3D visual studies
Small Three.js scenes for testing geometry, materials, lighting, camera motion, and spatial interfaces.
Experimental workspace
A public shelf for 3D sketches, interactive visualizations, small games, and simulations built with Three.js and AI-agent workflows.
Live experiments land here as focused demos with notes on what was tested, how agents helped, and what changed between versions.
Lab viewport
render
queued
agent pass
queued
visual qa
queued
Tracks
The lab can grow from lightweight studies into richer demos without changing the rest of the site architecture.
Small Three.js scenes for testing geometry, materials, lighting, camera motion, and spatial interfaces.
Compact games and interaction loops where mechanics, feedback, and performance can be explored quickly.
Systems experiments around motion, particles, constraints, procedural rules, and real-time state.
Build notes
The goal is not only to publish a canvas, but to understand how AI agents can help design, implement, test, and refine interactive 3D work.
Keep each experiment isolated so heavy rendering code does not slow down the main site.
Pair live demos with short notes on the model, prompting approach, agent workflow, and technical tradeoffs.
Prefer measurable interaction quality: frame rate, loading behavior, accessible controls, and clear fallbacks.
Use the lab as a public practice surface for AI-assisted 3D engineering, not as a polished portfolio theater.
Experiment shelf
Published lab entries stay isolated from the core site and include a short note on the interaction model, AI-assisted workflow, and technical tradeoffs.
A deterministic Three.js miniature railway with branching tracks, station dwell timing, automatic routing, sound cues, fullscreen controls, and reversible transport controls.
A Three.js sandbox game for assembling a fictional stainless-steel reusable launch vehicle, saving the build, and replaying a launch-and-return sequence.
Planned
A controllable scene for generating forms, materials, lighting states, and camera paths.
Planned
A compact mechanic with scoring, motion constraints, collision feedback, and agent-assisted iteration notes.
Planned
A demo that records how an AI agent helped shape the scene, implementation, tests, and visual QA.
A strong lab piece can be a single interactive system with clear controls, visible behavior, and a short note on what the agent generated versus what needed human taste and debugging.