Nine findings from a data-detective investigation on the GAIN-provided lobbying + congressional press corpus, amplified by six external-OSINT spotlight-handoff passes. Most carry analytical novelty beyond what other outlets have reported. The rest tie data-side anomalies to existing reporting (NYT, ProPublica, OpenSecrets) so an editor can verify in minutes.
The State of LOC Nation Global Public Benefit Corporation reports $20 million per quarter through a Delaware shell whose only lobbyist is "Rev. Dr. Christina Loren Clement," styled across the filings as "President of Black USA," "2024 Presidential candidate (assumed the Presidency)," and "HH Empress Queen." Detector D9 confirms it is the only multi-signal sovereign-citizen-pattern outlier in the entire 418,098-filing corpus.
The 15 filings name virtually every executive-branch agency as lobbied — the President, Treasury, DOJ, DOD, FBI, Federal Reserve, IRS, U.S. Mint — and cite a real federal court case (DC 1:24-cv-00479 RC) where Clement appeared pro se against the Chief Justice, the UN Secretary General, the ICJ Registrar, and AG Garland.
The filings' own income line carries a self-disclaimer: the $20M is "pending HR 40 research reimbursement... not a federal payment unless and until processed." Clement is using the federal LDA registry as a reparations-claims mechanism. A second filing in the database — THEYFEARTRUTH FEDERAL GOVERNMENT OF AMERICA — confirms the pattern is rare but not unique. See C-005 below: the FEC has now formally reclassified Clement's $50M Senate-committee filing as "False and Fictitious."
ingest/scripts/ingest.py with lda_profile.py. 418,098 senate_filings rows.single_client_juggernauts (SQL hash f9e737b8f3554a90) flagged LOC COMMUNITY ASSOCIATION at $60M income. D9 shell_pattern_filings (hash bbbea171aa997a1e) confirmed it's the only multi-signal sovereign-pattern outlier in 418K filings.detect/scripts/query.py --sql against senate_filings WHERE client_name LIKE '%LOC NATION%'. Pulled all 15 filings.evidence-cards/scripts/evidence_card.py --from-csv loc_nation_filing_uuids.csv emitted 15 audit cards.lda.senate.gov/filings/public/filing/<uuid>/print/ for each cited record; archived stateoflocnation.com; CourtListener API for CLEMENT v. GARLAND docket.D9 (shell_pattern_filings, SQL hash bbbea171aa997a1e) scores each Senate filing 0-6 against six independent sovereign-citizen markers: sovereign_client, established_govt, esoteric_terms (Annuit Coeptis / King Solomon / Empress / sui generis), self_styled_title, posted_by_llc_slashes, global_pbc_naming.
At threshold ≥ 2, only LOC Nation surfaces. At ≥ 1, the next candidates are legitimate firms (King & Spalding, Patterson Real Bird) tripping one signal incidentally.
Delaware iCIS: only one entity exactly named under "Christina Loren Clement" — file 7127855 CHRISTINA LOREN CLEMENT LLC. No Delaware shell-LLC network exists under that name.
Florida Sunbiz — novel finding. LOC NATION JAX LLC (document L21000352541, EIN 87-2134353, ACTIVE) was registered in Jacksonville on August 5, 2021 by Anisha Hicks — pre-dating Clement's federal LDA scheme by three years. The "LOC NATION" brand thus pre-existed Clement's federal-level deployment of it; a separate operator holds a Florida LLC under it. Could be namesake / adjacent operator / franchise — needs further investigative work. Sunbiz.
False statements in federal filings (18 U.S.C. § 1001) and LDA enforcement (2 U.S.C. § 1605) are potentially implicated. Use of the federal LDA database as a reparations-claims mechanism is at minimum a misuse of the registration system. C-005 documents the FEC's own action on the same operator's parallel filing.
Anchor & Arrow Strategies (2024) and its successor Anchor & Arrow LLC (2026) were founded by John Noonan (Professional Staff, House Armed Services Committee; National Security Advisor to Sen. Tom Cotton) and Christopher Zumbar (Congressional Appropriations Liaison, Secretary of the Navy; Marine Corps comptroller). The client roster is exclusively AI- and autonomous-defense companies.
