The comparison of our current methods of interacting with LLMs (back and forth text) to old-school terminals is pretty interesting. I think there's still a lot work to be done to optimize how we interact with these models, especially for non-dev consumers.