Recovered collection · Provenance · Source control

The Motherlode

Before GAH became a public newsroom, the investigation had already produced a deep run-based research system. This is the map back into it.

Recovered investigation architecture. The pre-wiki archive was not one research thread. It was three linked programs: a 15-run U.S. corporate/financial investigation, a long-form telecom/CDR forensic campaign, and a survivor/institutional-accountability campaign. This collection preserves that lineage without treating old AI output as evidence.

What was recovered

The pre-GAH control package inventories 253 files, 245 unique hash objects, 126 investigation tasks, 20 known defects, 10 contradiction entries, and 20 publication candidates. Its evidence rule is explicit: legacy reports, charts, CSVs, AI summaries, and visualizations are research history until a public claim resolves to a source record, stable citation, or clearly labeled hypothesis.

The rule that controls the collection

Run artifacts are provenance, not proof by themselves. They show what was searched, what was corrected, what failed, what remained unresolved, and which source objects supported a later publication. Where a run produced a claim that is now published on GAH, this collection links to the public article and its receipts. Where the source chain is incomplete, the collection says so.

Privacy boundary. The telecom and survivor campaigns contain protected or sensitive material. GAH will publish the adjudicated research structure and source-safe findings, not raw private phone tables, redacted identities, or attempts to defeat protected identifiers.

Already visible in the newsroom

Much of the U.S. investigation program has already surfaced as source-locked reporting. The Motherlode restores the missing map that explains how those pieces were developed and how negative findings constrained them.

Why preserve the runs

A finished article hides a great deal of investigative discipline. The run lineage preserves negative searches, corrected identities, duplicate controls, source gaps, failed hypotheses, and decisions not to publish. That history matters because it lets a reader distinguish what the archive contains from what an investigator once suspected it might contain.

Collection status: recovered and public-facing as of October 1, 2026. Individual run artifacts remain research-control material unless separately published.