Cycle 02 — Synthesis Review 1

Date: 2026-10-12. Internal synthesis after D-0017's accepted RP006 result and D-0018's graph review. Evidence cutoff: c0168659a47b4c9f9cd133d19066ea8e145d1d91. This review compares the five finite Results without pooling their unlike denominators or treating graph links as support.

Evidence basis and distinct outcomes

These denominators count different units and must not be summed or reported as an aggregate success rate. Statistical intervals in RP002A do not apply to deterministic enumeration in RP003–RP006. The packages are not one intervention or cross-domain validation study.

Conditional synthesis

The recurring design lesson is to declare the inquiry and its admissible information before assessing a representation, path, planner or procedure. The finite outcomes depend on different contracts: training budget and generator in RP002A; origin, path and tolerance in RP003; genealogy queries in RP004/RP006; and source, horizon and supplied policy in RP005A. This is an internal synthesis of these examples, not a new universal Claim.

RP002A's prediction task and RP004/RP006's lineage task remain distinct. In RP002A, history can change an algorithm's predictive performance under one fixed benchmark. RP004/RP006 ask whether retained construction fields answer specified genealogy queries. The RP006 mask results add no unseen cases and do not show that realized history is reducible to predictive representation. Q-3 and Q-4 remain separate ACTIVE questions.

RP003's local tolerance counterexamples and RP005A's planner outcomes also concern different mechanisms. One compares path-local judgments with endpoint/prefix criteria; the other evaluates finite target coverage after a supplied restriction. Their shared need for declared comparison origins does not establish a common mathematical mechanism or transfer theorem.

Negative outcomes and review limits

Keep RP002A's equivalent and indeterminate decisions, RP003's counterexamples and passing controls, RP004/RP006's conflicts, UNKNOWN and incomparable masks, and RP005A's zero/harmful differences. They delimit each finite commitment rather than disappear in a summary. RP006's independently replayed output establishes exact finite classifications for reused fixtures; it adds no fresh empirical confirmation.

D-0018's graph inventory verifies typed traceability only. It does not infer transitive support. No Claim is promoted, no Question is resolved, and no broad sufficiency, novelty, universal necessity or cross-package unification follows. External peer review has not occurred.

Question implications and disposition

Q-0, Q-2, Q-3 and Q-4 remain ACTIVE; Q-1 and Q-5 remain OPEN. The current evidence does not establish completeness or independence of the six-question set. DAM/DAM X primary-source and survival limits remain separate. Foundation Review 2 separately assesses the frozen contracts and current mathematical capabilities.

D-0019 accepts this synthesis and the separate Foundation Review 2, resolving only Cycle 02's synthesis and foundation obligations. Cycle 02 remains OPEN at 5/9. What Changed, What Remains Open, transition and close bundle remain unfinished. This is not external peer review, baseline amendment, Cycle transition or authorization of another run.

Reopen through a new Decision if a material defect invalidates an accepted Result, changes the registered finite interpretation, or produces a reviewed source-backed transfer result. Preserve this dated synthesis and the original evidence.