99K deduplicated trades. 61K bills with full text. 940K bill-to-ticker impact mappings. 224M FEC contributions. 3.2M corporate insider events. One 7-layer ML model, one deep-learning Bill Pass Index (AUC 0.74), one stack of LLM behavioral profiles — fused into tiered, labelled signals.
Every congressional stock trade sits at the center of a three-way money flow: donors fund campaigns, committee assignments unlock market-moving information, and trades capture the upside. Each layer is publicly disclosed somewhere — we cross-reference all three so the pattern is unmissable.
Donors · Committees · Trades — three publicly disclosed layers, one cross-referenced map.
Layer 1 · Donors
224M FEC contributions across 5 cycles, mapped to politicians and PACs by 24 industry tags. Plus 2.8M PAC contributions traced back to corporate ownership.
Layer 2 · Committees
3,908 committee assignments cross-referenced with 940K bill-to-ticker impact maps and 17,590 committee meetings. Who has the agenda power and when?
Layer 3 · Trades
99,546 deduplicated STOCK Act disclosures from 354 members, plus 23,408 corporate Form 4 insider events from execs at the same companies.
How it works
From disclosure to signal in three machine-learning passes
Other trackers stop at the trade. We start there — then put every disclosure through a 7-layer weighted scoring model, a deep-learning passage-probability model trained on 37K historical bills, and per-politician LLM behavioral profiles built from 70 features each. You see the insight, not the noise.
Step 01
Congress files a STOCK Act disclosure
Every purchase or sale over $1,000 must be disclosed within 45 days. We ingest all of them in real-time — 99,546 on file. 12,766 were filed late (a violation itself).
e.g. "Pelosi buys NVDA calls · $1M–$5M · filed 38 days after trade"
Step 02
7-layer ML fusion + deep-learning bill passage model
Politician quality (20%) · Herd convergence (20%) · Bill correlation (16%) · Technical context (12%) · Sector momentum (12%) · FEC contribution patterns (10%) · Lobbying alignment (10%). When 3+ layers fire on the same trade, a convergence multiplier (1.3x → 2.0x) amplifies the score. Each ticker also gets a deep-learning Bill Pass Index probability (AUC 0.74, calibrated on 37,132 historical bills with known outcomes).
A Claude Sonnet-trained behavioral profile per politician (70 features each) and per company (iron-triangle map of donors, lobbyists, bills, insiders) explains why the signal fired in plain English. You get tier + direction + convergence + freshness + related bills + 7/30/90-day backtested return — every claim cited to STOCK Act, SEC, FEC, or Congress.gov.
e.g. "STRONG_HERD · BILL_CORRELATED · 7-day lead on markup"
The stack
What's actually under the hood
Most platforms shout "AI" and stop there. Here's the exact data + model stack we run against every congressional disclosure, with row counts and references you can verify.
Layer 1 · Raw federal data
8 direct federal sources, no scrapers
Approved API connections to government data systems — plus licensed market feeds. Refreshed daily, archived forever, fully cited at the row level.
STOCK Act disclosures99,546
Bills (full text, 99.4% pop.)61,601
FEC contributions (5 cycles)224.9M
SEC EDGAR filings indexed391,213
Federal contracts (USASpending)39,408
Daily price bars (FMP+Yahoo)1.94M
Layer 2 · ML models
7-layer weighted fusion + deep-learning Bill Pass Index
Each trade scored across 7 independent intelligence layers; convergence detected when 3+ fire on the same ticker. The Bill Pass Index is a separately-trained probability model, calibrated against 37,132 closed-loop bills (known outcomes).
Claude Sonnet 4.6 profiles every politician on 70 features (donors, sectors, ideology, trading style). Grok 4 maps each company's iron triangle (lobbyists, bill exposure, insider activity). Every claim cited; outputs power the signal narrative and AI chat.
Politician LLM profiles (Sonnet 4.6 v3)2,469
Company LLM profiles (Grok 4 v3)6,151
Bill winners/losers (LLM-derived)6,349
Iron-triangle scores per company0–100
Conflict scores per politician0–100
Layer 4 · The record, in context
Timing, late filings and shared trades, deduped to one feed
Layers 1–3 are joined into one view of each trade: when it was filed, how late, which committees the member sits on, and which other members traded the same company.
Disclosures filed late12.8%
Median disclosure gap28 days
Herd convergence detections46
Markup advance-warning window7 days
What the filings show
1 in 8 trades was filed late.
The STOCK Act gives members 45 days to disclose a trade. Across 99,546 deduplicated disclosures, 12,766 missed that deadline. The median trade surfaces 28 days after it happened; the average, 50. That's the window GovGreed is built to close — by showing who filed, what else was happening at the same time, and what the member oversees.
QuiverQuant ships raw disclosures. Unusual Whales ships options flow. Capitol Trades ships a leaderboard. We ship the ML stack that scores all of it — with cited LLM behavioral profiles on top.
Feature
GovGreed
QuiverQuant
Unusual Whales
Capitol Trades
Congressional trades (STOCK Act)
✓
✓
✓
✓
7-layer weighted ML signal scoring
✓
✗
✗
✗
Deep-learning Bill Pass Index (AUC 0.74)
✓
✗
✗
✗
Bill → ticker impact maps (940K)
✓
Partial
✗
✗
LLM behavioral profiles per politician (Sonnet 4.6)
✓
✗
✗
✗
LLM iron-triangle profiles per company (Grok 4)
✓
✗
✗
✗
FEC contributions (224M rows, 5 cycles)
✓
✗
✗
✗
SEC Form 4 insider events (3.2M cross-refs)
✓
✗
✗
✗
Committee markup alerts (7-day lead)
✓
✗
✗
✗
Cabinet OGE filings + conflict scores
✓
✗
✗
✗
AI chat with live DB access (cited)
✓
✗
✗
✗
Paper trading portfolio (Alpaca pricing)
✓
✗
✗
✗
Free public tools (no login)
✓
Limited
Limited
Limited
Feeling overwhelmed? That's fair — there are 60+ pages under the hood because every data layer has its own view. You only need 4 of them day-to-day. The 5 steps below get you full value in about 10 minutes.
Onboarding
Your first 10 minutes
Each step saves your progress automatically. You'll get full value in about 10 minutes.
1
Follow 3 politicians we flagged as top-signal
S-tier and A+ tier on our 7-layer signal engine. We picked these based on committee access, trade frequency, and signal score.
2
Watch your sectors of interest
Each sector surfaces every related bill in play, federal contracts flowing to beneficiary tickers, and the Executive Branch officials overseeing it — plus which Congress members have been buying in that space.
3
Enable daily alerts
One email at 8am ET — every new signal, trade, and markup touching your watchlist. Zero noise from things you don't follow. Unsubscribe from your account settings anytime.
4
Place your first paper trade
$100K virtual portfolio via Alpaca live price feeds. Click any A-tier signal → Paper Trade → enter size. Zero risk, real price execution, tracks P&L over time. Good way to verify the signals before going live.
Claude Sonnet with live access to the full intelligence stack. Try: "Which bills would benefit NVDA?" · "Has Pelosi ever traded Lockheed?" · "What's in markup this week and who's been buying in that sector?"
Every plan includes public data. Membership adds questions in your own words and alerts on every stock you follow.
Not financial advice · All data from public federal disclosures (STOCK Act, SEC EDGAR, FEC, USASpending.gov, OGE, Congress.gov) ·
Questions? team@mmamodel.ai