I found this experiment interesting because it doesn’t frame AI as “breaking” a system. Instead, it treats AI as a new kind of participant interacting with infrastructure that was built around human assumptions consistency, behavior, timing, and intent…