Aura

A cognitive architecture that simulates how personality shapes thought, language, and social behavior.

What is it?

Aura is a 15,000-line Python simulation where AI agents develop unique personalities, form social bonds, learn language, publish papers, and make decisions—all from emergent processes rather than pre-programmed scripts. No two agents behave the same way.

Core Architecture: Cerome

Each agent has a Cerome—a five-layer personality model inspired by neuroscience and psychology. L1 (temperament) is genetic, L2 (values/beliefs) develops through experience, L3 (social ties) tracks relationships, L4 (drives) fluctuates like real moods, and L5 (working memory) handles moment-to-moment cognition. Together they create agents that feel genuinely individual.

Key Innovations

MFA Attention Field — attention follows magnetic-field mathematics (my own research theory). AuraLang — agents develop a shared language from scratch through social interaction. Zero Presets — strategy and personality emerge; nothing is hardcoded.