I find it shocking and disappointing how rare plainly logical and intelligent reactions like this are.

Not that I would ascribe actual intelligence to "AI", but those fans who claim it's better than your average human almost seem to be proven right by the sheer amount of incredibly stupid behavior around "AI".

I had a situation recently where I had to onboard to a new process, I spent the day analysing the onboarding docs raising PRs working through all the issues we hit using Claude. At the end of the day my Claude was primed with all the knowledge from the days efforts. When a colleague asked me why his onboarding was having a specific error my session was able to quickly debug it and give him an accurate answer. His agent could have absolutely done it but it would have taken (slightly) more effort to get it primed rather than proxying through me to my agent