Iteration 6 — H0 frozen at merge, branch rebased, import complete: 50 entries

Iteration 6 — H0 frozen · branch rebased · import complete 12 IMPORTED · 50 TOTAL RULEBOOK HASH 68596b5a

2026-07-07 · sixth firing · ledger rows 10 (NOTE) + 11 (HARVEST) · import batch I-02 · campaign branch loop/matrix-harvest-campaign

In plain words: a milestone firing on two fronts. First, the constitution is now carved in stone: the operator merged the scaffold PR, and the squash commit 68596b5a is now the permanent, tamper-evident fingerprint of the campaign's frozen rulebook — every future claim of "we followed the pre-registered rules" points at that hash. The campaign's working branch was rebased onto it (277 mechanical conflicts in auto-generated navigation files, resolved and regenerated; every research record byte-checked afterwards). Along the way two incidents went on the record: the live dashboard was briefly wiped by a deploy race minutes before the merge (detected, redeployed, root-caused — and permanently fixed BY the merge itself), and the navigation now carries the new Evaluation hub from main. Second, the attic is empty: the May catalog's last twelve importable instruments re-earned their places — conditional-independence tests, information decomposition, factor-structure stability, and two multiple-testing procedures built for dependent data. Two entered and were immediately shown the door for having no working code anywhere (the rule is merciless: no implementation, no candidacy) — honest entries, honest exits. The catalog stands at 50 of 300, every boundary now has at least one verified instrument, and the FAIL boundary is two entries short of its first interrogation batch.

The lifecycle events (ledger row 10)

2026-07-07 00:01:33 UTC bigblack deploy (rsync --delete, main-based) wipes campaign pages └─ detected on operator link request → redeployed → origin-verified 200 2026-07-07 00:06:03 UTC operator merges scaffold PR #581 (squash) └─ 68596b5a = FROZEN H0 pre-registration hash (row-2 plan executed) this firing campaign branch rebased onto main → force-pushed (PR #582 OPEN) └─ 277 nav-HTML conflicts → replayed side taken → nav regenerated LAST └─ wipe race PERMANENTLY FIXED: main now carries the campaign folder

Import batch I-02 — the last of the May attic

id	Instrument	Boundary	Note
H-039	Generalised Covariance Measure (CRAN)	FAIL	hardness-aware CI test; June in-project evaluation noted (WATCH, different context)
H-040	QR partial copula (JMLR 2021)	FAIL	SCREENED-OUT(NO-IMPL) — no code anywhere; pacotest recorded as the implemented neighbor lead
H-041	BROJA-2PID (Apache-2)	FAIL	first partial-information-decomposition instrument
H-042	O-information (hoi, BSD-3)	FAIL	synergy/redundancy sign; June evaluation noted
H-043	Ledoit–Wolf shrinkage (sklearn)	FRAGILE↔COND	NOT the boneyarded MP thresholding — distinction on record
H-044	Tucker's congruence φ (psych, active 2026)	FRAGILE↔COND	factor-loading similarity across regimes
H-045	RV coefficient (hoggorm)	FRAGILE↔COND	whole-matrix configuration similarity
H-046	Romano–Wolf stepdown (wildrwolf)	COND-scoping	FWER under dependence — the "dependence-tighter" slot G11 lacks
H-047	Benjamini–Yekutieli (statsmodels)	COND-scoping	FDR valid under ARBITRARY dependence (BH needs PRDS)
H-048	Robust distance covariance (2024)	FAIL	SCREENED-OUT(NO-IMPL) — author-code claim did not re-verify; re-check quarterly
H-049	CRQA (PyRQA, Apache-2)	FRAGILE↔COND	zero distributional assumptions — phase-space coupling stability
H-050	Effective Transfer Entropy (CRAN)	FAIL	finance-native shuffle-corrected TE; complements H-036's closed-form null

Instrument

Boundary

Note

H-039

Generalised Covariance Measure (CRAN)

FAIL

hardness-aware CI test; June in-project evaluation noted (WATCH, different context)

H-040

QR partial copula (JMLR 2021)

FAIL

SCREENED-OUT(NO-IMPL) — no code anywhere; pacotest recorded as the implemented neighbor lead

H-041

BROJA-2PID (Apache-2)

FAIL

first partial-information-decomposition instrument

H-042

O-information (hoi, BSD-3)

FAIL

synergy/redundancy sign; June evaluation noted

H-043

Ledoit–Wolf shrinkage (sklearn)

FRAGILE↔COND

NOT the boneyarded MP thresholding — distinction on record

H-044

Tucker's congruence φ (psych, active 2026)

FRAGILE↔COND

factor-loading similarity across regimes

H-045

RV coefficient (hoggorm)

FRAGILE↔COND

whole-matrix configuration similarity

H-046

Romano–Wolf stepdown (wildrwolf)

COND-scoping

FWER under dependence — the "dependence-tighter" slot G11 lacks

H-047

Benjamini–Yekutieli (statsmodels)

COND-scoping

FDR valid under ARBITRARY dependence (BH needs PRDS)

H-048

Robust distance covariance (2024)

FAIL

SCREENED-OUT(NO-IMPL) — author-code claim did not re-verify; re-check quarterly

H-049

CRQA (PyRQA, Apache-2)

FRAGILE↔COND

zero distributional assumptions — phase-space coupling stability

H-050

Effective Transfer Entropy (CRAN)

FAIL

finance-native shuffle-corrected TE; complements H-036's closed-form null

Kills: RIE (MP-lineage), CCIT (classifier-CI, RCIT precedent), standalone KSG-MI (already in Tier-2 production). Scope call: predecessor C2 portfolio frameworks (HRP/cMDA) = decision layers, not status instruments — out of scope, on the record.

Coverage counters after this firing — every boundary now populated

Boundary	Minimum	Verified	Screened-pass	Fed	Starvation
STABLE	100	23	10	10 (B-01)	77
FRAGILE↔CONDITIONAL	60	5	0	0	55
CONDITIONAL-scoping	40	3	0	0	37
FAIL	40	12	0	0	28
EXPIRY	40	6	0	0	34
cross-cutting	20	1	0	0	19
Total	300	50	10	10	250

Boundary

Minimum

Verified

Screened-pass

Fed

Starvation

STABLE

100

10 (B-01)

FRAGILE↔CONDITIONAL

CONDITIONAL-scoping

FAIL

EXPIRY

cross-cutting

Total

300

250

▶ Next iteration

Iteration 7 = STABLE harvest, family 2 (anytime-valid / e-process monitors) — STABLE remains most-starved (77) and family 2 serves it directly (test martingales, e-detectors, confidence sequences — the SAVI ecosystem around seed entry H-003). Alternative slices if steered: FAIL top-up (2 more entries → 10-entry screen cohort → screen batch S-02), or the fresh WebSearch leads (local Gaussian correlation, KLD instability, sliced Wasserstein).

Merge point (operator): campaign PR #582 is now rebased cleanly on main and contains everything through this iteration — mergeable at any time; the loop will cut the next cycle's branch automatically after a merge (rule 10).

Steering point: no directive = family-2 harvest next firing. Directives are ledger-recorded verbatim before taking effect.

Machine-readable record: ledger rows 10–11 + catalog §I-02 + taxonomy iter-6 leads. Pages are append-only.