"Language composed of grammar" excludes the aspect of diachronic and synchronic variation from language. To quote Sapir: "Language flows down in the current of it's own making". Like Hilbert's program, Chomsky's program of generative grammar was a disastrous failure. The abstract modelling felt apart in the first contact with empirical reality of language. Since that, Chomsky's scientific approach to general linguistics has been just fancy words reducing to "brain reduction". To be fair, his philosopher side is not as bad as his ritual of scientism in his own academic field.
Gödel proved that static models of language can't be complete, hence trying to
force a static model over dynamic process leads to inconsistency. Which is why
this is
complete BS.
To get a better grasp, compare how both-and and neither-nor behave in static and dynamic tetralemmas.
STATIC
1) True
2) Not true
3) Both true and not true
4) Neither true nor not true
DYNAMIC
1) Increases
2) Decreases
3) Both increases and decreases
4) Neither increases nor decreases
For a more general shorthand, various qualitative verb-polarities can be replaced with
more and
less, ie. relational operators < and >. As long as we keep in mind that they really mean verbs, not nouns. Static models can be derived from the 4th lemma of the dynamic, but there is no consistent or coherent way to derive other way around. There is a dependence hierarchy, static depends from dynamic, but the dynamic does not necessitate the static.