Independent builds, ordered by first commit
Portfolio / Evidence inventory / Case study
Build by Build.
An inventory of the tests
behind the portfolio.
Eleven independent projects, brought into one view. Trace the checks each build carried, the results they recorded and the limits that remain.
Test-surface definitions, with evidence and limits
Additional definition introduced with Claim Integrity
The problem
Make the evidence
easier to inspect.
Separate project pages explain individual builds. A shared inventory makes their controls comparable without turning the portfolio into a ranking.
Find the check
Rebuilds, source governance, numerical checks, presentation review and operation sit in a common vocabulary.
Read its actual state
A report can exist while a check fails, remains deferred or covers only part of the requirement. Its presence alone cannot establish a pass.
Keep the history
The original ten builds retain every recorded cell. Older projects are not retrofitted to fill the table, and an absent historical artifact creates no remediation debt.
Reading the comparison
Five states.
A clearer distinction.
Shape and color carry the state together. Select a mark in the matrix for the path, recorded result and available evidence.
Artifact recorded
A qualifying artifact was recorded. Its scope and actual outcome still matter.
Limited evidence
The record retains a failure, deferral, open item or other limit. Completion is not established.
Not carried
The original inventory did not record a qualifying artifact for that surface.
Not assessed
The retained evidence cannot establish presence or absence against the additional definition.
Not applicable
The requirement does not apply at this build scope or review phase. It is not counted as a pass or an absent artifact.
This is a categorical inventory. Artifact counts are not quality percentages, maturity scores or business-performance measures.
The Claim Integrity extension
Expected work.
Completed checks.
The original matrix covered ten builds and fourteen surfaces. Claim Integrity adds a build and the expected-event definition, while retaining the earlier evidence.
What should have arrived?
Expected event accounting starts from independently retained triggering facts. A claim never constructed or a promised response never received remains expected work, with its cutoff and timing basis.
The obligation itself is explicit, extending population reconciliation beyond the records that arrived.
Read the counting ruleWhat passed, at which scope?
Claim Integrity records 98/98 applicable golden-fixture variants, with 649/649 assertions compared; 81/81 applicable public diagnostic injections; and 22,173/22,173 numerical cells matching.
The chart review and restricted four-seat simulated panel are complete. Scheduled operation is N/A for the frozen snapshot.
Read the completed checksThe ten new cells for earlier builds are not assessed. The original fourteen-column inventory cannot settle the additional capability, and no first-ever capability claim is made. This is an addition to this portfolio inventory, not a change to the shared standards repository.
The implementation
One retained baseline.
A traceable extension.
Python combines local JSON records into a static HTML, CSS and JavaScript page. The same inventory feeds the homepage count and miniature.
- 01 / Retain
Original freeze
The ten-build inventory and its design-standard history keep their October 7 freeze.
- 02 / Extend
Pinned evidence
Claim Integrity's row adds source identities, recorded outcomes and their scope.
- 03 / Separate
Artifact and result
A dedicated evidence register keeps a path distinct from the conclusion it can support.
- 04 / Render
Page and miniature
A deterministic builder renders the comparison; the website consumes the same extended inventory.
A table that can be read
Real row and column headers preserve context. Build and group filters narrow the comparison; sticky headers and build names stay visible while scrolling.
Evidence within reach
Keyboard-operable cells open the evidence panel with the path, recorded outcome, scope and available artifact or verification summary.
A small static stack
Local JSON and a static page fit a categorical comparison. No chart library or database is needed for this inventory.
Architecture and provenance
- Original baseline
data/builds.jsonanddata/standard.jsonretain their October 7 freeze at66b9de6. No reread or retrofit of those ten repositories is implied.- Claim Integrity extension
- Repository evidence is pinned to
9e21fc7. Recorded outcomes come from the frozen public verification snapshot. Served snapshot bytes and later repository artifact bytes retain separate identities where they differ. - Deterministic inventory
src/extend_inventory.pycombines the baseline and authored extension. Check mode verifies the unchanged baseline, extension hashes, row and column shape, and explicit not-assessed states.- Publication builder
src/build_expanded_page.pyrenders the matrix, definitions, scope notes and historical design-standard lane. Website counts and its miniature use the same expanded inventory.
The extension is additive. It does not claim all eleven repositories were audited on one date.
Inspect the definitions and provenanceTesting and scope
State what was checked.
Keep the limits visible.
Focused integration checks support this interface update. They do not replace the underlying builds' assurance records or certify the effectiveness of their controls.
What was checked
- Preservation of all 140 original matrix cells.
- Ten explicit not-assessed cells in the additional legacy column.
- Deterministic output, recorded outcomes and internal destinations.
- Responsive rendering, navigation, filters and keyboard interaction.
Check receipts accompany the candidate. The original inventory's validation and panel records remain historical evidence.
What the checks establish
The retained inventory can be reproduced, its evidence can be followed and its current interface can be reviewed.
An earlier panel did not review this redesigned layout. No new blind panel or assurance cycle is claimed for this UI update.
A recorded artifact can carry failed, blocked, deferred, open or historical results. Those states remain visible.
Build forward
What a next iteration
could establish.
These are future investigations, not work represented as completed in this inventory.
Which older artifacts meet the additional definition?
Inspect the earlier candidate artifacts under an explicit scope. Presence or absence needs evidence; it cannot be inferred from the old fourteen columns.
Detection evaluation.
Outside the current test matrix for all eleven builds. Future work could establish independent custody, a complete held-out fault roster, clean opportunities and timing-aware scoring.
The historical unblinded attempt met must-catch in 20/26 represented classes, with six missed targets and ten unrepresented classes. These exploratory results do not establish final-control effectiveness.
Read the future candidate and retained context ↗Explore the work
Choose a cell.
Read what it supports.
Independent portfolio projects, not client work. Synthetic data is labeled by the build that uses it. No score or business-performance claim is inferred from this inventory.