Read this before you trust an answer.
Answers here are written by a language model (Claude Haiku) from document excerpts the search index picked out. That means two things. It can miss documents, because it only sees what the search found. And like any language model it can state something confidently that the documents do not actually say.
Treat every answer as a pointer to the sources, not as a finding. Open the documents and check. If an answer and a document disagree, the document is right.
For exhaustive or precise work, use the
full boolean search instead, which shows you the corpus directly with nothing in between.