Congressional trades were always half the picture. The executive branch — the people who run the agencies that award the contracts, write the rules, and approve the drugs — files financial disclosures too. Almost nobody reads them, because they live in scattered PDFs across the Office of Government Ethics. We built a pipeline that reads them from the primary source. Here is what shipped, and what it found on day one.
The STOCK Act made congressional trades public, and an entire category of tools sprang up to track them. But Congress writes the laws — it doesn't run the government. The SEC Chair decides enforcement priorities. The FDA Commissioner approves drugs. The CMS Administrator runs an agency that moves more money than most countries. When those officials hold or trade securities, the Ethics in Government Act requires them to disclose it — on OGE Form 278 filings that end up as PDFs, often scanned, spread across agency ethics offices and the Office of Government Ethics.
Nothing structured that data. So we built the pipeline that does: discovery across the appointee universe, retrieval of the original filings, extraction directly from the source documents, and normalization into the same signal architecture that already scores 30,000+ congressional trades.
The extraction layer reads the actual filed documents — including the scanned ones — rather than relying on third-party transcriptions. That choice cost us more engineering than any other part of the system, and it was worth it, because our data-integrity rule is absolute: wrong data is worse than missing data. Every extracted transaction carries a provenance anchor back to its position in the source filing. Anything ambiguous — an unreadable amount range, a ticker that doesn't resolve cleanly, a duplicated row — goes to a quarantine table instead of the production dataset. It shows up nowhere until a human or a stricter pass resolves it.
Discovery reconciled more than 1,500 senior officials against the federal appointee universe. The extraction pipeline has already normalized 26,000+ executive-branch transactions, and it drips through the remaining filing backlog daily. Among the officials whose disclosures are now tracked:
Having both branches in one dataset makes a new kind of question answerable — not "what did this person trade?" but "what recurring patterns exist across government trading behavior, and where is each one active right now?" That is the Pattern Library: a public, numbered catalog of behavioral patterns, each with a precise definition, a detector that runs hourly, and an evidence ledger for every signal it produces.
The library launched with these two patterns and 390+ signals, each carrying its full evidence chain. PAT-002 is the one the executive pipeline makes possible: on its first production runs it flagged overlaps like a CMS Administrator's disclosed UnitedHealth position while CMS awards flowed to the company, an FDA Commissioner's disclosed holdings in a life-sciences supplier with FDA-adjacent contracts, and an SEC Chair's disclosed positions in market-data firms that appear in SEC procurement. Every one of those sentences is built from two public records placed side by side — a disclosure filing and a federal spending record. That juxtaposition is the entire claim.
While the executive pipeline was landing, we re-derived the conviction score's thresholds from realized outcomes. The original high-conviction gate was set where intuition suggested; the recalibration set it where the backtests say the win-rate curve actually inflects. The gate moved from 65 to 57, the alert threshold from 75 to 70, and the score buckets were re-cut to match. The effect isn't cosmetic — the high-conviction cohort now tracks the boundary the data supports, and every threshold published on this site was re-verified against the live database as part of the release.
The same discipline applies to every number in our public copy: the alpha figures on the homepage are winsorized (per our methodology), the source count matches the pipelines that actually run, and the cluster rules published on the methodology page are the ones in the detection code — tiered member minimums and all.
The Executive pageis built to answer three questions that were previously unanswerable anywhere: Which tickers have both Congress and a cabinet official positioned in them right now? Which official-to-member matches are actually meaningful — sector-aligned, tight window, high-conviction member — rather than coincidental? And which officials are most worth watching, ranked not by transaction count but by how their agency's jurisdiction intersects active congressional signals?
Multi-branch convergence — the ⚡ signal — already existed as a concept on this platform. What changed is the substrate: it now runs on primary-source executive data covering an order of magnitude more officials, refreshed daily, with provenance on every row. When two branches of government point at the same company at the same time, the feed shows it, and the evidence ledger shows why.
The pattern catalog grows from here — PAT-003 is in development — and the executive dataset keeps deepening as the filing backlog drains. The longer arc is government decision intelligence: correlating policy events with trading behavior across both branches, on the same provenance-first foundation. If you want to kick the tires, the Pattern Library is public, and the executive feed is live on paid plans.