WIRED | 2026-08-27

Relay, a startup founded by two former Nothing executives, is debuting its software today alongside the announcement of the Relay Q — a dedicated AI voice-input microphone that won't be available until early 2027. The founders, who used Wispr Flow's transcription at Nothing, built Relay after hitting a pain point: "In the office, we never really could get the right volume for input," CEO Zhu told WIRED — whispering too quietly for the Mac to pick up, talking too loud for coworkers to ignore. "It calls for dedicated hardware." The macOS app works with or without the dedicated mic: press and hold a hotkey (default Fn) in any text field and Relay activates. The voice-to-text engine is powered by Google's Gemini models. It also performs contextual actions dubbed "Skills" — asking it to message a colleague in Slack prompts Relay to launch the app, find the recipient, and draft the text; it can also craft calendar events in Google Calendar, with preset skills for WhatsApp and Gmail and user-defined skills for other apps. That automation requires sweeping system permissions: Relay needs microphone access and can record a snapshot of the screen anytime it's activated. The company says it stores data locally, where users can edit or delete it; when data is sent for processing it passes through Relay's servers in transit to Google's servers, but Relay does not retain it. WIRED's beta testing found the transcription isn't quite as polished as Google's Rambler feature on the Pixel 11 series, with occasional missed words and punctuation errors, though it handles self-corrections and emoji insertion. Relay is macOS-exclusive at launch, with Windows, iOS, and Android planned.

Read more