01
Agents
Agentic systems
Agents that plan, act, and adapt across long horizons — pursuing a goal, using tools, and knowing the edges of their own competence.
01 — Intelligence division
Machines that reason and act.
We build artificial intelligence for the hardest environments on earth — where an agent has to be right, has to be fast, and has to be trusted. Agentic systems and frontier research, applied without compromise.
Capabilities
02
Reasoning
Judgment under pressure
Intelligence whose conclusions you can interrogate, not just accept — and that holds when the situation turns adversarial.
03
Perception
Machine perception
Sensing and interpretation across modalities and signal bands the human eye never reaches.
04
Alignment
Control & trust
Work on interpretability and control, so that capability and safety are made to advance together rather than apart.
05
Scale
Training substrate
The environment that trains and tests everything above — measured in outcomes, not slides.
06
Edge
Intelligence anywhere
Capability compressed to run where there is no cloud, no latency budget, and no second chance.
Active directions
AI-01Agents that pursue goals over long horizons
AI-02Intelligence that improves itself
AI-03Judgment you can verify
AI-04Many minds, one objective
Publication of results is decided case by case. Absence of a paper is not absence of a result.