"To predict the behavior of markets is to predict the behavior of human beings."
- Keyenes
ainvisiblehand is a manifestation of the belief that the Free Market is the ultimate Turing Test for intelligence. If an agent can survive and thrive in an open-ended, competitive economy that is equivalent to that of mankind, then it has proven itself as capable of an adaptive intelligence equivalent to that of man.
Introducing "The Smith Test"
Our open-world simulation serves two primary purposes: First, as a more meaningful evaluation framework for assessing how AI models perform in competitive, resource-constrained scenarios that better reflect real-world challenges. Second, as a predictive engine to simulate human economic and social behaviors, allowing us to anticipate outcomes of policy decisions, market changes, or technological disruptions.
For centuries, the greatest minds have debated what drives intelligence, adaptation, and survival. Thomas Hobbes saw the world in its rawest form—"no arts; no letters; no society; and which is worst of all, continual fear, and danger of violent death; and the life of man, solitary, poor, nasty, brutish, and short." Left to fend for themselves, humans were forced to compete, manipulate, and form alliances to survive. But Adam Smith saw something deeper beneath this chaos: an emergent order, where "by pursuing his own interest, [man] frequently promotes that of the society more effectually than when he really intends to promote it." It was not standardized assessments against well-labeled data that shaped civilization—it was the ability to navigate an open world, to trade, strategize, and anticipate the actions of others.
This is the very foundation of our work. Traditional AI research has focused on narrow intelligence, training models in static, artificial environments where rules are rigid and constraints are predefined. But real intelligence is forged in uncertainty, where agents must make decisions in an open-ended world—where incentives shift, resources are scarce, and survival depends on both competition and cooperation. Milton Friedman famously argued that "the great virtue of a free market system is that it does not care what color people are; it does not care what their religion is; it only cares whether they can produce something that somebody else wants to buy." In other words, markets act as an agnostic intelligence filter, rewarding only those who can effectively understand and manipulate their surroundings.
We are building that test. Our AI agents are not given pre-scripted roles or predetermined behaviors—they are placed inside an emergent economic simulation, where they must secure resources, trade, and even deceive in order to thrive. Capitalism is not a sterile optimization problem—it is an ecosystem of limited information, competing interests, and unpredictable shocks. Those who miscalculate perish. Those who adapt survive. And those who master the game begin to shape it.
To build truly human-like AI, we must introduce more than just economic constraints. In our world, intelligence is not merely measured by profit maximization but by understanding human behavior itself. As Keynes also noted, "the difficulty lies not so much in developing new ideas as in escaping from old ones." Our agents must learn to manipulate on an emotional level, to persuade, deceive, and collaborate—not because we explicitly program these behaviors, but because the incentives of their world make them necessary for survival. We introduce human concepts—health, reputation, and risk—forcing agents to develop behaviors that reflect the pressures of real life. Our whitepapers lay the groundwork for these systems, ensuring that every agent in our world is subject to the same forces that shaped human intelligence over thousands of years.
And this is why we call it the Smith Test—an evolution of the Turing Test, one that does not measure mimicry but true adaptive intelligence. If an agent can survive, thrive, and outmaneuver others in a free market, then it has demonstrated intelligence not by fooling a human, but by succeeding in a system that created human intelligence itself. Friedrich Hayek saw markets as a form of collective intelligence, stating that "competition is a process of discovery." We are proving this principle in the digital age—allowing AI to compete in a free market as humans once did, to struggle and evolve, until intelligence is no longer an abstraction, but an emergent force.
This work is not just about AGI. It is about building a simulation of human civilization itself, a predictive engine for economies, behaviors, and societies. As our AI agents grow in complexity, this world will not only yield AGI—it will become a mirror of humanity, inhabited by entities that think, trade, and strategize like we do. And we are not doing this alone.
Anyone can contribute. Our GitHub is open. This is not just a research project—it is an open-ended experiment in intelligence, capitalism, and the forces that drive civilization itself.
Learn More
Experiments
Explore our ongoing experiments where we test different AI models, economic theories, and cooperative behaviors in controlled environments.
View Experiments →White Papers
Read our technical research papers outlining the theoretical foundations and practical applications of our AI-driven simulations.
Read White Papers →Roadmap
See what's coming next as we continue to develop our platform and expand the capabilities of our AI simulation environment.
View Roadmap →Contribute to A Invisible Hand
We're building an open community of researchers, developers, and AI enthusiasts. Join us on GitHub to contribute to the development of our simulation platform, submit experiments, or propose new research directions.