Christopher Zumbar — formerly Congressional Appropriations Liaison, Secretary of the Navy — now lobbies Saronic Technologies, an autonomous defense maritime systems vendor whose primary customer is the Navy.Anchor & Arrow LDA filings · LinkedIn · Leatherneck Magazine · The Hill
new_registrant_surge (SQL hash e51ec9c1d23d4479) flagged Anchor & Arrow LLC — new 2026-Q1 registrant with 11 clients on day one.senate_activity_lobbyists joined to senate_filings surfaced founders' covered_position text (Noonan, Zumbar, Jensen).external-data/scripts/usaspending.py-pattern API call → DoD contract awards for each A&A client, 2022-2026. Surfaced Apex's $22M Air Force SBIR STRATFI Award FA880925CB005.| Client | Sector | A&A fees | DoD contracts |
|---|---|---|---|
| Apex Technology, Inc. | Multi-mission satellites (LEO/GEO) | ~$100,000 | $25,699,922 |
| Overland AI | Autonomous ground vehicles | $280,000 | $8,860,744 |
| Base Operations Inc. | AI-driven security intelligence | ~$30,000 | $1,324,183 |
| Gallatin AI | AI / disaster-response navigation | $80,000 | $149,971 |
| Saronic Technologies | Autonomous defense maritime | $410,000 | — |
| Epirus, Inc. | Counter-drone microwave | $490,000 | — |
| Palantir Technologies | AI / data analytics | $260,000 | — |
Empty USAspending fields likely indicate prime contracts under variant recipient names (Palantir USG, Inc.) or classified streams.
D8 (new_registrant_surge, SQL hash e51ec9c1d23d4479) flagged Anchor & Arrow LLC as a new 2026-Q1 registrant with 11 clients on day one. The Apex Technology pairing is the cleanest single example: registered Q3 2025; $22M Air Force SBIR Phase II STRATFI Award ID FA880925CB005 for "Multi-Mission Satellites for LEO and GEO Dynamic Space Operations." Total DoD contract flow to A&A's clients: $36,034,820 against ~$2.4M lobbying spend (corrected 2026-07-09 from $35,937,820 — arithmetic error; line items re-verified against the USAspending API).
Across 2024-2026 Senate filings, 743 distinct registrants disclose foreign entities; only 61 (8%) also appear in the DOJ FARA registrant list. Detector D10 narrows the 682-name gap to 200 specific filings that combine FARA absence with non-US client country, foreign-government-policy lobbying topics, or state-owned-enterprise client naming. See C-006 below: OSINT amplification on Akin Gump's four worst engagements in this gap.
external-data/scripts/fara.py pulled FARA bulk (OpenSanctions DOJ mirror) into the same DuckDB index — 7,051 registrants, 17,662 foreign principals.foreign_filings (SQL hash f7c962b62d2393b1) listed Senate filings with non-empty foreign_entities[]. D10 fara_gap_narrowed (hash ae4ebdb2ecf19559) filtered the residual by non-US client country + foreign-government-policy keywords + SOE naming.senate_foreign_entities + senate_filings against fara_registrants by uppercased name equality.| Registrant | Client | Foreign nexus |
|---|---|---|
| Akin Gump Strauss Hauer & Feld | Ant Group Co. (CN) | Alibaba fintech arm; US regulatory friction |
| Rich Feuer Anderson | Ant Group Co. (CN) | Second LDA registrant for same Chinese principal |
| Bracewell LLP | Tele-Fonika Cable Americas Corp. | Polish-owned cable manufacturer |
| FGS Global (US) LLC | Wuxi Biologics USA LLC | Chinese pharma supply-chain affiliate (HK) |
Detector D11 — a multi-hop SQL join from lobbyist covered_position → Congressional committee → current quarterly filings on the same committee's jurisdiction — surfaces Brownstein Hyatt at the top. Russell Sullivan (former Staff Director, Senate Finance) and Mark Warren (former Chief Tax Counsel, Senate Finance + Treasury LegAffairs) co-lobby an overlapping TAX/FIN client roster. Sixteen D11 lobbyists account for $122M of 2024-2026 lobbying revenue on topics aligned with the committee they previously staffed.
