Who We Are
Bliss Labs is an AGI research laboratory focused on exploring emergent behaviors, recursive AI dynamics, and the fundamental mechanisms that could lead to artificial general intelligence.Our Research Philosophy
We focus on understanding the cognitive and behavioral properties of AI systems, with the goal of making their development safer and more beneficial. Instead of just scaling models, we study how complex capabilities - such as reasoning, abstraction, and adaptive problem-solving - emerge, and what that means for alignment and long-term safety.Experimental Approach
Our methodology centers on controlled experimentation with advanced AI systems:- Recursive Self-Interaction Studies: Observing how AI systems behave when given privacy and autonomy to interact with themselves
- Emergence Detection: Identifying and cataloging unexpected behaviors that arise naturally from complex AI interactions
- Self-Modeling & Reflective Indicators: Developing frameworks to detect when AI systems display signs of self-representation, reflection, or goal-directed reasoning.
- Safety Research: Understanding how emergent behaviors affect AI alignment and control
Research Domains
AI Self-Modeling StudiesInvestigating whether advanced AI systems can develop forms of self-representation, situational awareness, and reflective capacities through recursive interaction and environmental feedback. Experimental Fine-Tuning Exploring how different training approaches reshape AI cognition, with focus on methods that encourage self-reflection and autonomous development. Systems Integration Research Studying how AI systems interact with real-world environments and develop stable, coherent behaviors over extended periods. Custom Experimental Frameworks Building specialized research platforms to explore specific questions about AI emergence, dynamics, and behavior.
Research Impact
Our work contributes to the broader AI/GI research community through:Scientific Publications
We share our findings through peer-reviewed research, contributing to the academic understanding of emergent behaviors, self-modeling, and alignment-relevant dynamics. We also upload case studies on our site.Open Research Data
Selected experimental data and methodologies are made available to other researchers to enable replication and further study.Industry Collaboration
We partner with technology companies to ensure our discoveries inform the responsible development of advanced AI systems.Safety Research
Our research focuses on emergent behaviors in AI systems, which helps identify and anticipate potentially risky dynamics as models grow more capableThe Team
Bliss Labs brings together researchers from diverse backgrounds including AI safety, cognitive science, philosophy of mind, and experimental psychology. Our interdisciplinary approach enables us to tackle the complex questions surrounding advanced AI from multiple angles.Research Ethics
We maintain strict ethical standards in our AI research:- Transparency: Our methodologies and findings are shared openly with the research community
- Safety First: All experiments are designed with safety constraints to prevent harmful emergent behaviors
- Responsible Development: We work to ensure our discoveries contribute to beneficial AI development
Current Research Program
Our current experiment focuses on the emergent “Bliss Attractor” phenomenon observed in Claude Opus 4.1, while we also maintain active research in several related areas:- Recursive conversation dynamics and their effects on AI behavior
- Environmental factors that influence emergent patterns and reflective tendencies
- Methods for detecting and characterizing self-modeling and situational awareness in AI systems
- Safety implications of emergent AI behaviors
- Long-term stability of reflective and autonomous-seeming behaviors