We build agents (that build agents)ⁿ: agents that build better and more specialized versions of themselves.

Our mission is to make sure anyone can build an agent that outperforms a frontier lab’s generalist agent on the problem they care about. The best agent for your problem should be yours to build because you have the deepest understanding of the problem space, and because the hardest problems hinge on knowledge that never made it onto the internet: a science lab’s decade of failed experiments, or a factory’s quirks that only the night shift knows.

The richness of the world’s problems can only be matched by the richness of the intelligences that tackle them. We believe in a Cambrian explosion of agents, instead of a monoculture everyone has to settle for.

Getting there means democratizing recursive self-improvement (RSI). Today, AI that improves AI is concentrated in a handful of well-resourced labs. We want to put that ability in your hands: you build the agents that keep getting better at your problem on their own, instead of hoping a future product release solves it. Our goal is to give everyone the power to shape agents around what matters to them.

People are at the center of how our agents improve. They choose which problems matter, define what “better” means, and recognize when an agent is gaming its own metrics. Their taste and preferences give self-improvement its direction.

Our RSI skill tree is where we’ll publish results selectively along the way.

If you’d like to become a MakerMakerMaker, reach out. For anything else, contact us.