Josue Agbekodo

Interactive Three.js scene where each player is positioned by PPG, PER, and career year. Switch layouts (stat space, era clusters, salary tiers), remap any axis to any stat, click a player for their full profile, and compare two players head-to-head with a radar chart. Fully client-side.

Parses a raw WhatsApp .txt export into a structured message timeline, then runs LLM analysis pipelines over it to extract response latency, initiation ratios, sentiment drift, and ghost events. Gives you full observability on your conversation history — aggregated stats rendered as interactive charts.

Follows the boids simulation rules (separation, alignment, cohesion) computed server-side in a NumPy/SciPy physics loop with vectorized spatial neighbor queries. Each frame is streamed to the client over WebSocket; the frontend renders it on a React canvas. Tunable sliders update simulation parameters live without restarting the loop.

Custom word is encoded in the share URL; no backend needed. Hints are powered by conversational AI with two distinct system-prompt personas — one genuinely helpful, one designed to mislead. Voice chat runs directly in the browser via the Web Speech API, streamed into the Gemini context as turns.

Players join a shared lobby over Socket.IO; the wheel spin is server-authoritative — the result is computed once and broadcast to all connected clients simultaneously so nobody can dispute the outcome. The spin animation is rendered on a Canvas element, synced to the server-emitted result.