@mgorny You are not wrong. The article is great. I have made the bottleneck argument before. The erosion of expertise to an LLM eating its own garbage output is a huge worry for not only the experts but for us all.
I'm saying it would be great to have a situation where the bottleneck discussion could be the focus rather than being distracted by the water use and provenance discussion, or to have a situation where we don't even need to discuss where the bottleneck is because the LLM is startlingly obviously not returning value.