OpenAI’s new real-time transcription collapses voice latency toward zero — and exposes your playback hardware as the bottleneck.

Google’s Lyria 3.5 can compose almost anything. The last 30 centimeters between your speaker and your ears decide whether you actually hear it.