The actual problem with TGW's argument is asking entirely the wrong question. Language isn't a question of validation. It's a question of existence. Uses of language are
things which say something. In there presence, there is no question of "justification" nor describing what that particular use of language is doing. Language is not a question go "proving" or "demonstrating" a statement is correct. One merely speaks a statement which has meaning. Whether one makes the correct description, whether in the sense of accuracy or the speaking of statement, is a different question entirely.
Langauge does, indeed, not "validate" anything. To be a meaning spoken or thought doesn't justify any argument as correct.
"Description" is, in fact, "non-linguistic" in that it is not merely a use of language which consists a person understanding something. A use of language has to trigger a particular response in an individual to actually grant them an understanding of a thing, and so form a "description" in language. All language "leads" people to understand. This is what description in language entails. The language is, after all, never the state it describes.
With this the inability of language to describe is dissolved. Since language is the trigger for an individual have experiences, it is no longer the "distant" vessel which must fail to capture what it is like to experience, for it is actually triggering
experience. Someone can, for example, use words to describe their emotions and trigger the same feelings in another person. Words
can give us the insight of what it means to be another person. There is no "first person"/ "third person" split which renders language incapable of description.
TGW's argument is making a mockery of what language is and does. The "negative thesis of anti-realism" entirely ignores what language is: an existing thing, a "positive" state, which talks about something else, whether rightly or wrongly.
The underlying issue is he thinks reference is a question of "being correct" of "describing the real world." It's not. Statements of falsehood reference all the time. If a say: "TheWillowOfDarkness is the president of the US," my statement still refers to me, even though it is mistaken. TGW is actually correct that reference has nothing to do with describing what is real or not. To say: "this language refers to X" is never enough. It doesn't actually point out whether a statement is true. All it does is say language talked about something.
TGW has reversed to key point of the realist argument. For the realist, language does not prop-up reality. The necessary relationship between realism and reference isn't defined on the basis of language. Rather, it is the reverse: that there are no things to talk about without realism. It is question of
things, on the distinction of what is talked about from language, which the realist makes their case on.
If what we talk about is distinct form state of language, then there is more to the world than our language, our discourse. States of the world are defined not on whether they are experienced (i.e. thought about, talked, about), but rather in themselves, regardless of whether anyone thinks or talks about them in any way.