Parallax

Parallax is the research system DisclosureWorks uses to support its work. It turns primary sources into a checkable record, so that sourced claims in our Sourcebook can be traced to the documents that support them. It extracts and fingerprints primary documents, makes the record searchable, ties claims to the pages that support them, and has independent AI reviewers check each change before a person approves it.

Why it exists

Oversight research now runs into more documents than people can read. A single investigation can release millions of pages, and no reader, newsroom or committee staff can keep every claim anchored to the page that supports it, or notice when a later release changes what an earlier page meant.

Doing that work reliably at this scale needs research methods that are automated and repeatable but still directed by people. Parallax is built that way: AI does the heavy reading, sorting and checking, and people set the questions, review the results and approve what is published.

What it does

How we use it

Parallax maintains our knowledge base, the internal working record behind everything we publish. The Sourcebook and our Publications are drawn from it after review and approval. It is built for any documentary record. So far we have used it on national security oversight, through four historical cases and the Jeffrey Epstein case, and it could be applied to other oversight domains.

Availability

Parallax is not yet available outside DisclosureWorks. It will be released for others to use; we have not yet set a date or decided the form of the release. To follow our work, including news about Parallax, follow DisclosureWorks on GitHub.