The 15-repository comparison: tables and method
Study overview and reading order
These are the frozen values behind the study’s main result: 13 of 15 exact candidates passed their checks, 1 had recorded acceptance, 0 were traced as a whole to their source and 0 of 25 published surfaces were independently observed running. The three distributions below give each question’s full set of states. The case table at the end is the one to read across, because it shows which repository holds which state; the separate distributions cannot be lined up to show that.
Historical evidence, not live status. All figures describe the frozen schema-v1.1 Project 03 comparison. An accepted W0–W8 package means the evidence was reviewed for the study—not that its software was accepted, deployed, healthy or operating today.
Case 01–15 are stable editorial pseudonyms, not guaranteed anonymity. Already-published examples may make some cases recognizable. This is not the complete underlying dataset; source references and the case-to-repository mapping remain private.
Candidate verification
15 repositories. What check results were observed for an exact candidate?
| Recorded state | Count |
|---|---|
SUCCESS_OBSERVED | 13 |
MIXED | 1 |
NOT_ESTABLISHED | 1 |
The mixed case includes two successful checks and one failed check. It is not a pass. Candidate verification is separate from engineering acceptance.
Engineering acceptance
15 repositories. Was the required candidate-specific acceptance authority established?
| Recorded state | Count |
|---|---|
ESTABLISHED | 1 |
UNKNOWN | 1 |
NOT_ESTABLISHED | 13 |
These are separate distributions, not a cross-tab. They do not identify which repositories overlap. The separately recorded 11/15 pairing is explained in Passing checks and software acceptance are different findings; it cannot be deduced by aligning the chart totals. Package acceptance is not engineering acceptance.
Independent runtime observation
25 projection observations—not 15 repositories. These are not necessarily twenty-five distinct deployed services.
| Recorded state | Count |
|---|---|
ESTABLISHED | 0 |
NOT_ESTABLISHED | 18 |
NOT_OBSERVED | 4 |
NOT_COLLECTED | 1 |
UNKNOWN | 2 |
Zero positive independent observations do not prove that the systems were offline. These historical records do not establish current uptime. Published code is not observed runtime explains why publication, runtime claims and independent observation answer different questions.
The fifteen-case comparison
Each column asks a different question. The row states are not grades. Whole-candidate lineage can remain unestablished even where provenance for a smaller declared subset is positive.
Wide table: swipe or scroll sideways to read every field. Keyboard users can Tab to the table region and use the left/right arrow keys. On a narrow screen, the qualification column repeats its case label.
| Case | Candidate verification | Engineering acceptance | Whole exact candidate lineage | W0–W8 package review | Reading qualification |
|---|---|---|---|---|---|
| Case 01 | SUCCESS_OBSERVED | ESTABLISHED | NOT_ESTABLISHED | ACCEPTED | Explicit candidate-specific engineering acceptance; public projection relation is not whole-candidate lineage. |
| Case 02 | SUCCESS_OBSERVED | NOT_ESTABLISHED | NOT_ESTABLISHED | ACCEPTED | Successful exact-head checks do not establish acceptance; one public projection observation has no recorded stable ID. |
| Case 03 | MIXED | NOT_ESTABLISHED | NOT_ESTABLISHED | ACCEPTED_WITH_QUALIFICATIONS | MIXED: two successful checks and one failed check. Positive provenance is limited to a declared 27-file subset. |
| Case 04 | SUCCESS_OBSERVED | UNKNOWN | NOT_ESTABLISHED | ACCEPTED_WITH_QUALIFICATION | Engineering acceptance UNKNOWN remains unresolved; absent runtime observation is not proof of service failure. |
| Case 05 | SUCCESS_OBSERVED | NOT_ESTABLISHED | NOT_ESTABLISHED | ACCEPTED | Seven separately keyed projections; one different direct-public technical surface is not an additional projection. Historical visibility drift remains. |
| Case 06 | SUCCESS_OBSERVED | NOT_ESTABLISHED | NOT_ESTABLISHED | ACCEPTED_WITH_QUALIFICATIONS | Four successful checks; public surface lacks a source-recorded projection ID; no whole-candidate lineage established. |
| Case 07 | NOT_ESTABLISHED | NOT_ESTABLISHED | NOT_ESTABLISHED | ACCEPTED_WITH_QUALIFICATIONS | Qualifying exact-head verification NOT_ESTABLISHED; formal activity collection gap must not be called a failed check. |
| Case 08 | SUCCESS_OBSERVED | NOT_ESTABLISHED | NOT_ESTABLISHED | ACCEPTED_WITH_QUALIFICATIONS | Three keyed projection observations; independent runtime observation not positively established. |
| Case 09 | SUCCESS_OBSERVED | NOT_ESTABLISHED | NOT_ESTABLISHED | ACCEPTED_WITH_QUALIFICATION | One public surface without a stable recorded projection ID; exact-candidate relation not established. |
| Case 10 | SUCCESS_OBSERVED | NOT_ESTABLISHED | NOT_ESTABLISHED | ACCEPTED_WITH_QUALIFICATIONS | Final W0–W8 reviewer lock retains qualifications; publication review is not engineering acceptance. |
| Case 11 | SUCCESS_OBSERVED | NOT_ESTABLISHED | NOT_ESTABLISHED | ACCEPTED | Successful checks are bounded to offline/mock/backend-disabled conditions, not actual cloud runtime. |
| Case 12 | SUCCESS_OBSERVED | NOT_ESTABLISHED | NOT_ESTABLISHED | ACCEPTED_WITH_QUALIFICATION | Two keyed projected candidates; historical visibility conflict persists and runtime observation remains UNKNOWN. |
| Case 13 | SUCCESS_OBSERVED | NOT_ESTABLISHED | NOT_ESTABLISHED | ACCEPTED_WITH_QUALIFICATIONS | Three distinct projection candidates; a live claim is not a positive independent runtime observation. |
| Case 14 | SUCCESS_OBSERVED | NOT_ESTABLISHED | NOT_ESTABLISHED | ACCEPTED | Generated, history-light visible projection; recorded zero public activity does not establish upstream inactivity. |
| Case 15 | SUCCESS_OBSERVED | NOT_ESTABLISHED | NOT_ESTABLISHED | ACCEPTED_WITH_QUALIFICATIONS | Private generated export; bounded zero activity not upstream inactivity; repository-level W4 without a projection instance. |
How to read the limits
NOT_ESTABLISHEDmeans the required evidence was not established in the frozen collection—not that the system failed or never existed.UNKNOWN,NOT_OBSERVEDandNOT_COLLECTEDretain their distinct recorded meanings.- Whole exact candidate lineage is a complete candidate-binding question, not a denial of every repository relationship or scoped provenance result.