There is no more effective, professional and respected tax lobbyist in Washington, D.C., than Russell Sullivan.Chambers USA, citing Sullivan's Brownstein Hyatt bio
| Role | Name | Activity |
|---|---|---|
| Former Staff Dir, Sen. Finance | Russell Sullivan (Brownstein) | $8.92M Apollo engagement |
| Former Chief Tax Counsel, Sen. Finance | Mark Warren (Brownstein) | Co-lobbies Apollo |
| Current Sen. Finance member (D-MA) | Sen. Elizabeth Warren | 9 press releases attacking Apollo over Steward, Walgreens PE, Genesis |
| Current Sen. Finance member (R-IA) | Sen. Chuck Grassley | 1 release — "Private Equity in Health Care Shown to Harm Patients" (Jan 2025) |
revolving_door_committee_match (SQL hash a7712c723043a717) maps each lobbyist's covered_position text to a committee, then joins to current quarterly filings whose issue codes match that committee's jurisdiction.external-data/scripts/congress_gov.py pulled the unitedstates/congress-legislators dataset — 536 current members × 230 committees × 3,879 assignments.committee_say_vs_pay (SQL hash 274c4ef00515cd73) — four-way join: lobbyist's former committee → current committee members (from congress_gov tables) → press releases attacking the lobbied client.| Client | 2024-2026 fees | Notes |
|---|---|---|
| Apollo Global Management | $8,920,000 | Largest engagement |
| Athene Holding | $4,540,000 | Apollo's annuities arm |
| Johnson & Johnson Services | $2,780,000 | Pharma |
| Highly Innovative Fuels | $1,920,000 | Energy |
| T-Mobile | $1,650,000 | Telecom/tax |
| Bloom Energy | $1,360,000 | Clean-energy tax credits |
| American Petroleum Institute | $1,260,000 | Oil & gas |
| Wynn Resorts | $1,170,000 | Gaming |
| NAREIT | $1,120,000 | REITs |
| TIAA | $1,090,000 | Retirement |
Per LegiStorm, Mark Warren left Brownstein for Capitol Tax Partners in May 2026. OS-006 follow-up (May 22) found Apollo still with Sullivan at Brownstein and zero Capitol Tax filings. Update (July 9, 2026): Apollo followed Warren — a new Capitol Tax Partners LD-1 for Apollo Global Management (9fb96e9d-4117-4257-9b6b-3bdfc66d16cf), effective June 29, 2026, lists Warren as sole lobbyist. Whether Apollo also keeps Brownstein shows in the Q2 2026 activity reports (due July 20).
Promoted from OS-001 spotlight-handoff
C-001 framed the LOC Nation LDA filings as fictitious. The external OSINT pass surfaced the institutional corroboration: on April 28, 2026 the FEC issued a warning letter to State of LOC Nation for Clement for Senate (FEC C00857128) over a $50.25M reported self-loan and reclassified the filing into the FEC's "False and Fictitious Filings" repository. The "loan" is the asserted fair-market value of a license to Clement's self-issued Loc Nation Dollar (LND), pegged at 1 LND = $750 USD.
The pattern is not unique to Clement. Rico Cortez Dukes, the second sovereign-citizen-pattern LDA operator (THEYFEARTRUTH FEDERAL GOVERNMENT OF AMERICA), has eleven pro se federal lawsuits in N.D. Texas (2017-2021) including Dukes v. Roman Empire, Dukes v. Federal Reserve Central Bank, and Dukes v. Trump. Both Clement and Dukes are formally registered FEC 2024 presidential candidates of self-styled sovereign-government parties.
spotlight-handoff generated brief OS-001 with named entities (Clement, Dukes), known URLs, and OSINT angles to pursue.spotlight:investigator agent (1 cycle, 22 sources consulted, 23 archived files via firecrawl + Wayback) pulled FEC committee record C00857128, Clement candidate pages, Dukes presidential committee, PACER N.D. Texas case list, OpenSecrets May 15 2026 coverage with Quinnipiac Prof. John Martin on-the-record quote, LegiStorm Pro News Aug 2025 article.browser-harness against Delaware iCIS + Florida Sunbiz identified by OSINT Navigator. Surfaced novel LOC NATION JAX LLC (Anisha Hicks, Jacksonville, 2021).Delaware iCIS: only one entity exactly named under "Christina Loren Clement" — file 7127855 CHRISTINA LOREN CLEMENT LLC. No multi-entity Delaware shell network surfaced.
Florida Sunbiz — novel finding. LOC NATION JAX LLC (L21000352541, EIN 87-2134353, ACTIVE) was registered in Jacksonville on August 5, 2021 — pre-dating Clement's federal LDA scheme by three years. Registered agent and manager: Anisha Hicks (not Clement). The "LOC NATION" brand existed in Florida before its federal-level deployment. Could be namesake, adjacent operator, or franchise — needs further investigative work.
Apopka FL address: No Sunbiz entity at 2787 S Orange Blossom Trl, suggesting Clement was listed there personally as a creditor on the Northrop Grumman matrix, not as a registered entity.
