public 3d-motion repository

emergence-sims

// aifriend/emergence-sims

Interactive simulation portfolio — emergence, minds, agents & markets, live in the browser (Next.js 15 + Three.js)

$ git log --oneline --stat

stars:0forks:0updated:2026-07-11

README.md

readonly

Emergence & Intelligence Across Scales

An interactive simulation portfolio that walks one arc — the math of emergence → minds → agents → societies. Built with Next.js 15, React 19, TypeScript, Tailwind CSS 4, and Three.js. Every model is computed live in the browser from first principles; nothing is pre-rendered.

Run

npm install
npm run dev      # http://localhost:3000

npm run build produces a fully static, prerendered site (every page is SSG).

What's inside

4 premium 3D heroes — /sim/<id>:

Sim	Idea
Strange Attractors	Lorenz · Rössler · Aizawa — deterministic chaos, live in 3D
Attention, Visualized	the Transformer self-attention mechanism, `softmax(QKᵀ/√dₖ)`, as a 3D graph
Watch an Agent Learn	reinforcement learning balances a cart-pole from scratch, with a live learning curve
Intelligence Without a Goal	the same pole balanced by causal-entropic forcing — no reward, no learning

A 20-model Simulation Lab — /lab, a blueprint-styled gallery of live canvas models across three movements:

Emergence & complexity — flocking, Game of Life, reaction–diffusion, route-to-chaos, predator–prey, Kuramoto synchrony, SIR epidemic, ant colony
Minds & learning — Hodgkin–Huxley neuron, Hopfield associative memory, reward-prediction (TD-learning), Q-learning
Societies & markets — spatial Prisoner's Dilemma, Schelling segregation, echo chambers, voting methods, minority game, agent-based market, wealth & inequality, and a causal-entropy particle

Architecture

Each sim is a pure model + a renderer + an SSR page, which keeps the physics testable and the heavy client code out of server rendering:

Path	Role
`lib/*.ts`	Pure, framework-agnostic models (integrators, agents) — unit-testable in isolation
`components/*Client.tsx`	Client-only `ssr:false` wrappers that keep Three.js / canvas off the server
`app/sim/<id>/page.tsx`	Server component — SSR metadata + prose for SEO, hosts the canvas
`lib/lab/sims.ts`	Lab metadata registry (server-safe) — drives routing, static params, the gallery
`components/lab/`	Shared blueprint hooks + controls; one renderer per model in `components/lab/sims/`
`app/lab/[sim]/page.tsx`	Per-model detail sheet (SSG via `generateStaticParams` + metadata)

The reinforcement-learning agent's hyperparameters were tuned by an automated keep/discard optimization loop over a fixed 5-seed validation set (mean episode length 314 → 500).

Deploy

The site is fully static, so it deploys anywhere — Vercel, Netlify, Cloud Run, or any Node host:

npm run build && npm start

License

metadata.json

TypeScriptcellular-automatacomplex-systemsemergenceinteractive-visualizationnextjsreactreinforcement-learningSimulationstailwindcssThree.jstypescriptWebGL

[INFO] 3 topics link to curated motion topic pages.