Who We Are

We build AI systems trusted to run unattended. Ten of them are shipped. One trades real capital under a statistical gate that killed 978 of 981 candidate strategies. One maps 466,689 buildings across two power grids. One holds people's medical records and never sends them anywhere.

AI agents write most of that code. The engineering that decides whether any of it can be trusted is the part we sell: architecture gates that reject a commit and explain the rule it broke, servers that discard any AI-sourced fact arriving without a citation, and monitoring that catches a system when it drifts.

We work from Brooklyn and Vancouver, and we take on a small number of projects at a time.

What we believe

Intelligence, artificial or otherwise, should serve the flourishing of all life. We build that in as architecture rather than stating it as a value.

Elsewhere