ok that's cool haha. but i'm genuinely curious, what will the architecture look like, considering we dont need TTS and STT? surely, the latency will be really fast, but i wonder how the latency will compare vs the one from chatgpt app once it releases based on your rough prediction