Ever notice how a single offhand comment from a public figure can move a stock within hours? I wanted to actually track that — not just react to headlines after the fact, but systematically catch mentions and see what happens to the price next.
So I built Wire Desk: a small automated pipeline that watches for statements referencing specific companies or sectors, classifies them as direct (his own words, via Truth Social/White House) vs. indirect (news coverage reporting on him), then follows the stock's price for four trading days so I can judge the reaction myself instead of trusting a headline.
🔗 Live site: https://akshay-bhatnagar-05.github.io/trump-stock-tracker/
🔗 Source: https://github.com/akshay-bhatnagar-05/trump-stock-tracker
How it works
- Three sources monitored daily — Truth Social (via an RSS mirror), official White House press releases, and news wires via NewsAPI
- Keyword matching against a watchlist of companies/tickers I care about
- Direct vs. indirect tagging — this turned out to be the most useful design decision. A statement from someone carries very different weight than a journalist writing about them, and treating those the same was the biggest flaw I noticed in similar tools
- 4-day price tracking via yfinance, so I can see the actual before/after reaction instead of just the headline
- Daily email digest + a live dashboard, fully automated via GitHub Actions — the whole thing runs on a schedule with zero manual steps
The stack
- Python (requests, yfinance, feedparser) for the detection/tracking logic
- GitHub Actions for the daily cron job
- A static HTML/CSS/JS dashboard, reading from a JSON file the script regenerates every run
- GitHub Pages for hosting — the whole thing costs $0 to run
What I deliberately avoided
While researching this, I found an open-source project that brute-forced 31.5 million rule combinations against historical posting data to find a "61.3% hit rate." Their own README honestly flags the risk: testing that many combinations basically guarantees some will look great by pure chance (the classic multiple-comparisons/data-snooping problem). I wanted the opposite — simple, explainable logic over a big number that might just be noise.
Honest caveats
This is a research/awareness tool, not a trading signal. By the time a mention is caught here, it's already public, and a "confirmed" multi-day trend is usually already priced in. It's useful for staying informed fast — not for predicting anything.
Would love feedback on the architecture, or if anyone's built something similar with different source coverage. Repo's open, feel free to fork it.
Tags: #python #githubactions #webdev #opensource #finance
Top comments (0)