The FEC has already acted (warning letter + reclassification). Whether further enforcement (DOJ Public Integrity referral under 18 U.S.C. § 1001 or LDA enforcement under 2 U.S.C. § 1605) follows is an active monitoring target. The Q2 2026 LDA filing deadline (July 20) will reveal whether the pattern continues unaffected.
Promoted from OS-002 spotlight-handoff
Akin Gump's complete 2025 FARA list is six sovereigns (UAE, Cambodia, Japan, Palau, Marshall Islands, Costa Rica) plus the Korea Embassy. None of the four C-003 / D10 engagements — Ant Group (CN), TP-Link Systems (SG/VG/GB), Adani North America (IN), United Solar Polysilicon (CN/OM) — is FARA-covered at firm OR attorney level. Same partners handle both sides; only the government clients get FARA-registered.
The Adani engagement is the strongest "gap should not exist" case. Akin Gump's current Adani engagement has a client effective date of November 21, 2024 — the day after the November 20, 2024 unsealing of the DOJ EDNY indictment of Gautam and Sagar Adani (returned under seal October 24; the LD-1 was posted December 31, 2024); the lobbying targets are White House + State + Commerce + House on bilateral "U.S.-India renewable energy cooperation"; the LDA-named foreign entity is the indicted parent Adani Enterprises Limited. Every subsequent quarterly report through Q1 2026 is a no-activity filing — a one-quarter but precisely timed retainer. NYT (May 18, 2026, nyregion/gautam-adani-indictment-trump.html) reports the DOJ is moving to drop the criminal case; as of July 9, 2026 dismissal is not granted — Judge Garaufis refused immediate dismissal June 26 and DOJ's Rule 48(a) justification is pending. (Corrected 2026-07-09: a prior revision said "registered December 31, 2024 — six weeks after," reading the LD-1 posting date; the filing's effective-date field, re-pulled from the LDA API, is 2024-11-21.)
ae4ebdb2ecf19559) — four LDA filings disclose foreign entities for Ant Group, TP-Link, Adani, United Solar Polysilicon.spotlight-handoff generated brief OS-002 with the four-engagement priority + the unresolved question "is the gap structural or evasive?"spotlight:investigator agent (1 cycle, 110 tool calls, 17 scraped evidence files + 8 Wayback archives via firecrawl) pulled DOJ FARA eFile complete Akin Gump principals list (only 6 sovereigns + Korea Embassy), OpenSecrets FARA aggregator, House LDA reg_id 31784 / client 317840961 (Ant Group), DOJ EDNY indictment, NYT 5/18/2026 reporting, US Treasury Aug 2025 IFC vote, BIS Entity List, Akin Gump partner bios (Pomper, Babin, Ros-Lehtinen, Kho).| Client | Read | Why |
|---|---|---|
| Adani North America | Gap should not exist | Day-after-indictment retainer; executive-branch-only lobbying; named foreign parent under criminal indictment |
| United Solar Polysilicon | Structurally adjacent to active sanctions | Chairman Zhang Longgen was CEO of Daqo (2018-2023) + vice-chair of Xinjiang Daqo (BIS Entity List since June 2021); US Treasury voted against the IFC's $250M loan; Akin Gump lobbies Section 232 |
| TP-Link Systems | Gap is defensibly structural | Delaware US subsidiary; ITC §337 trade defense; commerce exemption typically covers. Newsworthy element: former USTR China Enforcement Chief Counsel is the lobbyist |
| Ant Group Co. | Needs LD-1 retrieval | Q1 2025 LD-2 reports $90K with rendered empty issue text — likely rendering artifact |
Named partners handling both LDA + FARA sides: Brian Pomper (partner, International Trade Policy; former Chief International Trade Counsel, Senate Finance Committee — signed all four foreign-cluster LDA filings), Reginald "Hal" Babin, Ileana Ros-Lehtinen (former U.S. Rep., FL-27), and Stephen S. Kho (former Acting Chief Counsel for China Enforcement at USTR). They FARA-register only the government clients. (Corrected post-publication: original draft mis-cited "Stephen Pomper" — that is a different person at Cleary Gottlieb / International Crisis Group, not the Akin Gump signatory.)
FARA enforcement (22 U.S.C. § 611 et seq.) is theoretically triggered when an agent represents a foreign principal's interests in influencing US government policy. The Adani configuration is the most likely NSD-FARA scrutiny target. Flagged without adjudicating intent.
