The home is the trust boundary

Most smart-home products are designed around a simple bargain: convenience in exchange for data. A voice request leaves the house, is processed somewhere else, and an instruction comes back. The device is in your kitchen, but the intelligence is not.

That bargain made sense when useful AI needed a data centre. It makes less sense now that capable language and vision models can run on hardware small enough to sit on a shelf.

What changes when intelligence is local

• Latency drops, because requests do not travel across the internet and back.
• Resilience improves, because core functions can keep working during an outage.
• Privacy becomes an architectural property rather than a policy promise.
• Memory becomes possible, because a household can keep context it would never upload.

Memory is the real prize

A connected home follows commands. An intelligent environment needs context: who prefers which lighting, which evenings are quiet, what a normal morning looks like. That context is intimate. We think it should belong to the household, be visible to the household, and be deletable by the household.

Our design goal: the home is the trust boundary. Data crosses it only when a person explicitly decides it should.

What we are not claiming

Local-first does not mean perfectly private or never online. Weather, software updates and some integrations depend on outside services. The point is that those connections should be optional, visible and authorised, not the default path for everything a home knows.