Intelligence that reasons
Moving beyond pattern recognition toward systems that think through problems, test ideas, and understand uncertainty.
Reasoning · Planning · World models
A new perspective on reasoning
Koodos
Research for a more
human future.
We study how intelligence works—and how to build AI that reasons, learns, and helps us move forward. For everyone.
Independent. Nonprofit. Open to possibility.
Big questions. A shared direction.
ReasoningLearningUnderstandingHumanity
The questions that move us
More capable.
More understandable.
More useful to humanity.
General intelligence is an open scientific challenge. We approach it with curiosity, careful experiments, and the conviction that progress should serve people.
Learning that lasts
Exploring how AI can adapt to unfamiliar situations and keep learning from experience without losing what it knows.
Continual learning · Foundation models
Progress with purpose
Making safety, interpretability, and rigorous evaluation part of the science—not an afterthought.
Alignment · Interpretability · Evaluation
Upcoming project
The Media Agent
— an agent that makes media.
Give it a goal, and it plans, produces, and edits complete media — scripts, images, video, and sound — as one system. With provenance and disclosure built in.
Explore the program
Good science starts with good questions.