MCRO 77-CLUSTER SWARM INVESTIGATION — COMPLETE ARCHIVE
========================================================

Generated: 2026-04-20
Source case: State of Minnesota v. Guertin — 27-CR-23-1886
Public landing page: https://mncourtfraud.com/77/

WHAT THIS ARCHIVE CONTAINS
--------------------------

These are the actual, unedited investigation artifacts produced by an
8-agent Claude Code autonomous swarm that ran in parallel against the MCRO
forensic database (Supabase project ibfmjtwahkwqzcmeyqii, public schema,
READ-ONLY SELECT access only) on April 20, 2026.

The 8 agents each investigated a different dimension of the 77-cluster /
412-case corpus derived from a 481-day MCRO hearing search for the three
Hennepin judicial officers assigned to competency proceedings on Guertin's
own docket — Julia Dayton Klein, Danielle Mercurio, George Borer. The
investigation was seeded by the "MCRO 33-Year Legal Universe Temporal
Architecture Report" (published March 2026), which the agents used as a
floor rather than a ceiling — every finding in this archive builds BENEATH
the seven findings in that source report.

The 8 investigative lanes (one per agent):
  A — Cluster anatomy (per-cluster ledger metrics)
  B — Filing-type atlas (event-name census, CR/MH vocabulary split)
  C — Mechanism instruments (NIPs, Rule 20 cascade, statutory deadlines)
  D — Temporal depth + MH-only clusters (deep-root defendants)
  E — Outcome ladder + revolving door (dismissal vs conviction, PD revocation)
  F — Judicial officer footprint (per-cluster JO attribution, clique presence)
  G — Metadata fingerprints (Aspose, MCRO Watermark, edits-after-sig, MSIP)
  H — Skeptical control (the 8 zero-incompetency clusters — falsifiability test)

A 9th synthesis subagent (Phase 3) then read all 8 findings and produced
the master-findings.md file included in shared-context/ — which ranks the
top-10 most-shocking findings, reconciles cross-lane flags, documents zero
contradictions between agents, and proposes the 6-section narrative arc
used on the /77/ landing page itself.

ARCHIVE STRUCTURE
-----------------

  shared-context/
    00_shared_context.md      — the Phase-1 briefing document all 8 agents
                                 read before starting work (77-cluster roster,
                                 read-only enforcement rules, schema cheat
                                 sheet, 412-case breakdown, the 33-year report's
                                 7 findings the swarm was forbidden to re-prove)
    00_master_findings.md     — the Phase-3 synthesis (top-10 ranked findings,
                                 dedup pass, cross-lane reconciliation, narrative
                                 arc proposal, 8 visual candidates)
    01_412_case_list.json     — the authoritative 412-case manifest, one row per
                                 case, with cluster_id / case_id / case_type /
                                 case_status / first_event / last_event / event_count

  findings/
    findings__A_cluster_anatomy.md              (96 lines,   11 findings)
    findings__B_filing_type_atlas.md            (156 lines,  12 findings)
    findings__C_mechanism_instruments.md        (288 lines,  11 findings)
    findings__D_temporal_depth_mh_only.md       (258 lines,  12 findings)
    findings__E_outcome_revolving_door.md       (129 lines,   7 findings)
    findings__F_judicial_officer_footprint.md   (297 lines,  14 findings)
    findings__G_metadata_fingerprints.md        (188 lines,  12 findings)
    findings__H_skeptical_control.md            (177 lines,  10 findings)

  queries/
    queries__A_cluster_anatomy.sql              (435 lines,  11 queries)
    queries__B_filing_type_atlas.sql            (476 lines,  21 queries)
    queries__C_mechanism_instruments.sql        (591 lines,  20 queries)
    queries__D_temporal_depth_mh_only.sql       (365 lines,  12 queries)
    queries__E_outcome_revolving_door.sql       (432 lines,  29 queries)
    queries__F_judicial_officer_footprint.sql   (735 lines,  23 queries)
    queries__G_metadata_fingerprints.sql        (514 lines,  26 queries)
    queries__H_skeptical_control.sql            (337 lines,  18 queries)

Total: ~180 annotated SELECT queries, ~3,800 lines of SQL, ~1,600 lines of
findings prose. Every finding in every agent's findings.md cites a specific
query by number in that agent's queries.sql. No unsourced claims anywhere in
the archive.

