What's this about?
"Using AI turns out to be a very tiring mental activity. You have to stand one level above where you normally do — thinking not about 'how do I do this thing' but about 'how does this *class* of thing get done', 'how can it be done better and with less effort'. These are questions very few people previously needed to consider, or did consider. It is like having solved the problem of the writing implement, and now having to think about how to write well."
Why is that level harder?
Because it is methodology rather than execution, and methodology has no feedback loop. When you do the work yourself, you find out immediately whether your approach worked. When you direct an agent, you are choosing an approach and then evaluating an outcome produced by a process you did not observe — so the connection between decision and result is much weaker, and you have to reason about it rather than feel it. That reasoning is the tiring part, and most people have never had to do it explicitly.
Why is it worth noticing?
- The writing-implement analogy is exact: the tool problem is solved, and what is left is the hard part that was always there.
- It explains the widely-reported experience of AI-assisted work being simultaneously faster and more exhausting.
- It is the individual-level version of dotey's argument about technical judgement being what now separates people.