Today, my new walking friend told me that she was writing a book — with the Claude chatbot.

Needless to say, I was horrified.

She offered some examples of how Claude is asking her questions that is forcing her to think about what she's writing and expand on it. When I told her that Claude doesn't actually understand any of the discussion, she said she knows — but then offered me another example.

I don't know what to tell these people. They are so sold on AI systems.

@mlanger there is something to be said for what is known as rubber ducky debugging. But most people that are gung ho on LLMs just haven't used them enough, or hard enough to see the many many issues.