Whonym

My role
Solo: design and build.
Year
2026
Type
iPhone and Apple Watch app
Status
On TestFlight

(Work the app)

Tap through its screens. In the live session, tap an introduction to name that voice.

When a class or meeting starts, it offers to take notes.

Thursday, Nov 12

Whonym

Now · Meeting

Design review

Studio 3

A recreation of the app’s screens in HTML, from its SwiftUI code and colours: not screenshots. Every name, place and line is sample data.

Status
On TestFlight. I’m building it on my own.
Next
Before I share the app with anyone, its cloud calls would move behind a small server; a public release would also need Google’s security review for Gmail.

Whonym is a note-taker that remembers people, not just meetings. It tells voices apart, learns a name the moment someone introduces themselves, and remembers who I met, when and where. I’m building it on my own, for iPhone and Apple Watch.

Limits

iOS keeps Messages history from apps, so Whonym drafts replies only to texts I paste or share, and no app can start the iPhone microphone from the background.

The look

An index card: paper, ruled lines, a red line down the left and a highlighter for each speaker.

The Whonym app icon: a cream index card, tilted, with pale blue rules, a red line down the left and one line highlighted in yellow, on a dark ground

The app icon: one index card, one line highlighted.

How it fits together

Five-layer diagram: the Watch or iPhone microphone feeds on-device captions and speaker labels on the phone; introductions tie names to voices; after the session, optional cloud models re-check speakers and draft follow-ups; everything ends in the people memoryLive work stays on the phone; the cloud re-check and the drafts come after the session, with my own keys, and are optional. The Watch only records and relays.CaptureWatch or iPhone microphoneRaw audio, not noise-suppressed:that would hurt speaker matchingAudioLive, on the phoneApple speech, FluidAudio (CoreML)CaptionsWho is talkingSpeaker labelsNamesThree introduction cuesA name tied to a voiceLabelled transcriptAfter the sessionpyannoteAI, Claude; optionalSpeakers re-checkedNames checked, follow-ups draftedNotes and draftsPeople memoryWho, when, where, what was said

Live work stays on the phone; the cloud re-check and the drafts come after the session, with my own keys, and are optional. The Watch only records and relays.

Facts

My role
Solo: design and build
Company
Tully Tech
Runs on
iPhone and Apple Watch, iOS 26 and watchOS 26 or later
On the phone
Apple’s on-device speech recognition; FluidAudio (CoreML) for who is talking
After a session
pyannoteAI re-checks speakers; Claude checks names and drafts. Both optional, with my own keys
Status
On TestFlight