Audio & privacy.

Useful context should not require an indefinite archive of your conversations.

Design overview · Product in development. This page describes our intended direction, not verified protections or a finalized privacy policy.

Waitlist signups.

If you join the waitlist, we store your email address, chosen charm options, signup time, and your agreement to receive early-access and product updates. Records are stored in private Vercel Blob storage accessible to our site backend and authorized project administrators. Your email is not published or shared with other visitors. This site does not record audio. We retain your signup until you request removal or the waitlist closes.

Audio-only sensing.

The charm concept captures audio, not photos or video. Transcripts and saved notes can still contain personal information and need their own protections.

Keep less.

Our intended flow is encrypted backend processing, deletion of raw audio immediately after processing, and retention of only high-level notes. Audio recordings and full transcripts are not intended to be retained. These protections are in development and are not yet verified. We still need to define processing time limits, failure handling, temporary-transcript cleanup, note selection, and deletion from providers and backups.

Make the controls clear.

We’re defining how people can tell when capture is active, stop capture, review saved information, and delete it. These controls and their limits need to be verified before we describe them as available.

Explain where information goes.

The final documentation will identify processing on the charm, phone, or servers; any service providers; who can access information; whether data is used for model training; and when deletion completes. We are not yet claiming that processing stays on-device, that only the user can access data, or that audio deletion is already enforced.

Respect the people around you.

Audio can include other people’s voices. Let people know when capture is active, ask before capturing a conversation, and stop if someone is uncomfortable. Deleting audio later does not replace that consideration.

Draft updated October 24, 2026. Product privacy details will be updated as the product develops.