The Synapse proof of concept successfully transformed the AHAB cognitive testing infrastructure into engaging gameplay. Built with React, Vite, and D3.js, players navigate semantic pathways between words using cosine similarity relationships, finding the shortest route between randomly generated word pairs.
Core Implementation
Technical Foundation:
- Semantic Graph: 5,000+ words connected via cosine similarity (k=7 neighbors each)
- Pathfinding: Dijkstra's algorithm treating similarity as inverse cost
- Visualization: D3.js rendering t-SNE coordinates with interactive path display
- Game Logic: React Context managing state, move validation, and performance analytics
Key Features:
- Adaptive difficulty through intelligent word pair selection
- Real-time optimal move detection and accuracy tracking
- Post-game analysis showing player vs. optimal paths
- Interactive graph visualization with WordNet definition tooltips
Validation Success
This proof of concept showed that semantic similarity could create compelling puzzle mechanics, establishing the foundation for Synapse's evolution into a comprehensive application.
Try Original Proof of Concept - Experience the original implementation described in this post.
Play Synapse Beta - Experience the evolved version as a production game with daily challenges.
Repository: github.com/neumanns-workshop/synapse (PoC submodule)