Nothing new here yet — notes on speech, audio LLMs, and building real-time voice systems are on the way.