“Mati Staniszewski believes the breakthrough for real-time, production-ready AI voice models occurred approximately one year ago.”