A privacy-first ambient memory. Trace turns what you see into typed, cited text, on your device and in real time, then destroys the frame. Nothing is recorded, so surveillance is structurally impossible.
Today's assistants record and store the world so they can search it later. Trace inverts the pipeline: it perceives first, keeps only the meaning, and throws the raw footage away.
Capture everything, store the raw audio and video, then search through it when you ask. The footage of your life, and everyone around you, lives on a server, breachable and subpoena-able forever.
Perceive on the device, convert to typed evidence, discard the frame in milliseconds. When you ask, the answer is drawn from text you already own, never from stored recordings, because there are none.
Raw pixels exist for milliseconds and are destroyed. There is nothing to leak, breach, or subpoena. A natural fit for strict EU / German GDPR expectations, because compliance is a property of the design, not a policy you have to trust.
Every answer cites the evidence it was built from, or it refuses. Trace is measured on a reproducible benchmark and answers “I don't know” instead of hallucinating.
One pass, entirely on your own device. The raw frame never leaves the phone, and never survives it.
Local AI reads the frame. No cloud, no upload.
“desk · black keyboard · label reads DEWALT · 14:32”
The pixels are discarded in milliseconds.
Only the text is kept: local, encrypted, never deleted.
Cited answers, drawn from evidence you own.
Research-grade and improving in the open. Firm counting and a wearable form factor are on the near-term roadmap, and we'll tell you exactly what isn't done yet.
Leave an email and we'll invite you to the closed beta. No tracking, no newsletter machine, just an invite when it's ready.