Intelligence Infrastructure

The machine behind
the briefings.

Public records are public for a reason. The problem has never been access — it has been scale, fragmentation, and the absence of infrastructure capable of resolving relationships across incompatible federal and state databases simultaneously. We built that infrastructure.

280B+

Raw data points ingested

500K+

Influence chains mapped

204

Sovereign jurisdictions tracked

Source schema formats normalized

Federal & state record systems

72M+

Property parcels cross-referenced

What we do

We monitor 11 federal and state record systems continuously — FARA, FEC, Senate LDA, SEC EDGAR, county property appraisers, Secretary of State filings, congressional financial disclosures, and state voter rolls. Every new filing is ingested, normalized across 23 incompatible schema formats, and resolved against a master entity graph that links people, organizations, addresses, and filings across all sources simultaneously.

We identify relationships no single database can show you. The foreign agent registered with DOJ whose firm name also appears in 60 million FEC contribution records as a donor employer — that connection does not exist anywhere in government data. It only exists in our graph, because we built the resolution layer that creates it. The same applies to the shell company that surfaces in both a bulk foreclosure acquisition and a Secretary of State filing two states away under a marginally different name. Or the congressional staffer whose STOCK Act periodic transaction report discloses a position in a company currently under their committee's oversight jurisdiction.

We score every detected relationship by financial magnitude, political seniority, country-of-origin risk classification, and temporal proximity between lobbying activity and donation dates. The highest-scoring chains become briefings. Every briefing cites the exact filing it came from. Nothing is inferred. Nothing is speculated. If the government record says it, we report it. If it doesn't, we don't.

The intelligence pipeline

Six stages from raw government filing to published briefing.

01Acquisition▼

Continuous multi-source acquisition across 11 federal and state record systems. Asynchronous ingestion pipelines maintain persistent authenticated sessions with each upstream authority, executing structured document retrieval under jurisdictionally-appropriate request cadences. Delta detection compares checksums against versioned snapshots to isolate net-new filings without full corpus re-ingestion.

02Normalization▼

Raw government records arrive in 23 distinct schema formats — XML, CSV, fixed-width, HTML tables, PDF, and proprietary bulk feeds. A multi-pass normalization engine deduplicates across source-specific ID namespaces, resolves inconsistent date formats, applies jurisdiction-aware address standardization (USPS CASS-compliant), and canonicalizes entity names through phonetic hashing and edit-distance clustering before committing to the master record store.

03Entity Resolution▼

The core challenge: a FARA registrant listed as "J. Smith Consulting LLC" and an FEC donor employer listed as "John Smith Consulting" are the same entity — but neither database knows the other exists. Our entity resolution layer applies a hybrid deterministic-probabilistic matching model: exact EIN match → employer name fuzzy match (Jaro-Winkler ≥ 0.91) → registered agent cross-reference → address tokenization → phonetic fallback. Resolution confidence scores drive tiered verification queues before any relationship is asserted.

04Graph Construction▼

Resolved entities are loaded into a directed multigraph where nodes represent people, organizations, addresses, and filings — and edges represent typed relationships: REGISTERED_AS, DONATED_TO, LOBBIED_FOR, OWNS_PROPERTY, FILED_WITH, SHARES_ADDRESS_WITH. The influence chain detection algorithm performs multi-hop traversal (depth 1–4) across the FARA → FEC edge class, weighting paths by dollar magnitude, temporal proximity, political office seniority, and country-of-origin risk tier. Over 500,000 chains have been mapped across 204 sovereign jurisdictions.

05Scoring & Ranking▼

Each detected chain receives a composite virality score derived from six independent signal dimensions: aggregate financial magnitude (log-scaled), recipient political office level (POTUS → municipal), country of origin classification (OFAC/FATF risk tier), temporal proximity between lobbying registration and donation (days delta), media salience of named entities (prior coverage density), and pattern novelty versus the existing published corpus. Scores above threshold route to the editorial generation queue; sub-threshold chains are retained in the graph for future multi-hop compound detection.

06Editorial Generation▼

High-scoring chains are structured into a canonical briefing schema — filing citations, dollar amounts, entity names, office held, registration dates — and passed to a large language model operating under a strict factual constraint: no inference, no attribution beyond what the filing states, no narrative embellishment. Every output sentence is grounded in a specific document reference. The model's role is structure and clarity, not interpretation. Human review gates publication for anything touching a sitting federal official.

What comes out the other end

The output is intentionally simple. The complexity lives in the infrastructure.

→A headline. A dollar amount. A name. A filing number. A date.

→No unnamed sources. No background characterizations. No editorial interpretation.

→Every claim in every briefing traces to a specific government document.

→If the document doesn't say it, we don't say it.

→The record is the story. We just found it.

Record systems monitored

Source	Scale	Refresh
DOJ FARA Foreign Agents Registration Act	7,050+ active registrants	Weekly delta
FEC Federal Election Commission	60M+ contributions	Monthly bulk + daily delta
Senate LDA Lobbying Disclosure Act	1.9M+ filings	Daily (non-headless session)
SEC EDGAR 13F / 13D-G / Proxy filings	Full corpus	Quarterly bulk + daily 8-K
County Appraisers Property ownership & assessment	72M+ parcels, 35+ states	Monthly per-county
State SOS Business registrations & officers	10 states, expanding	Weekly delta
Voter Rolls State voter registration files	85M+ records	Post-election cycle
eFD / STOCK Act Congressional financial disclosures	2012–present	Within 48h of filing

Corrections policy

Because every claim is tied to a specific filing, factual disputes are resolvable against the source document. If a government record has been updated, amended, or if our entity resolution produced an erroneous match, we will issue a correction prominently and immediately.

[email protected]

Tips & document drops

Have a lead on a lobbying arrangement, beneficial ownership chain, or property transaction that should be in our database? We accept tips and document submissions. All sources treated confidentially.