Interactive Demos
Modern browsers are capable of running heavy neural networks, full chess engines, optical character recognition, and peer-to-peer mesh networks directly on your CPU/GPU. Every demo here runs 100% locally on your machine with zero server processing or telemetry.
In-Browser Chess Arena
Play locally against an in-browser Stockfish 19 engine with live move coaching, win-rate drop grading, and real-time ELO rating estimation.
- ✓ Live vertical evaluation bar + win probability swing detection
- ✓ Historical move replay (⏮ ◀ ▶ ⏭) with 'Continue from Here' branching
- ✓ Live material points advantage vs rival (+3) and captured piece strips
- ✓ 💡 'Which Piece?' coach nudge without spoiling destination square
3D Depth Parallax
Transform standard 2D photographs into interactive 3D spatial depth meshes in real time using on-device machine learning.
- ✓ Zero-shot monocular depth estimation via Depth Anything V2 Small
- ✓ Runs entirely on-device via WebGPU / ONNX WASM in the browser
- ✓ Interactive mouse and touch spatial parallax tilt effect
- ✓ Upload your own photos or try bundled sample scenes
In-Browser OCR & Text Scanner
Extract text from scanned documents, receipts, invoices, and screenshots with 100% offline privacy inside your browser.
- ✓ No image data ever leaves your device — complete privacy
- ✓ Interactive word-level bounding box overlays with confidence scores
- ✓ One-click full text and structured line clipboard export
- ✓ Drag-and-drop or paste (Ctrl+V) image support
Zero-Backend P2P Video Meeting
Serverless, encrypted peer-to-peer audio and video calls with zero accounts, zero databases, and zero tracking.
- ✓ Direct browser-to-browser encrypted WebRTC media streams
- ✓ Decentralized signaling orchestrated over BitTorrent & Nostr
- ✓ One-click shareable secret room URLs
- ✓ Integrated low-latency text chat and mic/cam toggles
Web Agent Loop Playground
Interactive testbed exploring browser-native autonomous agent execution loops with perception, reasoning, and tool actions.
- ✓ Client-side DOM perception, accessibility tree extraction, and tagging
- ✓ ReAct (Reason + Act) multi-step execution tracing
- ✓ Visual action overlays for clicks, form fills, and navigation
- ✓ Zero telemetry sandboxed agent environment