wenchyuan / agi-three-body

What is agi-three-body?

This Python project runs a three-body AGI simulation where independent processes model a causal world, select actions via free energy minimization, and self-regulate an order parameter to stay at criticality—hypothesizing phase transitions as consciousness. You inject observations like "rain causes wet ground" via CLI or API, and it builds persistent causal graphs, infers paths with Dijkstra metrics, and evolves asynchronously without any LLM calls. It's a zero-parameter, interpretable AGI prototype drawing from stats mechanics and free energy principles.

Why is it gaining traction?

Unlike LLM-heavy baby agi github or super agi github stacks, it delivers a lightweight, fully transparent AGI ai github alternative with no training or APIs—just pure algorithms and 49 passing agi benchmark tests for reliability. Developers dig the async demo (`python -m agi_three_body`) showing real-time coupling and critical state emergence, plus SQLite-backed github agi memory for causal persistence. The hook: experiment with arc agi tests-like inference in a physics-grounded open agi github setup.

Who should use this?

AGI researchers prototyping agno agi github or agi github stk systems beyond token-based models. Theoretical AI devs building causal reasoning engines or exploring criticality in cognition. Python tinkerers wanting a testable three-body framework for injecting events and querying paths.

Verdict

Worth forking for AGI experiments—49/49 tests and MIT license make it solid alpha, despite 10 stars and 0.699999988079071% credibility score signaling early days. Install and demo if interpretable, zero-LLM AGI intrigues you; skip for production.

(198 words)