Atlas r0056 · measured, not scored

Evidence quality, with the seams visible.

Coverage, completeness, transport health and resilience—published as inspectable facts without turning heterogeneous evidence into a single synthetic score.

Claims with sources

3,258 / 3,258

At least one explicit evidence reference.

Claims with reachable evidence

2,564 / 3,258

At least one exact URL join to a stored reachable route.

Counterevidence

2,603

Claims retaining a counterevidence or uncertainty field.

Update criteria

2,776

Claims stating what could change the reading.

Reachable routes

1,926 / 2,484

Stored transport result only—not a truth score.

Blocked routes with fallback

24 / 558

Registered owner-preserving paths; partial support stays partial.

Accepted Living Signals

0

No candidate is promoted before its gates pass.

There is no magic quality number.

Corpus completeness and stored transport state only; this is not a truth, importance, freshness or model-performance score.

Inspect quality JSON →Audit signal readiness →Browse source records →

Coverage by domain

Domains overlap by design. Counts below are memberships, not additive corpus totals.

Source resilience records

W71-SR-001
PDF cover names NIST AI 100-1, the identical DOI, title, agency and publication date; 48 pages inspected.
W71-SR-002
XML carries DOI 10.1136/bmj-2024-081123, BMJ article ID, title, author list and complete article body; PMCID inspected against the PMC record.
W71-SR-003
Publisher PDF title, authors, study design, N=30, tables and conclusions match the DOI article; all 12 pages inspected and claim-relevant table/text located.
W71-SR-004
Repository cover cites the identical DOI and says accepted version; 57-page manuscript contains the reported studies, incentive-compatible choices, vulnerability and conspicuous-label moderators.
W71-SR-005
PDF first page contains exact title, authors and DOI 10.1086/592130; 14 pages and claim-relevant multinational/local-firm and Hindi/English experiments inspected.
W71-SR-006
Repository cover maps authors, title, journal and DOI; 24-page text contains exact claim values 21.5/22.1 sessions, $721/$773 sales, +3%, +7% and nonsignificance.
W71-SR-007
PDF metadata and first page give exact title, authors and DOI; 19 pages include the randomized field experiment, redemption and one-/four-week expenditure comparisons.
W71-SR-008
Repository cover cites exact title, authors, journal, pages and DOI; 50-page manuscript contains affective/cognitive review content and referral/(re)purchase analyses.
W71-SR-009
PDF metadata/first page match exact title, authors and DOI; 21 pages contain disclosure, human-oriented messaging and employee-workload mechanism.
W71-SR-010
PDF metadata and first page identify exact title, authors and DOI; 17 pages contain the field experiment, sales activity, units sold and tailored human-centered implementation results.
W71-SR-011
Same authors, title, eBay experiment and core findings; 35-page working paper explicitly reports no measurable brand-keyword benefit, 0.66% total paid-search sales estimate, frequent-user spend concentration and negative average returns.
W71-SR-012
Full-text XML has the identical DOI, title, authors, randomized experiment N=704 and all reported warning-label, purchase-intention and nutrient-claim results.
W72-SR-001
The 192-page JMP PDF has the identical report title and contains the cited 48%→58%, 3.4 billion, 354 million, 429 million, fragile-context gap and 18-fold acceleration results.
W72-SR-002
The 30-page UNIDO PDF carries the Q4 2025 report title and the claim-relevant 0.5% global growth, regional ordering, 0.2% lower-tech growth and above-1% higher-tech result.
W72-SR-003
The 86-page UNEP repository PDF is the full report linked by the canonical resource page and contains the more-than-tripling demand, 7.2 GtCO2e business-as-usual and 2.6 GtCO2e pathway values.
W72-SR-004
The 236-page WHO IRIS PDF has the same ISBN/title and contains the 15.8% global, four age-group and 24.3% versus 11.0% income-group estimates with uncertainty intervals.
W72-SR-005
The current FDA page differs only by removal of one 'for' token in the slug; title, release date, authorization, 30-minute result and all six positive/negative agreement values match.
W72-SR-006
The Budget chapter directly supports completion of the Consumer-Driven Banking Act, Bank of Canada oversight, national-security safeguards and write/payment-initiation work by mid-2027; it does not restate the canonical framework page's accreditation, consent and liability detail.
W72-SR-007
The 37-page PDF is the identical ICO report under the current /media2/ tree and contains workplace concentration, discrimination-risk, directly-produced neurodata and excluded eye/gait/heart-rate examples.
W72-SR-008
The 67-page authors' working paper contains the same deployment, productivity outcome and skill/experience heterogeneity, but its main analysis reports 5,179 agents and about 14%; the QJE version reports 5,172 and 15% after revision.
W72-SR-009
The 174-file-page OECD PDF carries the identical DOI, title, ISBN and CC BY 4.0 notice and details the five-module scenario/stress-test method; its role in the D-class Metatron synthesis remains contributory rather than independently causal.
W72-SR-010
The XML contains DOI 10.1073/pnas.2422633122, matching title/authors and full methods/results, including the 48%, 127%, −17% and guarded-tutor unassisted-result distinctions.
W72-SR-011
The full XML matches DOI, title and author list and explicitly reports immediate intent versus control, no superiority to public-health materials, no sustained 45-day intent and no uptake difference.
W72-SR-012
The XML matches DOI 10.1093/pnasnexus/pgaf170, title and authors and contains the belief/credibility reductions and nonsignificant standalone AI-label sharing-intention result.

Completeness, transport, resilience — separately

Self-reported health is not evidence

Corrections are counted in public

New to this publication?