The verification layer for visual intelligence.

The prevailing narrative frames AGI as a language problem. But language is only a fraction of human intelligence. The vast majority of economically valuable tasks are embodied - physical, spatial, visual. Our founders believe AGI will be multimodal visual intelligence: systems that see, reason, and act within the physical world.

Video understanding and video generation are converging into unified omni-models. This convergence is already underway at every frontier lab. But alignment for visual systems is fundamentally harder than alignment for language. The signal space is high-dimensional. Subtle pixel-level variations carry enormous semantic weight. Existing alignment pipelines, designed for text, break down when applied to video.

We are building the verification layer that enables frontier labs to train video AI systems that are not just capable, but safe, controllable, and aligned with human intent.

We write to clarify our ideas and share our perspective on where video AI is heading. These essays explain the intellectual foundations of our work.

We are a small team at the intersection of machine learning, visual perception, and philosophy of mind. We come from frontier AI labs where we built multimodal systems, video world models, and large-scale evaluation infrastructure. We left because we believe the verification problem for visual AI deserves dedicated, first-principles attention.

We are based in San Francisco.

We are hiring people who want to work on the hardest unsolved problem in AI alignment. If you believe the alignment problem extends far beyond language, we would like to talk.