So many things wrong with this.


I am not a programmer by trade, and even though I learned programming in school, it’s not a thing I want to spend a lot of time doing, so I do use AI when I need to generate code.


But I have a few HARD rules.




I execute all code and commands. Nothing gets to run on my system without me.




Anything which can be even remotely destructive, must be flagged and not even shown to me, until I agree to the risk.




All information and commands must be verifiable by sourcing documentary links, or providing context links that I can peruse. If documentary evidence is not available, it must provide a rationale why I should execute what it generates.




Every command must be accompanied by a description of what the command will do, what each flag means, and what the expected outcome is.




I am the final authority on all matters. It is allowed to make suggestions, but never changes without my approval.




Without these constraints, I won’t trust it. Even then, I read all of the code it generates and verify it myself, so in the end, if it blows something up, I bear sole responsibility.