Among the October 23 UAP releases, the December 2008 Project Capella document is worth reading as a research plan. The government-listed D167 file, available in a readable archival mirror, describes how to organize a UAP information system—and how to prevent its contents from acquiring more certainty than their sources justify.

A plan for checking the database

The report, prepared for Bigelow Aerospace Advanced Space Studies, proposed common structures across separate catalogs rather than immediately collapsing every collection into a single chronology. It warned that case selection could favor a preferred hypothesis.

Its staffing proposal placed a quality coordinator outside the other data team’s reporting structure. That role would check source material, coding and completeness, with independence intended to protect critical review. The relevant discussion appears on PDF page 10, numbered page 9.

The plan also favored a practical initial spreadsheet platform. More elaborate analytical capabilities were a later possibility, not evidence that a functioning automated discovery system already existed. A proposal specifies intended work; it does not demonstrate that every position was filled or every check performed.

New access, older public history

Capella was not first revealed this month. In a 2014 paper hosted by CNES/GEIPAN, Jacques Vallée described a UAP data warehouse containing eleven separate databases under that name. He discussed inconsistent archival standards and the value of linking case records with supporting material.

That earlier publication provides a separate, dated record of the project’s public history. It is a participant’s account, not an independent audit of the warehouse’s contents. Read together, the sources establish continuity between the planning document and a later public description without establishing that every planned safeguard was implemented.

Why this matters for UAP evidence

A large catalog is not automatically a large set of independent observations. Reprints, translated reports and retrospective interviews may all refer to one event. Analysts need to preserve that relationship or risk counting repetition as corroboration.

Useful follow-up questions are therefore operational: Can a record be traced to its original source? Are estimated quantities distinguished from measurements? Are revisions retained? Can another analyst reproduce the classification? The release makes those questions more concrete. It does not turn the underlying reports into proof of any particular origin.

Hero image: original AI-generated editorial illustration. It does not depict evidence from the reported events.