While that makes it about 10 years younger than Postgres and MySQL, it's still not a flash in the pan.
Node has encouraged MongoDB usage because there were nice packages that made working with Mongo very easy. If Postgres supports JSON with powerful features then developers will make npm packages that support it and this will make Postgres very attractive.
And even if you don't like MongoDb or Node, JSON is clearly one of the most popular data formats - better support is surely a good thing.
But this is different because the postgres architecture is very much capable of great support here, and getting even better. Postgres has the best support for adding in new types, with amazing capabilities around indexing.
I'm not just talking about different kinds of BTrees, I mean real non-scalar indexing strategies like inverted indexes (used for full text, but also great for JSON), similarity matching indexes, unbalanced indexes, spatial indexes, etc.
The author merely wants to use this infrastructure as intended for the JSON use case. Easier said than done, of course, but no architectural compromises need to be made in the process of providing this support.