By definition, any formal system (machine, language, etc.) that can compute (simulate) any Turing machine or its equivalent (lambda calculus, recursive functions, etc.) is Turing complete.
I wonder if this property of the system can also be reduced to self-universality, that is, when the system can compute itself. As an example, consider a programming language in which one can create an interpreter program that computes (interprets) any valid string of that very language given on its input.
Would this property automatically make the system Turing complete? Or can one think of a system that is powerful enough to be self-universal but not enough to be Turing complete? Could you provide an example of such a system, please?