Viska research pipeline · internal

Parcel health — 2026-08-10

30 canonical ledgers (__ comparison runs excluded)

Engineering only — not a client artifact

69.7
/ 100
D

vs 2026-08-07: -3.3 (73.0 → 69.7)

Scores process health — whether the pipeline did its job. It is not a measure of how well the corpus was read; a day can score 100 and contain shallow readings.

What this number is not. Read quality is an explicit rubric, never an invented number. Every check below is a fact about structure — none is a judgment about content. Silent under-work is this pipeline's characteristic failure and is invisible to a shape check, so depth is a signal to go look, never a verdict.

Dimensions

DimensionScoreWeightWhat it measured
integrity77 (-13)40%23/30 ledgers structurally sound
traceability40 (-3)25%findings 802/802 · flags 0/118 · eureka 0/39
vocabulary60 (+1)15%71/118 flag horizons on the closed set · 4 ledger(s) mix BOTH
depth100 (+0)10%28/28 within half of their page-band median
provenance10010%60/60 in-force provenance fields present (cutover-aware)

Anomalies 41 ledger flag(s)

integrity — 7 ledger(s)

Cites that resolve, ledgers that are not empty, and no citation of disclosure or certification boilerplate.

  • sha1:0a49be1c21d (cites boilerplate ['p11-disclosures'])
  • sha1:132c8691abd (cites boilerplate ['p19-company-disclosures'])
  • sha1:6c0059aaa2b (cites boilerplate ['p34-disclosures', 'p35-disclosures'])
  • sha1:716340ee8e0 (cites boilerplate ['p13-disclosures'])
  • sha1:7fe33bb5438 (cites boilerplate ['p12-disclosures'])
  • sha1:87b23fde807 (cites boilerplate ['p13-disclosures', 'p17-general-disclosures'])
  • sha1:bbf76dc23ef (cites boilerplate ['p15-pricing-disclosure', 'p17-coverage-disclosures'])

vocabulary — 12 ledger(s)

Whether closed sets were honoured — chiefly the two inverted horizon vocabularies that live in one ledger object.

  • sha1:2de2c82c66d (4/4 off-set)
  • sha1:37913e9d8e5 (7/7 off-set)
  • sha1:3c169a8b601 (5/6 off-set, MIXES both vocabularies)
  • sha1:50b914950de (1/1 off-set)
  • sha1:6bb96e63896 (4/4 off-set)
  • sha1:6c0059aaa2b (4/4 off-set)
  • sha1:83c97596e83 (3/5 off-set, MIXES both vocabularies)
  • sha1:a692a00072a (7/8 off-set, MIXES both vocabularies)
  • sha1:c2f4874d15c (4/4 off-set)
  • sha1:c6d35a8d638 (1/4 off-set, MIXES both vocabularies)
  • …2 more

traceability — 22 ledger(s)

Whether the fund can get from an assertion back to the page it came from.

  • sha1:0a49be1c21d (flags 0/4)
  • sha1:2de2c82c66d (flags 0/4)
  • sha1:37913e9d8e5 (flags 0/7)
  • sha1:3c169a8b601 (flags 0/6)
  • sha1:4253096b901 (flags 0/3)
  • sha1:4ffe6071033 (flags 0/3)
  • sha1:50b914950de (flags 0/1)
  • sha1:6bb96e63896 (flags 0/4)
  • sha1:6c0059aaa2b (flags 0/4)
  • sha1:716340ee8e0 (flags 0/37)
  • …12 more

Consistency by reader

An anomaly concentrated in one reader is a reader problem. One spread evenly across every reader is a contract problem — that distinction is the whole reason this table exists.

ReaderLedgersFlag horizons off-setFlags uncited
viska-d-res/gpt-5.6-luna2047/6363/63
viska-d-strat/gpt-5.6-luna100/5555/55

Scale collapse

A three-level scale with no written anchor collapses to its default, and then carries no information. Watched, never scored.

ScaleDistributionOn the default
red flag severity{'high': 37, 'medium': 78, 'low': 3}66% medium
eureka confidence{'medium': 34, 'high': 4, 'low': 1}87% medium