Skip to content

Agent skill for creative 3D visualization: turn an idea into a Three.js/Blender scene worth exploring. Claude Code + Codex plugin, 223 recipes, 440 knowledge records, 22 proved kits, 37 runnable studies.

1.5 KB
Raw Blame

Probability in Glass

A seeded three-dimensional Gaussian sample with a covariance ellipsoid and marginal-density floor curves.

From examples/, run pnpm install once, then pnpm dev and open /probability/. Use pnpm build for all examples.

Generation prompt

Use 3dviz-pro-max to create a seeded multivariate-Gaussian point cloud with a covariance ellipsoid. Let me vary population correlation, sample count, horizontal spread, seed, and shell visibility; report sample correlation separately; and explain exactly what the ellipsoid and floor projections do and do not encode.

Explore and implementation

Use Overview and Detail. Adjust ρ, sample count, and σx; draw another deterministic sample; toggle the shell. See scene.js, the shared scene, and math helpers.

Source and limits

The model follows Stanford CS229's multivariate Gaussian notes. Box–Muller samples are transformed by a Cholesky factor. The shell is Mahalanobis radius 2, not a 95% boundary in 3D. Correlation controls X–Y; the floor shadow is X–Z and does not display it.

Skill rules actually applied

The artifact corresponds to recipe.covariance-ellipsoid-axes in probability and statistics, especially population/sample separation and still-state rendering. This is a retrospective mapping, not original retrieval evidence.