This is really impressive engineering. I thought real time agents would completely change the way we're going to interact with large models but it would take 1~2 more years. I wonder what kind of new techs are developed to enable this, but OpenAI is fairly secretive so we won't be able to know their sauce.
On the other hand, this also feels like a signal that reasoning capability has probably already been plateaued at GPT-4 level and OpenAI knew it so they decided to focus on research that matters to delivering product engineering rather than long-term research to unlock further general (super)intelligence.
On the other hand, this also feels like a signal that reasoning capability has probably already been plateaued at GPT-4 level and OpenAI knew it so they decided to focus on research that matters to delivering product engineering rather than long-term research to unlock further general (super)intelligence.