I'm the author. Sorry for lack of context for people not familiar with Sourcegraph.
This blog post outlines changes to the backend services that enable precise code intelligence over the last year. Code intelligence in this context includes things like hover text, definitions, and references for a position in a source file, compiler/linting diagnostics for a file or directory, etc. To see it in action, try hovering over identifiers in this repository: https://sourcegraph.com/github.com/sourcegraph/sourcegraph/-...
Sourcegraph has search-based code intelligence, which works with over 30 languages without configuration, but is heuristic and not always accurate (for example, jump-to-definition may jump to a definition with the same name, but not the correct one). We had support for running language servers to provide _precise_ code intelligence, but we moved away from that strategy due to speed and resource concerns. Since then we've moved to pre-indexing source code to provide precise code intelligence, which is where we're at today (and what the blog post is about). The original announcement of these features can be found at https://about.sourcegraph.com/blog/code-intelligence-with-ls....
