I learned very early the difference between knowing the name of something and knowing something. Richard Feynman.
We have to develop new languages to cope with the growing complexity of scientific knwoledge, a more strategic approach than the "deep thoughts/self-referenced" paradigm. I am talking about a metalanguage such as the "object oriented" in computer programming. This is a fundamental problem in modern science both, at the educational and research level. It is also true that innovation in science comes often from the ability to deconstruct a language and this demands to understand lower grammars, for this reason this is just a complementary approach, but one which is in greater need now, and in the ages to come.
Approaching the problem of language creation  from a metalanguage perspective allows to rediscover the role of semantics, proscribed by the former paradigms of knwoledge (see above, citation of Feynman). Semantics emerges as a social meaning articulating conceptual objects  which are often -specially in science and technology- related at a more logical level. But this latter level, although critical in the moment of first articulation becomes a burden to the coming articulations which response to a more open demand, what we could call a social demand. There is an institutional role in safeguarding the fidelility between the first -logically based- relation and the subsequent -semantic based- relations. To some extent this is the duty of orthodoxy. How I imagine this new paradigm for the individual is as a complementarity between low grammar skills, limited to a small but highly specialized domain of knowledge, and metalanguage skills allowing to connect different fields of low grammar skills in a superficial but still, inteligent way. There will be though, those for which the metalanguage will be their focus i.e. their low grammar (following Deleuze's "the whole is also a piece").
 By creation we have to consider both procesess: (i) "inteligent design" understood as large contributions from a centralized location (f.e. Facebook's Social Network Protocols, Windows' OS, Apple's GUIs, etc.), and (ii) "emergent", as the aggregate of descentralized contributions (f.e. Linux, Wikis, etc.).
 This process, itself a structured system, allows the gearing of social practices.