As large language models (LLMs) expand into audio, progress has been breathtaking. “LLM-native” speech technology reached practical maturity roughly half a year ago, and the entire industry has surged forward. Two features mark this maturity: ultra-realistic speech and full-duplex interaction. A ...