great pod for learning how this kind of testing actually works, also miles made a really good point that it may be hard to "pull the plug" on a data center or whatever to prevent AI from doing something bad if you're already using AI throughout a bunch of critical systems