Larchmere Library and Community Hub (Demo Data)Evidence ledger, chain-diff, rolling baseline, trust score, slip-to-evidence

The documents in the evidence folder name 14 of the 286 activities that slipped between the baseline and the current update, and only 2 of the 50 slipped activities on the driving path to completion.

The largest slip, 86 working days, is shared by nine activities including Substantial Performance of the Work (A1040), which moved from 10 April 2026 to 13 August 2026; no document names any of them, so the record does not yet show what moved completion. The 14 named slips can be argued from the documents; the other 272 have only documents dated in the same window, which is context, not proof. Three records are needed to close the gap; section 3 names who holds each, and all are needed now.

The current update was filed at data date 5 August 2025 but records actual progress up to 14 October 2025, so it is read at 15 October 2025, as section 4 sets out.

The documents name 14 slipped activities, 2 of them on the driving path; each is listed below with the documents that quote its code, the driving-path slips first.

Documents That Carry the Record

Of the 21 documents, 19 quote the code of at least one slipped activity. The other 2 quote none: E-001 Notice to proceed and E-023 Schedule update transmittal. The ledger keeps them as context.

Three records stand between these documents and a supported claim; each is stated as a requirement with the party that holds it, and all are needed now.

The slips come from the schedules as filed and the matches from the codes the documents quote; the schedule chain scores 80 of 100 for trust, and every point it loses comes from network checks, not from the pattern of edits.

Each slip is the change in an activity’s finish between the baseline (data date 5 May 2025) and the current update (data date 15 October 2025), counted in working days on the activity’s own calendar. A document supports a slip only when its text quotes the activity’s code. A document merely dated inside the slip’s window is context, and the schedules themselves are never counted as evidence of their own slips.

What the Trust Score Rests On

None of the edit-pattern detectors raised a flag across the update chain. The network checks of the two schedules raised 2 warnings for parts of the network not linked to the rest. The score is 100 less 10 points for each warning, which gives 80; exhibit A-08 lists every finding. The score measures how the schedules were kept, not who caused a delay.

Changes Between the Schedules

263 changes between the schedules are logged for review: 96 logic links added, 55 activities added, 46 logic links removed, 34 predecessors swapped, 19 activities removed and 13 durations changed. 53 of them are at warning level. Each is an edit to check against the record, not a finding. The 55 activities added after the baseline are baselined at the dates each was introduced with, so their later movement can be measured (exhibit A-05).

What the Ledger Cannot See

The ledger matches activity codes; it does not read what a document says. A document that describes the work without quoting its code is not counted, and a document that quotes a code is not, by that alone, proof of what caused the slip. All 21 documents are dated from their filename.

How the Schedules Were Read

Basis of Measurement

The exhibits behind these findings follow in full, each under its drawing code:

• A-01Artifact Inventory
• A-02Trust Assessment
• A-03Schedule Chain Summary
• A-04Schedule Change Log
• A-05Rolling-Baseline Insertions
• A-06Evidence Ledger
• A-06bEntity Summary
• A-07Slip ↔ Evidence Pairings
• A-08Network Integrity Flags
• A-09Trust Flags Detail
• A-10Methodology & Provenance

Rolling baseline written to Larchmere_Library_and_Community_Hub_Demo_Data_Rolling_Baseline.xer. Preserved 343 original baseline activities. Inserted 55 new activities introduced mid-chain, each captured with as-introduced planned dates.

Corpus totals: 1 distinct people · 9 organizations · 1 RFI numbers · 0 change-order numbers. Method: tiered NER, regex plus curated construction dictionaries and an optional model. Correlation only.

Evidence-Supported Rate counts only slips whose paired evidence-document text actually references the slip’s activity code (match reason both or activity_code_in_text). Date-Window-Only slips have evidence documents that fall in the slip’s date window but do not name the activity (date_within_window); contextual co-occurrence, not evidentiary support. Schedule-File Context slips matched only the schedule XER snapshots under analysis. A schedule cannot be evidence of why it itself slipped, so these are excluded from every evidence rate. Any-Artifact Context is the all-inclusive ceiling (schedules included) and is not an evidence rate. Contextual-only candidates (date co-occurrence and the schedule snapshots under analysis) are rolled up to one counted line per slip: the same artifacts co-occur with every slip whose window covers their date, and the full slip-window and artifact-date lists on this page reconstruct each candidate exactly. Their ids ship in the register’s CSV export.

Findings are CORRELATION, not CAUSATION. compute_cpm_salvaging detects network-structural conditions (AACE RP 29R-03 §2.1 baseline schedule validation and §2.3 schedule-update validation; DCMA-14 logic check). Each flag warrants analyst verification against contract documents. It does not determine intent or manipulation.