โ€บNavigation

Iteration 5 โ€” Predecessor-import batch I-01 12 IMPORTED ยท 8 DEDUP KILLS WEBSEARCH CONDITIONALLY UNBLOCKED

2026-07-06 ยท fifth firing ยท ledger row 9 ยท import batch I-01 ยท campaign branch loop/matrix-harvest-campaign

In plain words: instead of hunting new literature, this firing went shopping in the project's own attic. A May campaign had already catalogued ~95 measurement instruments โ€” before today's rulebook existed. Twelve of its best entries were pulled out, and each one re-earned its place under the current rules: the original paper re-resolved against a journal or archive today, the open-source code re-checked today (license, version, last activity), and every entry passed the four-way duplicate gate. Four strengthen the STABLE front (new causal-discovery machinery built for autocorrelated data), six open the FAIL boundary (independence tests that catch dependence our Spearman gate is blind to), one opens FRAGILEโ†”CONDITIONAL, one deepens EXPIRY. Just as important: eight attic items were rejected at the door โ€” already owned, already killed, or already operational โ€” and each rejection is on the record. Bonus: the hosted web-search channel, blocked for four days, answered one query this firing โ€” it surfaced eight new leads, now parked in the taxonomy for future verification.

The import gate (every row re-verified TODAY, never trusted from May)

 2026-05-16 catalog (~95 instruments, pre-rulebook)
    โ”‚
    โ–ผ  re-verify provenance (arXiv API ร—8 ยท Crossref ร—5) โ€” TODAY's resolution or no row
    โ–ผ  re-check FOSS (PyPI ร—5 ยท CRAN ร—4 ยท gh api ร—5) โ€” license + activity or NONE-FOUND
    โ–ผ  dedup gate (certified stack ยท boneyard ยท this catalog ยท in-project use)
    โ”‚
    โ”œโ”€โ”€ 12 IMPORTED โ†’ CATALOGUED, imported-from: marked
    โ”‚     STABLE +4 : LPCMCI ยท CCM ยท Neural Granger ยท DYNOTEARS      (family 4)
    โ”‚     FAIL   +6 : Hoeffding D ยท ฯ„* ยท dCor ยท dHSIC ยท copula entropy ยท Reduced TE
    โ”‚     EXPIRY +1 : Robust BOCPD (ฮฒ-divergence)                    (family 5)
    โ”‚     FRAGโ†”C +1 : energy distance (no-classifier two-sample)     (family 13)
    โ”‚
    โ””โ”€โ”€ 8 KILLED at dedup (wins, recorded):
          Chatterjee ฮพ (already operational in-project) ยท e-BH (=G11)
          knockoffs ยท Marchenkoโ€“Pastur ยท CODEC/FOCI ยท RCIT/FastKCI (boneyard)
          stability-selection base (CPSS bar) ยท Pearson-family (baseline)

Highlights of the twelve

idInstrumentBoundaryWhy it matters here
H-027LPCMCI (tigramite, GPL-3, 1686โ˜…)STABLEcausal discovery built FOR autocorrelated series with hidden confounders โ€” deepens the H-004 line
H-028Convergent Cross Mapping (pyEDM, active 2026)STABLEthe only state-space coupling instrument in the pipeline; 2024 "causalized" fix removes future-leakage
H-031Hoeffding's D + fast rank computation (CRAN)FAIL (first entries!)catches non-monotone dependence the Spearman gate misses โ€” with an O(n log n) exact algorithm
H-034dHSIC joint independence (CRAN, active 2026-05)FAILjoint (beyond-pairwise) independence โ€” pairwise gates are blind to synergy structures
H-036Reduced Transfer Entropy (PRE 2025)FAILclosed-form null for finite data โ€” no surrogates; built exactly for our ~2K-bars-per-regime constraint
H-038Energy distance (CRAN/PyPI)FRAGILEโ†”COND (first entry!)regime homogeneity WITHOUT training a classifier โ€” sidesteps the exact failure that boneyarded Failing-Loudly

Channel status change

Hosted WebSearch answered a query this firing (regime-stability phrasing) after four days of filter rejections โ€” the block is query-dependent, not absolute. Recorded in ledger row 9 (partially supersedes rows 3/7). The provenance ladder stays primary โ€” WebSearch results are leads until arXiv/Crossref-verified. Eight surfaced leads (local Gaussian correlation regime test, KLD sliding-window instability, MS-vine copulas, positive-definite dependence measures, FNETS, RKHS-OT dependence, irreversibility analysis, robust dCor) are parked in FAMILY-TAXONOMY for future firings.

Coverage counters after this firing

BoundaryMinimumVerifiedScreened-passFedStarvation
STABLE100231010 (B-01)77
FRAGILE↔CONDITIONAL6010059
CONDITIONAL-scoping4010039
FAIL4060034
EXPIRY4060034
cross-cutting2010019
Total300381010262

โ–ถ Next iteration

Machine-readable record: ledger row 9 + catalog ยงI-01 + taxonomy appended-leads. Pages are append-only.