GROWTH / DATA · Build time: a weekend · Running cost: a few dollars a month in LLM calls · Level: comfortable with Python and SQL
What you'll build
A pipeline that checks the US House's financial disclosure site every morning, downloads new stock-trade reports filed by members of Congress, turns each PDF into clean rows (who, ticker, buy or sell, amount range, date), stores them in Postgres, and produces a short daily briefing of the most notable trades.
The output is a content engine: a newsletter, an X account or a page that people check every day because nobody else presents the data this cleanly.
Why it works
The data is public but painful. Under the STOCK Act, members report trades within 45 days. The reports are PDFs, some typed, some scanned. Whoever makes them readable becomes the source.
The curiosity is built in. People want to know what lawmakers are buying, and every big trade is a ready-made headline.
It compounds. Each day adds to a searchable history, which makes the archive more valuable than any single post.
The stack
Job | Tool |
|---|---|
Find new filings | The House Clerk's yearly index file (XML) |
Read PDFs | pdfplumber, with an LLM vision fallback for scans |
Structure the data | Any capable LLM with JSON output |
Store and query | Supabase (Postgres) |
Run daily | GitHub Actions cron |
Publish | Your newsletter platform, plus X |
Read this before you monetise. US law (5 U.S.C. § 13107) restricts using these disclosure reports for commercial purposes, with an exception for news and communications media publishing to the general public. A public newsletter fits that exception more comfortably than, say, selling a raw data feed to trading firms. Speak to a lawyer before you charge for the data itself. This is not legal advice.
Subscribe to our premium content to read the rest.
Become a paying subscriber to get access to this post and other subscriber-only content.
Upgrade