Every AI you’ve heard of learns from a record of the past. Language models learn from what people wrote. World models learn from what they observe. Decision models learn from examples of past judgments. Then they’re frozen, and the world moves on without them.
ExperienceFlow builds real Super Intelligence that learns from the consequences of its own actions. Our networks of reinforcement learning agents pursue goals people set, act continuously at machine speed, coordinate decisions that depend on one another, and get better with every result. The record of the past cannot contain every future. Experience can.
Founded by Prof. Richard Sutton
Led by Dr. Mark Ring
ExperienceFlow is envisioning a new future with experiential and continual learning. Every role in an enterprise has its private super intelligence that improves not only at an individual level but also as a network of other agents that it is collaborating with. We believe this distributed super intelligence architecture with networks of agents is a far more powerful way of enabling value to society without losing control to a handful of entities that would put humanity at risk.
Our architecture is built with security by design where the data (queries, answers, decisions, actions and weights) are all private at the level of the individual enterprise or agent.
Experience-powered real Super Intelligence matters wherever the world evolves faster than static models and human coordination can't keep up.
Deploy RL-based AI agents that learn from experimental outcomes — not just literature — explore hypothesis spaces beyond human priors and improve with every verified result. We’ll tailor a microsystem for you.
Reach outContinually improve public healthcare, education, and city planning with agents that learn from real-world outcomes, operate inside your infrastructure, and preserve data sovereignty. We’ll tailor a microsystem for you.
Reach outAccelerate research with RL-based AI agents that learn alongside your teams, provide hands-on exposure to experiential AI, and surface interdisciplinary connections. We’ll tailor a microsystem for you.
Reach out
What is Super Intelligence?
The super in Super Intelligence means beyond. Super Intelligence is machine intelligence whose abilities exceed what people can achieve on their own. Intelligence has many dimensions. Designing an experiment that no scientist has thought to try is one expression of it. Discovering a strategy whose benefits emerge only after a long sequence of decisions is another. At ExperienceFlow’s Super Intelligence Research Lab, our ambition is to build machines that can go beyond the human expertise we know how to teach them. People set the goals. Through reinforcement learning and continual learning, these machines can experiment, learn from the consequences of their actions, and discover ways to reach those goals that even our best teams could not find unaided.
How is Super Intelligence different from LLMs?
For all the advances they have brought, LLMs capture only a fraction of what AI can deliver. LLMs rely on the input of human data and cannot retain what they learn from new interactions once deployed. Their ability to evolve and improve relies on additional training and human input. Super Intelligence surpasses these limitations by continuously learning from new data, environments, and experiences without retraining from scratch. This autonomous learning enables a superintelligent system to understand what it’s working with, think through complex problems on its own, and learn from everything it experiences. It continually learns, adapts, and runs autonomous experiments to deliver new data and fresh insight every second of every day.
Why are world models important to AI learning?
World models are how AI pictures and understands the environment around it. That environment can be defined by a physical space, a company’s operations, a nation’s needs, or the entire, complex global landscape. Within this space, Super Intelligence considers a full spectrum of realities to determine potential outcomes. This spectrum includes markets, competitors, supply chains, regulations, human behavior, cultural factors, economic and political considerations, and historical data. Armed with this information, Super Intelligence uses the model to understand how all the factors interact and influence one another. By simulating scenarios, it can predict the outcomes of decisions, optimize strategies, and adapt in real time.
Does Super Intelligence replace the need for people in the workforce?
No. Super Intelligence is not designed to replace humans. It is designed to make us smarter, boost our abilities, and help us solve the problems and create the solutions we can’t on our own. The goal is partnership, not competition. Super Intelligence is meant to accelerate human progress and well-being, not to take over. While it’s true that Super Intelligence will displace jobs, it is equally true that it will create them. The World Economic Forum 2025 Future of Jobs Report forecasts 170 million new jobs and 92 million displaced, for a net gain of 78 million.
How is experiential learning controlled to ensure reliability?
There are multiple layers of oversight from both human input and AI learning to keep Super Intelligence reliable and prevent problems like bias and hallucinations. Human oversight includes designing clear, ethical guidelines and values for the intelligence systems to follow; testing to find problems early; and reviewing AI decisions to course-correct if and when needed. In extreme circumstances, there are built-in “kill switches” that people can use to pause or stop the AI. Built-in AI oversight includes learning continuously to improve; incorporating feedback from humans or other AI to fix errors; and continuously self- assessing and proactively reporting significant deviations.
What impact will experiential learning have on society?
Super Intelligence will unlock human potential like never before by solving grand challenges at lightning speed. It will accelerate scientific discoveries by thousands of times, helping cure diseases, extend lifespans and solve many challenges we once thought impossible.
