Themes: Gemini live audio, speech as an API surface
Detail
Gemini live audio
The one substantive item today is an external hands-on write-up of Gemini's Live audio mode — the real-time, speech-in/speech-out path rather than the text-first API. The reason this is worth logging even as a single data point: live audio is the part of the Gemini stack that gets exercised least by developer chatter and most by actual usage, so independent write-ups are the main signal on whether the latency, interruption handling, and voice quality hold up outside of demo conditions. It also sits squarely on the Gemini API / AI Studio surface, meaning changes here show up as developer-facing primitives rather than as a consumer app feature — worth watching for whether streaming audio gets treated as a first-class modality in tooling and agent loops, or stays a separate product mode.
Sources: Simon Willison — Gemini Live audio.