Angsi is a framework for turning brain-inspired AI architectures into runnable agents, test environments, and model implementations.

Angsi is an AI framework focused on experimenting with architectures inspired by the brain: neurons, circuits, cortical columns, scene-based cognition, memory, prediction, simulation, and action. The goal is not to wrap existing AI models, but to build and test alternative foundations for intelligent agents.

Core Idea

Angsi provides a practical path from theory to experiment. It is designed to support multiple levels of modeling: from biological neuron and circuit-level systems, through mesoscopic brain-inspired models, to higher-level Workspace of Scenes agents. This allows research ideas to be explored as working systems rather than only as diagrams or papers.

Key Directions

Runnable Agents – Agents that perceive environments, act, learn, and adapt through internal model dynamics.

Test Environments – Controlled environments for evaluating recognition, prediction, planning, action selection, and continuous learning.

Multiple Model Levels – Support for biological, mesoscopic, and Workspace of Scenes implementations under a shared framework direction.

Brain-Inspired Components – Experiments with neurons, macrocolumns, sparse representations, prediction, routing, memory, and scene-based architectures.

Research Status

Angsi is an early-stage research framework. It is being developed alongside the Workspace of Scenes architecture as an implementation path for testing brain-inspired AI ideas. The current focus is on building experimental agents, validating core mechanisms, and identifying which architectural components are necessary for more general forms of intelligence.