TLDRocket
11 October 2026
Voice AI’s “ChatGPT moment” still hasn’t arrived, and PolyAI and Otter are arguing that the reason is surprisingly technical: full-duplex gets you talking in real time, but it doesn’t yet deliver the fast, reliable reasoning that makes conversation feel coherent. PolyAI’s CTO Shawn Wen frames the next milestone as latency and thinking speed—plus better transcription context and language accuracy—so the model can actually keep up with what people mean, not just what they say. Even the user-facing details matter now, they say, because disclosure and notifications can’t be an afterthought when the system is effectively always listening.
Read the full briefing →