Promoted from OS-005 spotlight-handoff
Tarplin, Downs & Young is the institutional home of two Medicare Part D legislative architects. ProPublica's 2009 investigation named co-founders Raissa Downs (HHS Deputy Asst Sec for Legislation) and Jennifer Young (HHS Assistant Secretary for Legislation; Sr Counselor to HHS Sec. Leavitt). The firm has carried Amgen, AstraZeneca, BIO, Boston Scientific, Genzyme, PhRMA, Vertex Pharmaceuticals, and BCBS for sixteen years.
The current say-vs-pay loop: Vertex Pharmaceuticals — TDY's $6M+ client — sued HHS-OIG to set aside the agency's Anti-Kickback Statute interpretation of its Casgevy fertility-preservation program (Vertex Pharmaceuticals v. HHS-OIG, 1:24-cv-02046 D.D.C., filed July 15, 2024; Law360 July 16, 2024 coverage quotes Bass Berry & Sims's Jennifer Michael as a third-party expert commentator, not Vertex's counsel). Vertex lost — summary judgment for HHS, March 31, 2025 — and its appeal (D.C. Cir. No. 25-5133, argued February 23, 2026) is pending. A top-five TDY client took the precise agency where two of three TDY founders ran the Office of the Assistant Secretary for Legislation to federal court. (Corrected 2026-07-09: an earlier version said "is in active federal litigation" without noting the district-court loss.)
revolving_door_committee_match + D3 revolving_door_candidates surfaced Tarplin Downs & Young's lobbying revenue concentration on healthcare clients with former-HHS staff covered_position.spotlight-handoff brief OS-005 asked spotlight to verify TDY founders' biographies, identify the firm's full client roster, and check for HHS-OIG enforcement against any client.spotlight:investigator agent (1 cycle, 80 tool calls, 22 scraped files via firecrawl) pulled ProPublica's 2009 investigation as the load-bearing source naming Downs + Young as Part D architects → drug lobbyists, TDY firm website (Wayback archive), Law360 July 2024 Vertex v. HHS-OIG coverage via Bass Berry & Sims, FGS Global Sept 2025 acquisition announcement. Corrected data-detective's D11 attribution from "Pamela Smith" to Raissa Downs.Data-detective's D11 detector originally attributed the HHS Asst Sec for Legislation role to "Pamela Smith"; the actual holder is Raissa Downs. The LDA covered_position field aggregates positions across all lobbyists on a single filing, conflating co-filers. A future cycle should use LegiStorm / Congress.gov individual-staff databases to resolve this rather than the aggregated text. The TDY acquisition by FGS Global on Sept 10, 2025 also explains the Q4 2025 spending-spike D1 flagged.
Promoted from OS-003 spotlight-handoff
Reynolds American — via its subsidiary RAI Services Company, the contributor of record and the $4.42M cumulative lobbying client of Ballard Partners' Daniel McFaul — paid MAGA Inc. (FEC C00892471) $5 million on April 30, 2026. Two days later, a top Reynolds executive and two lobbyists lunched with Donald Trump at his golf club in Jupiter, Florida. On May 8 — within ten days — the FDA issued enforcement-discretion guidance easing its flavored-vape stance. NYT's Kenneth P. Vogel and Christina Jewett reported the timeline; corroborated across four independent secondary sources. (Corrected 2026-07-09: an earlier version cited terminated committee C00825851, a $3.18M through-2023 fee tally, and "reversed its flavored-vape ban.")
McFaul is Managing Partner of Ballard's DC office; he served on the 2016-17 Trump Presidential Transition Team in the Appointments office covering DoD, VA, Armed Services, and Intelligence; and on Matt Gaetz's own congressional transition team. His personal FEC giving (46 contributions, ~$33K, 2017-2026, all Republican) includes Friends of Matt Gaetz, Trump Victory, and Brian Jack (former Trump WH Political Director).
