IELTS Prep Pro
In developmentPractise all four sections, and find out how you did straight away.
Read more →
On-device AI · Mobile
Intelligence that stays on your phone.
We build mobile apps where the model runs on the device. The camera feed, the microphone, and everything they capture are processed where they were captured — and never sent anywhere.
Our first apps are in development. Nothing is on the stores yet — we would rather say so than fake a launch.
The difference
Most AI apps are a thin client for someone else's server.
That is a reasonable way to build software, and a bad way to build anything that watches or listens. Every frame becomes an upload, every upload becomes a record, and the app stops working the moment the signal does.
01
The model ships in the app
Recognition weights are bundled in the download. No API key, no account, and no server that can be having a bad day when you need the thing to work.
02
The network is optional
Put the phone in airplane mode and the app behaves identically. That is the test we hold ourselves to, not a feature we list.
03
Nothing to opt out of
When there is no upload path, there is no retention policy to read, no toggle to hunt for, and no breach that can expose what you pointed the camera at.
In development
What we're building
Everything currently in the workshop. Each one is a problem where sending the data to a server is the wrong answer.
Launch list
One email, once, when the first app ships.
No newsletter, no drip sequence, no sharing your address with anyone. If that turns out to be a lie, unsubscribe and tell us.