Intelligence,
put to work.
We research and develop agents that work 24/7 on everyday work, investment decisions, and the environments that train them.
Explore our work
One lab.
Three ways to put agents to work.
Gru
A general personal assistant that stays with your work and life, around the clock.
gru.airethinka
An investment workstation: research, testable ideas, and ongoing review from idea to conviction.
rethinka.aiGbox
Data, evaluation, and reinforcement learning environments, grounded in the work agents already do.
gbox.ai
Build in the world.
Learn from the work.
Our agents work in the lab itself, not only as products. They take on real lab work—so building agents and using agents are the same loop.
Feedback from live work is collected continuously. Then people and agents in the lab fix what breaks—improving the agent and the environments it trains in.