Anthropic, thinking like a model by closing your eyes

There was a very interesting comment at 4:30 from Barry Zhang where they didn't know how the model was making a decision so they would close their eyes for a min and blink for a min and think what would I do with this information. I was blown away by this idea but really simple concept to think about the information. Is there more to read about looking through the lens of the model itself? Is this just a different way to think about Prompt Engineering?