In our research we focus on creating a universal knowledge representation for any computation - a new lingua franca for AI. We rely on category theory, graph-based knowledge representation and computational theory of meaning to achieve it. New properties emerging from this approach are: interpretability, provable reliability, security, continual learning.

We are building a new AI architecture designed for factual accuracy, comprehensive memory retention, and continuous learning and evolution. Our system combines three innovations:

1. a new language based on category theory that defines knowledge through actions,
2. graph-like data structures that allow efficient System-2 reasoning, and
3. an extendable workspace where all components exist, collaborate and communicate seamlessly without translation losses between formats, data types or languages.