Tag
OpenAI introduces GPT‑Live‑1 in the API, a natural voice model that enables developers to build voice-enabled apps with smooth interruptions, tone customization, and reasoning delegation to other models.
The tweet highlights a hidden feature '/btw' in Hermes AI that enables handling interruptions without disrupting current AI agent tasks.
An analysis of 10,000 voice AI calls reveals key failure points: STT error rates, first-8-second chaos, interruption handling, extended silence, tool call latency, LLM failures, and broken escalation — offering practical insights for agent builders.
This paper identifies 'state inertia' in full-duplex spoken language models, where the model's internal predictive focus lags during user interruptions, and proposes a training-free activation steering method to improve interruption handling.