Yes, very interesting point. It can not know that they are its own rules. BUT we can know that. Therefore we the meta-level can reason about the subject more th

Yes, very interesting point. It can not know that they are its own rules. BUT we can know that. Therefore we the meta-level can reason about the subject more than it can of itself. For example the subject-system can not mechanically prove that it is looking at its own ruleset. So the term "unprovable" should probably best be always qualified with: in which system. There are facts about systems which the system itself can not prove. Statements like that can be expressed in some formalized subset of English and proven following the rules of that formalized English. What is that "formalized English" that we use in our reasoning? Or does it not matter?