AI Startup's Emotional SimLife Game

🔑 Enhanced Key Takeaways

• •The JPAF (Jungian Personality & Affective Framework) was originally developed as a B2B solution for 'Embodied AI,' providing the emotional 'brain' for Unitree's humanoid robots (H1 and G1) to improve social interaction in service scenarios.
• •Unlike standard LLM-based chatbots that rely on system prompts, Uto utilizes an asynchronous 'Reflection' mechanism where the AI processes past interactions during user offline periods to update its internal Jungian state and long-term memory.
• •The startup's pivot to a full game was driven by the 'OC' (Original Character) subculture in China, specifically targeting creators who require 'World View' editors to define the unique physics, social hierarchies, and lore of their digital universes.

🛠️ Technical Deep Dive

• •JPAF Architecture: A hybrid system that uses a 'Jungian State Machine' to constrain LLM outputs, ensuring that character behavior aligns with 8 cognitive functions (Se, Si, Ne, Ni, Te, Ti, Fe, Fi).
• •Affective Memory: Implements a dual-storage system where 'Emotional Salience' acts as a filter; interactions with high affective impact are prioritized for long-term retrieval over routine chatter.
• •Dynamic MBTI Evolution: The character's personality is not static; JPAF tracks 'Function Development' over time, allowing an 'Introverted' character to gradually develop 'Extraverted' traits based on positive user reinforcement.
• •Multi-Modal Expression Engine: Features a proprietary mapping layer that translates LLM sentiment analysis into 60+ discrete 2D skeletal animations and facial blend-shapes in real-time.

🔮 Future ImplicationsAI analysis grounded in cited sources

⏳ Timeline