How the draw works

After filters are applied, every eligible record has the same chance of selection. The browser uses Web Crypto values with rejection sampling to choose an unbiased index; multiple no-repeat results use a partial shuffle.

Recent-result avoidance temporarily removes up to ten public identifiers from the current browser session when enough choices remain. It never changes the 1000-record source pool.

A seeded draw uses a separate, versioned deterministic generator. The trimmed seed, dataset and algorithm versions, filters and quantity determine the result; ordinary session history is ignored.

A snapshot link stores public result identifiers instead. Seed and snapshot URLs may reproduce a selection, but neither contains an account identifier, email address or private draw history.

After an animal is selected, its editorial habitat data routes the result to one of 30 environmental art directions. Each environment has ten deliberately different spatial structures, producing 300 internal scene compositions—not 300 animals and not 300 claims about species.

Palette, contours, currents, strata, sonar-like rings, paper grain and registration marks are generated as abstract interface layers. They do not affect selection probability and should not be read as a range map, field observation or taxonomic evidence.

The animal name remains semantic live text in the page. A dark result layer travels from right to left across a white information layer while the title reveal is choreographed for orientation rather than spectacle.

When the browser requests reduced motion, the spatial travel resolves immediately or with only a minimal opacity change of about 120 milliseconds. The result and controls remain usable without animation.