indigo interprets Addy's view from the Claude Code team on what humans should think about when agents are working

indigo cites Addy from the Claude Code team on maintaining critical thinking when agents handle tasks. The article notes that as models grow stronger, critical thinking becomes harder; humans risk a three-step cognitive slide—unloading, delaying, and surrendering—gradually losing control. It recommends placing judgment at both ends of the task: defining assumptions and completion criteria beforehand, letting verification enter the execution loop during the process, checking hidden assumptions against evidence after the fact, and asking before delivery whether one can explain and take responsibility.