conscious geometry
a robot cannot distinguish you waving hello from you waving for help. an agent cannot distinguish whether your silence is approval or frustration. an automobile doesn't know whether you are alert or distracted before asking for input.
existing models are not conscious. and in a world shared between ai and humans, this is a problem.
we believe in a world where humans live alongside models that are aware across every human signal models that know when to speak and when to stay quiet, models that can read a room, models that recognize people’s emotion and infer their intentions.
to get there, we are driven by the relationship between consciousness, computation, and philosophy. we believe that humans’ inner states have a shape: thought, emotion, and intent are not formless, but are concrete and recognizable. we train conscious models to detect these states and produce actionable insights for autonomous systems such as robots, cars, and biological monitors.
our team has backgrounds in computer science and philosophy from harvard and oxford, and mathematics and ai from an MIT phd. we have backgrounds in quant trading, and are currently fellows at anthropic and MIT in san francisco. while working on human-aligned models, we realized existing models do not understand us. we believe conscious models are the most important problem to solve, so we left school, and turned down hedge funds and ai labs to work on it.