Diverge: Chess AI Review (Maia & Stockfish)

In-browser chess review with Maia human intuition vs Stockfish engine best moves. Move heatmaps & 100% private local analysis.

⚖️ FAIR PLAY FIRST · STRICT ANTI-CHEAT COMPLIANCE Diverge strictly adheres to the Fair Play and Anti-Cheat regulations of Chess.com and Lichess.org. • Zero Live Assistance: During live timed games, Diverge automatically engages a fail-safe lock. Engine computation completely halts (0% CPU), and all visual overlays are cleared immediately. • Legitimate Post-Game Use Only: Diverge unlocks exclusively when games are concluded, on analysis boards (/analysis, /study), and in tactical puzzles (/puzzles). ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ Understand how humans think, how engines calculate, and where the gap lies. Standard chess engines calculate like cold supercomputers, leaving you wondering: "Why did I play that move? What would another human at my rating play?" Diverge brings two powerhouse engines together right inside your browser: 1. Maia 3 (Chessformer Neural Network): Models real human chess intuition, predicting the most natural candidate moves and their probabilities at calibrated Elo rating levels. 2. Stockfish 19 (WebAssembly): The world's strongest open-source engine, delivering objective tactical evaluations and optimal computer moves. Both engines run 100% locally on your machine via WebAssembly. No game data or positions are ever uploaded to any server. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ ✨ KEY FEATURES ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 🏹 Side-by-Side Visual Comparison • Gold Solid Arrow: Human intuition candidate & probability (e.g., e4 64.3% at your Elo). • Emerald Dashed Arrow: Stockfish objective best move and position score (e.g., d4 +0.35). • Consensus Glow: Automatically highlights when human instinct matches engine perfection. • Target Heatmap: Aggregated human probability heatmap across all legal destination squares. 🎯 Calibrated Intuition Ratings Fine-tune Maia to mirror real human play at any skill level: • Quick presets: 1100 (Beginner), 1500 (Intermediate), 1900 (Club), and 2200+ (Master). • Continuous slider (600–2600) to benchmark against your exact rating bracket. 📊 Standalone Deep Review Studio (PWA) Take your post-game analysis to a dedicated full-screen workspace: • Win-rate & advantage momentum charts. • Divergence breakdown: identify human blunders, subtle inaccuracies, and clever traps. • Import via PGN string, .pgn file, or direct Lichess / Chess.com game links. • Installable as a standalone PWA for offline post-game study. 🔒 100% Local, Private & Fast • True Edge Computing: WebAssembly and Float32 neural network inference execute entirely in your local browser memory. • Zero Telemetry: No board positions, FEN strings, or browsing history leave your device. • Instant Cache: Model weights are cached securely in browser IndexedDB after the first load for offline zero-latency use. • Clean & Respectful: No ads, no analytics, no third-party tracking. 🌐 WHERE IT WORKS • Lichess.org: Finished game review, Analysis Board, Studies, Puzzles. • Chess.com: Finished game review, Analysis Board, Puzzles, Classroom. • Standalone Studio: Built-in analysis workspace or 4chess.cc. Make post-game review a bridge to genuine chess understanding.