Overall don't disagree with the post too much, but I don't like the example given of speaking to the AI "Charlie".
I work with AI on a daily basis, and frankly I don't think I'll EVER want to just prompt the LLM to do things like craft and send an email autonomously or draft a blog post. Andrej Karpathy's recent talk on programming discussed the generation -> verification loop. For any serious work, you don't want to just rely on AI to generate lots of content (textual, email, etc) without verifying what the AI has done. Maybe in a few decades when we've reached HAL 9000 levels of intelligence and the accuracy is 99.999%, but even then there is some something in me (perhaps neuroticism) where I ALWAYS went to check the content of the email before it is sent. For blog posting, I just want my natural voice to shine, so I would never use an LLM to wholesale write a blog post for me.
But maybe that's just me.
