you mean turing complete?

the point is the tradeoff between "the ability to say" and "the ability to get answers".
turing completeness means maximum expressiveness, can say anything, but even things such that it'll be impossible to get all answers about, even in theory.
however if we restrict ourselves only to settings in which we can answer all relevant questions, we have to restrict what we're able to say.

the following two blogposts contain some explanations about the proposed alternative (as there can be many non-turing-complete languages):
http://www.idni.org/blog/the-art-of-self-reference.html
http://www.idni.org/blog/the-new-tau.html