I let two local LLMs fight over how to optimize a Linux VM, and they destroyed it instead
… I ran the experiment in a few different framings, including a generic "optimize this machine" prompt and one that pointed both models at security hardening specifically. …