Cycle 01 — What Changed

Date: 2026-10-12. Evidence cutoff: 939d8dffa2064413f8f6f211f7c358219c07d0f9. D-0010 accepts this dated change record after Synthesis Review 1 and Foundation Review 1. It compares the frozen starting commitments with the retained outcomes and dispositions; it creates no new research data or final Cycle snapshot.

Scientific and operational changes

Starting commitment	Evidence now available	What changed, and what did not
RP002A preregistration predicts useful history in E1, present sufficiency in E0 and future-irrelevant history in E2. B1/B2 comparisons cannot establish universal memory necessity.	The fresh second confirmation has 60 units and 120 learned fits: two benefit, four equivalent and one indeterminate decision under the fixed margins and conditional intervals.	A prospective fixture hypothesis now has scoped, audited algorithm-level evidence. E1 history benefit and E0/E2 equivalence meet their registered rules. E1 B2 versus B1 remains indeterminate for its own benefit threshold; no post-outcome equivalence switch or demonstrated recurrent necessity follows.
RP003 explicitly plans transitive controls, tolerance counterexamples, criterion signatures and path-information boundaries.	All 11,100 declared finite paths replay. Exact local-to-endpoint and local-to-prefix controls have zero failures; tolerance local-to-endpoint has 848/5,396 and local-to-prefix 1,136/5,396 counterexamples. Variation bound passes on all 1,876 eligible paths.	The expected boundaries now have complete finite witnesses and coverage. Local tolerance, endpoint return and initial-reference prefix bounds are distinct. These were planned controls, not an unexpected discovery of a new identity theory.
RP004 plans to distinguish current-state equivalence from lineage relations with R0–R3.	Six constructions produce 216 query truths, 864 decoder rows and 60 pair/projection comparisons. R3 answers all 216 inquiries; R0–R2 have collision witnesses. All non-UNKNOWN answers are sound.	The representation requirement is explicit for the chosen inquiry family: some same-payload/projection constructions require further genealogy or birth-kind distinctions. R3 sufficiency is audited; minimality, arbitrary-domain sufficiency and universal full-history necessity remain unproved.
RP005A locks structural inclusion under pure edge removal and asks about horizon-limited action selection.	All 24 planner rows, 12 pairs and 35 source certificates replay. Two positive, four zero and six negative differences remain; both raw and B=3 inclusion pass.	Conditional steering now has designed, reproducible examples and an exact selection/deletion decomposition. No structural reachability expansion occurs. Sufficient horizon removes V1's benefit here; V2 and C0 preserve distinct sources of harm.

The change is from a declared research design to bounded, reproducible evidence and explicit package dispositions. The main qualitative expectations were present before execution. Deterministic fixture counts are coverage, not empirical frequencies; RP002A's conditional inferential units are not pooled with path, decoder or planner rows. No universal theory, novelty clearance or cross-domain transfer theorem is introduced.

Beliefs retained and narrowed

Retain the contract principle that preservation and representation judgments require an explicit inquiry and comparison context. The examples now make several boundaries concrete: adjacent steps do not determine initial-reference excursion; current payload does not determine all genealogy answers; a restriction's effect on selected endpoints is separate from fixed-source graph inclusion.

These are scoped interpretations. History helps in the declared aliased predictive world, but history varying in E2 does not automatically help future prediction. R3's success does not prove it is the smallest encoding. UNKNOWN decoder responses do not all establish information loss. V1's supplied full-information rule does not establish learned-rule benefit, indefinite safety or a normative need for restrictions. Q-4's possible reduction to Q-3 remains unproved.

Foundation Review 1 retains PC01–03 FROZEN, PC04–06 DRAFT and the existing formal-capability contract. No recurring formal deficit or discriminating test among parked foundation candidates was established. Retaining the current baseline is an explicit bounded adequacy judgment, not evidence that it is final or optimal.

Status and governance changes

RP-002A, RP-003, RP-004 and RP-005A now have CLOSED finite package dispositions under D-0004–D-0007. Their four scoped Results remain VALID. The original pre-execution registrations and audit-stage reports keep their historical states; current disposition records provide the later change. Q-0/Q-2/Q-3/Q-4 remain ACTIVE and Q-1/Q-5 OPEN. There are still no Claim or Assumption instances in the graph.

D-0008 completed an actual late Checkpoint A and graph review; D-0009 completed synthesis/foundation reviews. The graph review reconstructs its own immutable 33-record cutoff, while the living graph had 35 before this stage and has 36 after D-0010. Typed references and semantic relations remain separate. The checkpoint tag remains an immutable historical anchor, not a backdated execution gate.

Two process facts remain canonical. The first RP002A attempt is INCOMPLETE and has no Result; the separately authorized second attempt reused no samples/checkpoints and preserves its distinct timing record. Checkpoint A was missed before RP004/RP005A and later reviewed under an explicit timing deviation. Neither the successful retry nor the late review erases the original failure. The transition and close bundle must retain both records and make future execution gates explicit before execution.

Publication and remaining commitments

Both English sites now project the shared repository's package outcomes, deviations and review records. The ring measures resolved obligations, not scientific truth, question completion or time. Site source/version/deployment IDs are separately bound to a checked content commit in docs/operations/SITE_DEPLOYMENTS.json after each successful publication.

D-0010 resolves only What Changed and What Remains Open: Cycle 01 stays ACTIVE at 10/13 (76.9%). DAM X pilot, transition/Cycle 2 decision and close bundle remain unresolved. DAM X has a template and source-dependency README, not supplied historical source evidence or a survival judgment. No new baseline version, final snapshot, close tag, Cycle 2 commitment or new experiment is created.

Revisit this dated interpretation through a new Decision if evidence invalidation, a material method error, a mistaken comparison or source-backed new work changes it. Preserve original bytes and record the subsequent correction separately.

Traceable references

Source record

Repository path: cycles/cycle-01/reviews/WHAT_CHANGED.md

SHA-256: 8921d1aba90e3b2d9f55f42648e715a584dcd25cd3cdecb1420a111d24321941

This reading view is generated from the source. The downloaded Markdown preserves the original bytes.