Lived context for AI agents.
The knowledge you gather, the actions you take, and the decisions you make become the source of truth for AI, giving agents the context to complete work the way you would.
ARBOR-1.
Across individuals, teams, and a select group of Fortune 500 enterprises, ARBOR-1 is deployed to surface decision traces and the trajectories of how tasks are completed, creating the living SOR of knowledge workers.
500Deployment
Privacy research.
Together with Carnegie Mellon's security and privacy institute (CyLab), we research small language models that run on device, so sensitive information stays on your machine and never has to leave it.
Joint research
DeskBench.
DeskBench is a dataset built from 500,000 hours of real cross-industry work sessions, captured on macOS and Windows. It covers the context humans produce across long-horizon sessions, built to test whether agents can retrieve real human work.