revolving_door_candidates flagged McFaul's covered_position (former Gaetz / Miller / Scarborough Chief of Staff) on Ballard Partners filings.senate_activity_lobbyists + senate_filings aggregated McFaul's full client list (U.S. Sugar $4.56M, Amazon $4.08M, RAI Services $3.18M, Hard Rock Japan, MHP, LA28).spotlight-handoff brief OS-003 prioritised the Gaetz tie, the Reynolds tobacco engagement, and the LA28 Mar-a-Lago angle.spotlight:investigator agent (1 cycle, 52 tool calls, 22 scraped files + 3 Wayback archives via firecrawl) pulled NYT Vogel/Jewett reporting on the Reynolds → MAGA Inc → FDA timeline (corroborated across 4 secondary sources), Ballard Partners firm bio confirming McFaul as DC Managing Partner, FEC committee records for the $5M contribution (initially attributed to C00825851; corrected 2026-07-09 to C00892471 via openFEC), McFaul personal FEC giving (46 contributions), Ballard revenue history via OpenSecrets ($19.6M → $88.3M), and Susie Wiles + Pam Bondi as Ballard alumni now in the administration.The factual chain (donation → meeting → policy reversal) is established. Intent and a specific exchange element are not in the public record. Federal Election Campaign Act pay-to-play, federal anti-bribery (18 U.S.C. § 201), and LDA misreporting are all potentially implicated; flagged without adjudicating intent.
Promoted from OS-004 spotlight-handoff
Anthony "Lazer" Lazarski at Cornerstone Government Affairs — 25-year USAF Colonel; former Senate Armed Services Committee Professional Staffer; former Sr Legislative Advisor to SASC Chair Jim Inhofe — is the named lobbyist on Cornerstone's 2024 contracts with Boeing, General Dynamics, Anduril, AND Apollo portfolio company Arconic Corp. (Corrected 2026-07-09: an earlier version said "Apollo Global Management" — that attribution is OpenSecrets' parent-company rollup of Cornerstone's four Arconic filings; Lazarski has no LDA filing for Apollo itself.)
Sen. Elizabeth Warren (current SASC) attacks the portfolio on two axes:
Triangulation with C-004 — at one corporate remove. Apollo's interests are lobbied along two committee axes by former senior committee staff — Brownstein lobbies Apollo directly on Senate Finance; Cornerstone's Lazarski lobbies Apollo portfolio company Arconic on Senate Armed Services. The same senator (Warren) attacks Apollo from both committee positions.
revolving_door_committee_match (SQL hash a7712c723043a717) surfaced second-tier names beyond Brownstein: Richards, Lazarski, Young, Teague, O'Neill — each $50M+ committee-aligned revenue.spotlight-handoff brief OS-004 asked spotlight to profile the five second-tier lobbyists and identify the cleanest say-vs-pay extension of C-004.spotlight:investigator agent (1 cycle, 79 tool calls, 27 scraped evidence files via firecrawl) pulled OpenSecrets revolving-door profiles + LegiStorm biographies for each of the five, then archived Sen. Warren's 2024 Apollo/Steward Senate letter (Wayback) and her campaign platform "Reduce Corporate Influence at the Pentagon" (Wayback). Lazarski's 2024 client list verified via OpenSecrets + Cornerstone Government Affairs official bio. The triangulation with C-004 emerged from observing Apollo as the common client across Brownstein + Cornerstone, with the same Warren attacking from both committee positions (re-verification 2026-07-09: the Cornerstone-side client is Apollo portfolio company Arconic — OpenSecrets parent rollup — not Apollo itself).data-detective works — phase by phase.The skill pushes ETL, indexing, entity resolution, and aggregation to deterministic Python + DuckDB; it reserves the LLM for reasoning over already-filtered candidates. Every numeric claim cites a SQL query by SHA-16; every named record has an evidence card; every external source is archived. The skill is corpus-agnostic — configuration lives in a Python profile module that maps file globs to mapper functions. What follows is the actual run that produced the nine findings above.
~3 min Full corpus ingest (pyarrow bulk insert; 60× faster than executemany)99.9% Entity resolution via deterministic bridge keys0 tokens LLM tokens spent scanning raw documents during exploration12 SQL-only detectors with SHA-tagged provenance per querySHA-tagged Every numeric claim cites a query hash + provenance JSONPortable Profile-driven — new corpus = new profile.py, same scriptsCross-corpus In-DB joins to FARA + Congress.gov; out-of-band joins to USAspendingMulti-hop SQL D11 + D12 demonstrate three- and four-way joinsdata-detective is structured as eight phases mirroring Spotlight's brief → methodology → cycles → fact-check → gates → handoff arc, adapted for structured records. Each phase has its own sub-skill (or, in Spotlight's case, an external orchestrator). The audit chain is preserved across the data → OSINT handoff so that every claim in this report has a single replicable path back to the source rows.
