64nodes

💡 Inspiration

Chess improvement is historically locked behind expensive coaches. For over-the-board (OTB) players, analyzing handwritten score-sheets usually means painstakingly typing moves into an app one by one. We wanted to build a bridge between physical chess and state-of-the-art AI, creating a pocket Grandmaster that actually speaks human.

⚙️ What it does

64nodes is a natively compiled Flutter companion app that completely streamlines post-game chess analysis:

Physical OCR Scanning: Simply snap a photo of your handwritten tournament score-sheet. The app automatically extracts the notation into a clean PGN format.
Chess.com Integration: Instantly pull your recent online match history to analyze your latest games directly from the cloud.
Grandmaster AI: Powered by Google Gemini, the app acts as a "harsh but fair" coach. It identifies the turning point of your game and gives you one clear, actionable piece of advice.
Interactive Review: Move seamlessly from text analysis to an interactive, playable digital chessboard to manually step through and review the variations.

🛠️ How we built it

Frontend: Built entirely in Flutter/Dart for a fast, responsive, cross-platform mobile experience.
Vision & OCR: Powered by Google's Gemini 2.5 Flash multimodal API to interpret messy, handwritten chess notation from physical camera input.
Reasoning Engine: We engineered a custom prompt architecture using Gemini to extract game themes, turning points, and actionable advice from raw PGN data.
UI Components: Utilized flutter_chess_board for the interactive game review and flutter_markdown for rendering beautiful coaching reports.

🧠 Challenges we ran into

Integrating complex UI packages for the interactive chessboard alongside modern AI data streams caused significant native Android build conflicts. We had to carefully manage our dependencies, isolate stable UI packages (flutter_chess_board), and manually configure Kotlin build scripts (minSdk=21) to ensure the app compiled flawlessly while handling intense asynchronous API calls.

🚀 What's next for 64nodes

Migrating the AI architecture to Google Firebase Genkit for better prompt management.
Adding localized Stockfish 16 logic directly via Dart FFI for offline mathematical evaluation.
Adding voice-to-text input so users can dictate their moves to the coach in real-time.