I would like to see more articles about agent sandboxes. With agents gaining popularity we need a higher fraction of users to understand containers and sandboxes and their risk profiles, and then to communicate their understandings to friends and family. It is a harder task than explaining ChatGPT, and it often feels like a hindrance.