Comment by rickydroll

Well, if you throw a long-winded prompt over a fence and tell the agents to work it out on their own, yeah, you will spend a lot of time reviewing it. But if you use it as a form of pair programming and keep your scope of focus constrained, review time is built into your effort.

Also, keep in mind that my perspective is colored by the fact that I can't type much, and coding by keyboard is pretty much off the table for me.

And last, as I pointed out elsewhere, LLMs are useful for more things than writing code, and administrative work can be handled easily with the current generation of LLMs.