Research
Our research bet is that more capable AI, and in the long run forms of superintelligence, should be built on structure that can be inspected and verified, so capability and accountability grow together. This is a long-term horizon, not a near-term claim.