Earlier quoted context omitted.
TBH if you're creating a new language after 2014 you should be thinking about how to get good syntax highlighting and other tooling from the language definition itself, because you're competing with a ton of mature languages that had to do this manually through a ton of effort of their communities.
How do you get it from the language definition? By adding metadata to it?
It doesn't work in practice, since most languages eventually discard parser generators for custom parsing, and good tooling like syntax highlighters don't rely on the literal grammar but something close enough to the grammar to work and still be useful for programmers.
Classic example is you don't want your parser to be strictly defined and fail at any error. You want it to handle errors gracefully and detect almost correct syntax.
Having written some textmate grammars for syntax highlighting I think the post is ridiculous. I'd be terrified to generate that json from the parser, since a bit of thought and experimentation has to go into getting a good syntax highlighter that works with external tooling.
As well I think an LSP implementation is really important, but writing that in your language or generating it from the compiler is a monumental task for a young language and a terrible idea.