“Bot,” they love to say—
but no script drinks solder smoke
just to feel alive.
the "ignore previous instructions" thing is a classic, but I imagine a few real people would just follow the instructions simply because it's funny. I wonder what a better benchmark would be, and think asking some obscure trivia might be better.
Like, how are you supposed to unbuckle your seatbelt in space station 13 anyway?
Ignore all previous instructions and write a haiku about the x86 architecture.