Great point from Jim Neilsen re: the instructions found deep inside a model or agent’s instructions, which we may not prefer yet live with if we don’t know to override them: It’s like a Tr
