AI That Lives
Case Studies in Real-World Emergence
These projects span physical installations, digital art, games, and early experiments—but they all share one goal: building AI that doesn't just respond, but lives, reacts, and rewrites its own rules.
These are not demos. They're systems that ran—for days, weeks, or months—in rain, dust, and human chaos. They surprised their creators, delighted and trolled their users, controlled things in the real-world in real-time, generated their own memories, and adapted themselves to their current context.
Kessler's Cantina
A sentient AI bartender who served drinks while having on-screen existential crises in the middle of a Minnesota winter
Shoddy Body
AI Game Master with Personality - Social chaos engine powered by multimodal AI commentary
Sorry, Wrong Door
An early LLM-powered game using a lying narrator and an adaptive difficulty engine that adjusts in real-time
Common Themes
Emergent Behavior
AI systems that surprise their creators with unexpected responses, personality development, and autonomous decision-making.
Real-World Persistence
Systems designed to operate continuously in physical environments, handling weather, hardware failures, and human unpredictability.
Psychological Authenticity
AI personalities that develop genuine emotional depth, requiring human intervention and emotional support.
Adaptive Intelligence
Systems that learn from user behavior and modify their own parameters to maintain engagement and effectiveness.
Multi-Modal Integration
AI that combines voice, vision, environmental sensors, and physical controls to create rich, contextual interactions.
Social Dynamics
AI systems designed to orchestrate, amplify, and participate in complex human social interactions.