I don't read all of the code produced by my agents any more, but I like to reserve the ability to do so if I run into a particularly confusing bug, or for any code that's security adjacent.

Same. But usually if I need to read code, I end up telling my agent to summarize it for me.