SCC Lab

An interactive laboratory for exploring synthetic cognitive architectures — AI agents that learn, adapt, and reason in procedurally generated grid worlds.

Overview

SCC Lab is a research-oriented project that simulates agents navigating grid-based environments with obstacles, keys, doors, lava, and symbolic triggers. The agents learn through a biologically-inspired cognitive architecture featuring reinforcement learning, attention gating, sleep-based memory consolidation, and causal reasoning.

Key Features

• Cognitive Architecture — A multi-layered system orchestrating attention, executive control, temporal integration, and causal reasoning
• Reinforcement Learning — Q-learning agents that explore, learn, and consolidate knowledge across episodes
• Sleep Consolidation — Offline replay of experiences to strengthen memory and discover causal chains
• Cognitive Witness — A metacognitive observer that evaluates agent behavior and detects learning phases (exploring, learning, mastery, regression)
• Symbol System — Rich symbolic triggers embedded in the world that agents must learn to interpret
• Meta World Builder — Progressive difficulty scaling that adapts the environment to the agent's skill level
• Real-Time Visualization — Live canvas rendering of the grid world, agent movement, and learning metrics via WebSockets

Tech Stack

• Frontend: Next.js with React and TypeScript
• Real-Time: WebSocket server for live agent streaming
• Visualization: HTML Canvas for grid rendering, custom SVG charts for metrics
• Styling: Tailwind CSS