How come the LLMs cannot figure out interacting with command line programs and HTTP services with API docs, as humans can?
It seems counterproductive that we need to make the whole world adapt to these new super intelligences instead of them just figuring it out.