Great overview. I wonder if we'll start to see more text diffusion models from other players, or maybe even a mixture of diffusion and transformer models alternating roles behind a single UI, depending on the context and request.