@davidrevoy

It's always these inexperienced people. Somewhere in the 1,000-page manual, there's bound to be a hidden note for this case: "Never let your AI run autonomously without AI.md attached. In this rule file, you write down fixed principles that your AI must adhere to."

This file may contain something like: Before each action, assess whether this action has potentially harmful consequences. Only carry out the action if two independent checks have not predicted any harmful consequences