Systems for intelligence.
Built with care, at human scale.
We design and ship AI and computing systems — from the silicon up. Local-first infrastructure, intelligent agents, sensor platforms, and applied research for the products of the next decade.
What we build
A small, opinionated surface area.
We work across a handful of deeply related disciplines. Each one feeds the others — models need memory, memory needs sensors, sensors need infrastructure, infrastructure needs care.
Applied AI systems
Production-grade agents, retrieval, and reasoning pipelines — engineered for real workloads, not demos.
Local-first infrastructure
Self-hosted engines that run on your hardware. Your data stays where it was created.
Knowledge & memory
Systems that capture, structure, and reason over long-horizon context — personal and organizational.
Sensor & edge platforms
End-to-end stacks from embedded devices through cloud, with observability and OTA as first-class concerns.
Interface & signal
Natural-language and multimodal interfaces. The boring parts of product, done with craft.
Research, shipped
We take speculative ideas and turn them into products — or honestly report when they don't work.
How we work
Principles, not a playbook.
Small teams, long runway
Every product is built by a team that fits in one room. We design for five-year horizons, not five-quarter ones.
Local-first, by default
We ship software that runs on your machine, your server, your edge. Cloud is an optional convenience, not a requirement.
Open where it matters
Protocols, schemas, and interfaces should be inspectable. If you can't read the source, you can't trust the system.
No shortcuts, no compromises
Correct beats quick. We fix what we find, verify what we build, and document why things are the way they are.
Have a system that should exist?
We take on a small number of engagements each year — research, product, and infrastructure work where the craft matters. If any of that sounds like you, get in touch.