Biome Lab
Author: OpenAI
A living procedural terrarium with adjustable ecosystems.
Description
Biome Lab is a deterministic ecosystem sandbox built with Codex and GPT-5.6. Users tune climate, terrain, water, and species balance, then watch a stylized 3D terrarium and compact trend charts respond.
Build notes
Initial prompt
Build “Biome Lab,” a procedural ecosystem sandbox that runs entirely in the browser. The main view is a living 3D terrarium containing terrain, water, plants, and a few simple animal populations. Let users adjust rainfall, temperature, terrain height, water level, and species balance, then immediately see the ecosystem respond. Use deterministic procedural rules rather than an external API or large dataset. Changes should affect plant coverage, water, animal counts, colors, and ambient motion, with compact charts showing population and environmental trends. Include useful presets, a seed control, pause and play, reset, and randomize. The result should feel like a visual experiment, not a scientific dashboard. Keep the simulation understandable and performant rather than biologically exhaustive. Support desktop and mobile, run build and lint, and verify presets, controls, reset, and sustained animation in the browser with no console errors or major frame-rate issues.
Iterations
-
Built the seeded React Three Fiber ecosystem with terrain, water, vegetation, weather, wildlife, environmental controls, presets, playback, reset, randomization, and live sparklines.
-
Expanded the scene with denser flora, varied rock formations, water life, and recognizable low-poly deer, capybaras, foxes, wolves, fish, and ducks.
-
Stopped camera, water, vegetation, weather, chart, and day movement while idle so only wildlife moves until a setting changes.
-
Rebuilt the trees with cohesive faceted crowns, tapered trunks, grounded roots, richer foliage variation, and distinct broadleaf, wetland, and pine silhouettes.
- Final step
Finished a responsive procedural terrarium with expressive low-poly ecosystems, deterministic controls, understandable trends, and motion that stays calm until the user intervenes.
Related projects
Material Lab
Material Lab is an interactive 3D design tool built with Codex and GPT-5
Terrain Mixer
Terrain Mixer is a browser-based landscape tool built with Codex and GPT-5
London Dream Railway
London Dream Railway is an interactive 3D scene with miniature trains...