An AI memory that remembers contexts — not a chatbot that forgets the thread

You tell an assistant something important — a decision, a name, a plan — and weeks later you have to reconstruct it yourself. Chat threads forget. Browser AIs only know what you pasted into that one conversation. Your life is not one chat.

Claire is built for the opposite problem: your days create more information than you can reliably remember and file. Capture it, organize it, remember it, ask for it later.

Standard large language models generate text, code, and web answers. Claire’s primary job is local tracking: you start transcription when you want words kept, in the context you choose (personal, work, or private / mic off).

The voice notes exist whether or not you ever open AI. Transcription does not require a model. Nothing is sent while Claire is transcribing. You can delete a capture immediately. Your logs are not training material for a public model.

When you do ask, Claire behaves like a secretary who already has the file: “Summarize what happened in that meeting.” “What did we decide about the trip?” “Explain that part of class again.” The helper works from what you saved — it is not omnipresent.

A hallway decision, a class, a work call, a grocery thought — they are not the same pile. Modes keep lanes apart so a work question does not drag in a family dinner. Private mode stops recording when the room is not yours to keep.

You find things with on-phone search (text and date) even offline. AI is optional, needs the internet, and only receives excerpts for that question — not the entire archive.