this world has no audio at all. Pause and Reset are real buttons, reachable by keyboard, and everything the canvas shows is carried in words on this page.

From laboratory to client work

What this practises

The disciplines on show here, deterministic state you can replay, randomness you can audit, rendering that stays calm on a mid-range phone, are the same disciplines behind trustworthy data-to-motion. A chart that cannot explain its own state is decoration; a snowflake that prints its seed is a state machine.

That is the standard GA brings to dashboards: motion only where it changes a decision, and state that can be reconstructed on demand. The Samples · 200 Capability Levels page shows where generative visuals sit on the complexity ladder, and what each rung costs.

See the craft applied

The laboratory keeps the ideas honest. Samples shows the same craft aimed at business interfaces, with trade-offs stated at every level.