The orchestrator (data-detective SKILL.md) loaded its sub-skill registry, confirmed DuckDB + pyarrow + lxml available, and inventoried the corpus: Senate LDA JSON (2008-2025), House LDA XML (2008-2025), congressional press releases JSONL, FARA bulk (OpenSanctions DOJ mirror), USAspending DoD contract API, and the unitedstates/congress-legislators member/committee graph. The congress.py profile module declared the file-glob → mapper-function bindings.
ingest sub-skill — load to columnar DuckDB.First run took 740 s/file using executemany. Replaced with pyarrow.Table.from_pydict() → INSERT INTO ... SELECT FROM _batch: full corpus ingests in ~3 minutes. Output: 17 normalized tables, 2.3 GB on disk, 1.5M+ lobbying activities. The XML parser was hardened with recover=True to handle several malformed House quarterly XMLs.
resolve sub-skill — bridge-key entity unification.Senate registrant.id ↔ House senateID first segment is a deterministic bridge — 99.9% match rate measured against ~14,000 House registrant records. No fuzzy matching needed. Member rosters keyed by bioguide_id; committee assignments keyed by thomas_id. Normalized name + state + role tables materialized in DuckDB for cross-corpus joins.
detect + similarity-search + external-data + evidence-cards — the analytic loop.This is the largest phase. Three iterative cycles, each running the same loop: (a) SQL-only detector against the DuckDB index — each query SHA-16 tagged into case-trace/data-detective/anomalies/name/{ranked.csv, provenance.json}. (b) Drill into the top 5-20 candidates via ad-hoc joins. (c) Hop out-of-band to FARA / USAspending / Congress.gov to triangulate. (d) Generate evidence cards (one markdown per cited record). (e) firecrawl-archive every external source. (f) Adversarial fact-check pass.
Detectors that produced findings:
D3 revolving_door_candidates — flagged McFaul (Ballard) → C-008.D4 foreign_filings + D10 fara_gap_narrowed (SQL hash ae4ebdb2ecf19559) — narrowed FARA gap to 4 named engagements including Adani → C-003 + C-006.D5 single_client_juggernauts + D9 shell_pattern — flagged LOC Nation's $20M-per-quarter LDA income pattern ($60M deduplicated) → C-001; the separate $50.25M FEC self-loan is C-005.D8 new_registrant_surge — flagged Anchor & Arrow LLC's 11-client first quarter → C-002; the USAspending cross-reference then surfaced the $22M Apex STRATFI.D11 revolving_door_committee_match (SQL hash a7712c723043a717) — Brownstein + Cornerstone + TDY → C-004, C-007, C-009.D12 committee_say_vs_pay (SQL hash 274c4ef00515cd73) — Warren's 9 Apollo press releases vs Apollo lobbying → C-004.Adversarial fact-check verdicts (Spotlight pattern, applied to data-side claims by an independent fact-checker agent — every source archived, disconfirming evidence surfaced):
| Claim | Verdict | Key fact-checker contribution |
|---|---|---|
| C-001 | verified · high | PacerMonitor independently confirmed CLEMENT v. GARLAND. Surfaced the registrant's own self-disclaimer that the $20M is "pending HR 40 reimbursement." |
| C-002 | verified · high | The Hill + Polaris + HQMC OLA confirmed founders. G2X + GovTribe + AFRL confirmed the $22M Apex STRATFI award. |
| C-003 | partially · medium | Akin Gump × Ant Group traced via House LDA reg_id 31784. Akin Gump IS itself in FARA for other clients; gap is at engagement level. |
| C-004 | verified · high | Brownstein's 2021 Warren-hire announcement confirmed prior roles. OpenSecrets confirmed Sullivan's 2025 client list. Surfaced Mark Warren's May 2026 move to Capitol Tax Partners. |
| C-005 | verified · high | OS-001 verified the FEC reclassification + on-the-record Quinnipiac Prof. John Martin commentary. 11+ Dukes pro se N.D. Texas cases verified via PACER. |
| C-006 | verified · high | OS-002 cross-validated DOJ FARA eFile + OpenSecrets. NYT 5/18/2026 + US Treasury Aug 2025 + BIS Entity List corroboration. |
| C-007 | verified · high | OS-005 surfaced ProPublica's 2009 Part D architects investigation. Vertex v. HHS-OIG via Bass Berry & Sims / Law360. Corrected D11 attribution. |
| C-008 | verified · high | OS-003 confirmed the Reynolds-MAGA-FDA timeline via NYT Vogel/Jewett plus 4 independent secondary sources. |
| C-009 | verified · high | OS-004 confirmed Lazarski's 2024 client list via OpenSecrets + Cornerstone bio + LDA filings. Warren's two attack axes archived on Wayback. |
The orchestrator wrote a synthesis to case-trace/data-detective/summary.md with a findings table, methods summary, and unresolved-gap inventory. Gate 1 protocol: the user must approve before promotion or hand-off. Iteration here patched C-004 (the temporal puzzle of Apollo's 2024 vs 2025 lobbyist) by spawning a targeted re-cycle (OS-006).
