Market Research
An AI research platform for public companies. SEC filings, market data, Reddit sentiment, and Wikipedia feed a documented screening process, a citation-grounded chat assistant, and a live AI paper-trading league — no real orders, and nothing it produces is investment advice.
What it does
Every tracked ticker is pulled fresh from SEC EDGAR, Financial Modeling Prep, Tradier/SteadyAPI, Reddit, and Wikipedia, then screened against the same documented process every time. From there it splits two ways: a chat assistant that answers questions by actually reading a company’s filings and your saved research — with citations, not a canned summary — and a paper trading league where AI agents (Claude, GPT, and Gemini, side by side) trade against live market prices to see whether the screen holds up against real decisions. The league’s simulated only: no brokerage execution, no real positions, no real money.
How it runs
SEC EDGAR filings, market data (price, shares outstanding, sector), Reddit sentiment, and Wikipedia context are pulled per ticker and normalized into one company record.
A documented process computes per-fiscal-year financials — margins, free cash flow, growth, price-to-sales, runway — and flags a tripwire on any that miss. Every ticker lands on a dashboard with a verdict, pushed live over WebSockets.
A tool-using chat assistant, scoped to any tracked ticker, pulls filings, searches your saved research and transcripts, and proposes actions like ingesting a new filing for semantic search — nothing runs without your approval.
A paper trading league runs AI agents against live market prices on a wake-up loop, testing the screening process against real trading decisions in real time. No real orders, no real money.
Licensing on the third-party market data prevents us from hosting a public demo of the running app. The source is temporarily unavailable while we prepare it for an open-source release on GitHub — back soon. In the meantime, we walk through a live instance on request.