- W0–W8 package review concerns the evidence package, potentially with qualifications—not engineering acceptance of its software. Every case has one, so the column shows the study’s coverage, not a difference between repositories.
Consumer Lab shows this concretely: a mixed verification result and established provenance for a declared 27-file subset coexist with whole-candidate lineage NOT_ESTABLISHED. The Consumer Lab provenance explanation explains the scope distinction.
What is deliberately withheld
Repository names and mappings, private source coordinates, receipts, selectors, raw activity windows and screenshots are omitted by editorial choice. Withheld information is not the same as missing or unestablished evidence. Private workbook links are not supplied.
Activity windows were not fully aligned and historical conflicts remain unresolved. This extract does not total them, create productivity or readiness scores, or extend the original three-case conclusions to all fifteen.
Download the selected frozen extract
These are the unchanged values displayed in this comparison, offered in machine-readable form—not full source workbooks or newly collected records.
- Download the selected 15-case CSV — 15 repository evidence cases, six columns, exact state tokens, original qualifications, and stable
Case 01–Case 15editorial pseudonyms. Published examples can make a case recognizable; pseudonyms do not guarantee anonymity. - Download the three frozen distributions as JSON — n=15 repositories for verification, a separate n=15 for engineering acceptance, and n=25 projection observations for independent runtime observation. The two repository distributions are not a cross-tab.
The downloads are a selected frozen schema-v1.1 extract, not the complete underlying dataset. Raw activity windows are omitted because their intervals/cutoffs differed and some were incomplete. Historical source and visibility conflicts remain unresolved. The downloads provide no private source mappings, original identities, provenance receipts, service endpoints or current deployment status.
Field guide: what is recorded, what is withheld
The CSV’s candidate_verification records candidate-bound checks, not engineering acceptance. The distinct engineering_acceptance field records whether the required candidate-specific authority was established. whole_exact_candidate_lineage asks about a complete candidate binding, even when a narrower declared subset has positive provenance. accepted_w0_w8_package is an evidence-package review state, not software acceptance. Each reading_qualification is a bounded explanation accompanying the original case.
The recorded categories MIXED, UNKNOWN, NOT_ESTABLISHED, NOT_OBSERVED, and NOT_COLLECTED retain their different meanings. A historical runtime EXPLICITLY_NOT_CLAIMED state refers to what a surface claimed, not an independent finding of downtime. Deliberately withheld is a disclosure choice about a field excluded from the public extract; it is not another recorded evidence state and does not mean the field was never collected.
A projection observation lacking its original recorded identifier remains an unkeyed projection observation. A direct-public technical surface that is not a projection instance is a different category and is not added to the 25-observation count. Visible zero activity in a generated, history-light public projection also does not establish zero upstream work. None of these values is a productivity, quality, readiness, deployment-health or overall CI score.
Sources, method and reproducibility
The original three-case Cohort 1 explored contrasting repository profiles; the final estate describes 15 accepted repository evidence packages, without automatically generalizing every original three-case conclusion.
The retained study separated W0 source identity and scope; W1 bounded activity/history windows; W2 exact-candidate verification; W3 acceptance authority; and W4 publication, projection relationships, runtime claims and runtime observations. W5 normalized reviewed observations while retaining their source states and limits; W6/W7 formed bounded outcomes and analysis; and W8 selected publishable or qualified public claims. Collector material and reviewer dispositions were distinct, and the frozen Master and final comparison are downstream views of the accepted evidence, not independent repeated measurements.
The underlying source packages, workbook identities and original collection receipts remain private. Readers can reproduce the published presentation from these selected CSV/JSON values, including the displayed states and chart distributions; they cannot independently reconstruct or verify all underlying private repository, provider, screenshot or acceptance evidence from this public extract. This account reports reviewed historical records, not freshly repeated tests, provider reads or current uptime.
For a concrete research example, see Consumer Lab’s mixed checks and 27-file scoped provenance: that positive subset does not establish whole-candidate lineage. The separate Edge Evidence development and operational account reports historical engineering exercises, including a completed longer run with 75 source_gap_observed records and a retained limitation classification. Those exercises belong to the later engineering history, not the frozen Project 03 repository or projection denominators.
The Edge Evidence case study describes separate engineering demonstrations; it is not a current deployment attestation supplied by this frozen comparison.
Research basis: the previously reviewed frozen schema-v1.1 repository comparison and dashboard. Exact source references, field mappings and redaction decisions are retained privately. This presentation introduces no new measurements.