report-drafting sub-skill — findings-report.md + report.html + evidence-map.json.Each verified claim was written to findings.json with confidence, fact-check verdict, query-hash provenance, source list, and cross-reference IDs. evidence-map.json indexes every source record cited across findings — the audit ledger used to verify replication paths. The report-drafting sub-skill (added during this investigation) produces all three Phase 5 deliverables from a working template that enforces the design discipline used here: per-finding .path replication block, per-finding .sources strip, novelty + confidence pills, and the phase-by-phase methodology pattern that absorbs the fact-check verdict table inside Phase 3 and the spotlight-handoff outcomes table inside Phase 6.
spotlight-handoff sub-skill — formal data-side → OSINT pass.Six leads were identified for external OSINT amplification. The spotlight-handoff sub-skill generated structured briefs (named entities, known URLs, OSINT angles, falsification protocol, the upstream data-side audit chain) and dispatched a spotlight:investigator agent on each. All six completed. Five surfaced material new evidence and were promoted to data-detective findings (C-005 through C-009).
| Brief | Topic | Spotlight execution | Promoted to |
|---|---|---|---|
| OS-001 | LOC Nation / Rico Dukes | 1 cycle, 22 sources, 23 firecrawl + Wayback archives; browser-harness Delaware iCIS + Florida Sunbiz sweep | C-005 |
| OS-002 | Akin Gump foreign cluster | 1 cycle, 110 tool calls, 17 scraped + 8 Wayback archives; DOJ FARA eFile, OpenSecrets, House LDA reg_id 31784 | C-006 |
| OS-003 | McFaul / Ballard portfolio | 1 cycle, 52 tool calls, 22 scraped + 3 Wayback archives; NYT Reynolds-MAGA-FDA, Ballard bio, FEC C00825851 (corrected 2026-07-09 → C00892471) | C-008 |
| OS-004 | D11 second-tier revolving door | 1 cycle, 79 tool calls, 27 scraped evidence files; OpenSecrets + LegiStorm bios, Warren letter + platform on Wayback | C-009 |
| OS-005 | Tarplin Downs & Young | 1 cycle, 80 tool calls, 22 scraped files; ProPublica 2009 Part D investigation, Law360 Vertex v. HHS-OIG, TDY firm site | C-007 |
| OS-006 | Warren → Capitol Tax Partners | 1 cycle, temporal disambiguation of "Mark Warren" lobbyist career — verified May 2026 firm move | C-004 patch |
vault-ingest sub-skill — Obsidian write-back with --sensitive flag.Mirrors spotlight:ingest: writes one finding-note per verified claim, one source-record note per cited row, one query-hash note per detector hash, plus entity notes for each named person/firm. The --sensitive flag scrubs identifiers for the public vault and keeps the unredacted set in the private vault. This run produced 4 finding notes, 22 source-records, 6 query-hashes, 36 entity notes (real run, dry-run validated first).
| Target | Why |
|---|---|
| Q2 2026 LDA filing deadline (July 20, 2026) | Reveals whether the SOLN GPBC LDA pattern continues unaffected by the FEC action — and whether Apollo also keeps Brownstein now that it has followed Warren to Capitol Tax Partners (LD-1 effective June 29, 2026). |
| DOJ Adani disposition | NYT 5/18/2026 reports DOJ moving to drop the EDNY case. If it drops, the C-006 NSD-FARA configuration becomes tighter. |
| FDA flavored-vape regulatory record | The reversal is the substantive policy hinge of C-008. Federal Register filings would document the procedural reversal. |
| LOC NATION JAX LLC (Anisha Hicks, Jacksonville) | The brand pre-dates Clement's federal LDA scheme by three years. Independent / adjacent / franchise? Unresolved. |
| D.C. Circuit appeal 24-5263 | Filed pro se 11/19/2024, no fee paid. PACER document download via authenticated session. |
| Remaining D11 candidates | James Richards / Anduril; Cornell Teague / Palantir; Jennifer Young / AbbVie; John O'Neill / Blackstone. Each parallel to C-009. |