Bob learned how to use the agent to do work, which is just as useful to society. I wrote a short script to tproxy my connection using an AI and it took less time than doing it myself. More importantly, knowing it took less time gave me the motivation to actually ask the AI to write it

This is an uncomfortably well written and thought out article. Its also helped me put into words some of my own conceptual reluctance to fully embrace LLMs in my work. I know how I've learned from tackling problems on my own in my career. I worry that the seductive siren of the "right answer" LLM without the effort may be too strong for me to resist or too subtle for me to notice.

Whats the answer? Become the constant contrarian never able to trust anything these output? At that point the value of using the LLM at all is erased.

Bob learned how to use the agent to do work, which is just as useful to society.

Bob used the agent to do the work of a junior and it cost him the ability to learn how to do it himself as a senior in the future.