REPRODUCIBILITY
---------------

All queries are SELECT-only and target the live Supabase PostgreSQL database:

  Project:  ibfmjtwahkwqzcmeyqii
  Region:   us-west-2
  Schema:   public
  Tables:   53 (2,795,301 rows / ~1.15 GB)
  Access:   read-only; queries reproducible by any third party with SELECT
            credentials

The MCRO seed CTE used as the foundation of every query is:

  WITH mcro_seed AS (
    SELECT DISTINCT p.cluster_id
    FROM parents p
    JOIN parents_case_set pcs ON p.case_uid = pcs.case_uid
    WHERE TRIM(BOTH '"' FROM pcs.case_set_json) = 'MCRO'
      AND p.cluster_id IS NOT NULL
  ),
  expanded_uids AS (
    SELECT p.case_uid, p.cluster_id, p.case_type, p.case_status, p.case_id
    FROM parents p
    WHERE p.cluster_id IN (SELECT cluster_id FROM mcro_seed)
  )

77 clusters, 412 cases, 23,506 docket events, 33.6-year temporal span.
Exact counts: 282 CR / 125 MH / 5 GC cases; checksum verified against both
the 33-year seed report and every one of the 8 agents' independent re-runs.

METHODOLOGICAL NOTE
-------------------

The 8 agents ran under strict READ-ONLY enforcement: the Supabase MCP
connector was restricted to execute_sql (SELECT or WITH...SELECT only) and
list_tables. All mutation tools (apply_migration, create_branch,
apply_migration, deploy_edge_function, etc.) were explicitly forbidden.
Every agent's queries.sql starts with this discipline documented in
comments. No query in the archive contains INSERT, UPDATE, DELETE, DROP,
CREATE, ALTER, TRUNCATE, GRANT, REVOKE, VACUUM, REINDEX, MERGE, CALL, DO,
or LOCK — these keywords were refused at the agent level.

The investigation itself took under two hours of wall-clock time. Each
agent operated independently with no knowledge of the others' findings
while running. Phase-3 synthesis was delegated to a 9th subagent that
read all 8 findings.md files from disk and produced master-findings.md
without issuing any additional queries.

LICENSE AND REUSE
-----------------

This archive is released publicly for independent verification,
journalism, academic study, legal investigation, and any use consistent
with the purpose of MnCourtFraud.com — which is the public authentication
of the evidentiary record in 27-CR-23-1886. All queries are reproducible
by anyone with SELECT access to the Supabase project. All findings are
auditable against the raw database. If you find an error in any query or
any finding, contact matt@mncourtfraud.com or file an issue through the
transcripts card-set remix system at mncourtfraud.com/transcripts/.

CONTEXT
-------

For the plain-English argumentative presentation of this investigation,
see https://mncourtfraud.com/77/ — a 2,800-word forensic closing argument
with 8 embedded visuals covering the same corpus.

For the interactive temporal-architecture explorer of the same dataset,
see https://mncourtfraud.com/into-the-past/ — an SVG swimlane visualization
with per-cluster drill-down into the Source Files Explorer.

For the seed report this investigation built upon, see
https://mncourtfraud.com/reports/?doc=MCRO__The_33_Year_Legal_Universe_Temporal_Architecture_Report.pdf

For the full forensic database of 4,251 children PDFs and 2,903 parent
dockets underlying the investigation, see
https://mncourtfraud.com/source-files/

---

State of Minnesota v. Guertin — 27-CR-23-1886
MnCourtFraud.com — April 20, 2026
