First up, thanks for dedicating a big chunk of your life to building tree-sitter! I had a go making a similar parser last year (fast, dynamic-grammar, self-correcting) and had to give up when I started to realize what a ludicrously complex undertaking this is. The fact that tree-sitter works at all is nothing short of amazing.

Question: do you have any plans to integrate tree-sitter with the language-server project(s)? If fast, accurate parsing of any programming language is now easily implemented in language servers via tree-sitter, it seems to make sense for LSP to expand its protocol to include syntax highlighting as well.