I think it’s much more interesting that the user can write “brah” and the model can interpret that as “you are continuing to make the same mistake and it’s dumb

I think it’s much more interesting that the user can write “brah” and the model can interpret that as “you are continuing to make the same mistake and it’s dumb” and then follow suit with a weird role play of trying to hack itself.