Mission

WalNut is an artificial intelligence research and product company. We learn the principles behind people, build worlds that could exist, and turn them into a new kind of reward.

Humans, the missing piece

We can simulate nature, but not yet people. Simulations already describe weather, fluids and motion remarkably well. Modeling how humans think, feel and decide has mostly failed. We believe it’s the last missing piece.

Beyond verifiable rewards

Rewards from consequences, not rubrics. RL works where a verifier exists, as in math and code. Our worlds act as verifiers for everything else, letting agents act, see what happens and learn.

Principles, not a replica

Worlds that could exist. We don’t copy reality. We learn its principles and build many possible worlds from them. Reality gives one history; we want agents that hold up across all of them.

Research and product, together

One loop. Real interactions sharpen our principles, and better principles make better worlds and better rewards.

Join us

We’re looking for people who want to build worlds, not just models. Get in touch.