> The thing is, agents aren’t going away. So if Bob can do things with agents, he can do things.

But Bob can't do things with agents.

He can get a project from someone else and ask the agents to do that project. Then give the output of the agents back to that someone else, and that someone else reviews it, says why it's wrong, and sends it back. Bob feeds the review to the agents, gets something back, then gives the output back to that someone else who reviews it, etc. So,

1) The loop requires his advisor to know how to go about doing the thing.

2) Bob is absolutely unnecessary and should be discarded.

3) Alice will eventually be qualified to be an advisor.

edit: And the crisis that the article is really pointing out is that when the advisor is using the LLM (while Bob is driving an Uber), and his productivity goes way up because he's only handling the things only he can handle, what about Alice?

Let's say that pre-AI the advisor could either do the job in 2 months or assign it to Alice who could do it in 12 with a week of the advisor's supervision/review. Now, with the LLM, the advisor can do the job in 2 weeks without Alice. Before, Alice made barely any money and had no health insurance. After, Alice is also driving Uber.

Now the advisor has a heart attack and now the thing just can't be done. Also, Ubers become pretty much self-driving, so Bob and Alice are not only ignorant, but unemployed. They can't even afford to take an Uber.