What it does
Stocks View aggregates quotes, fundamentals, and news from free-tier market data APIs, then uses a large language model to correlate that news against the stocks you follow. The output is a set of structured directional hypotheses — a bias (bullish / bearish / neutral / mixed), a conviction score, a time horizon, an expected magnitude, the reasoning behind it, and the key risks that would invalidate the thesis.
The goal is to surface where to look and why — to compress a morning of reading into a ranked, sourced read on your watchlist.
The prediction log is the point
Every analysis is written to a prediction log alongside the price at the time it was made. A scheduled job later scores each call against the actual 1-day and 3-day returns and marks it a hit or miss. That feedback loop is the whole project: it lets you see — on the public calibration page — whether stated conviction actually tracks reality. If 70% conviction doesn't hit ~70% of the time, the model is noise and the approach gets redesigned rather than trusted.
Where the data comes from
Quotes, fundamentals, and analyst data come from Finnhub; macro series from the Federal Reserve (FRED); filings from SEC EDGAR; with Stooq and Google News RSS as no-key fallbacks. News is summarized and analyzed by the configured LLM backend (Google Gemini or Anthropic Claude).
What it is not
Stocks View produces signal and synthesis, not trading advice. It does not place trades, integrate with any broker, or execute orders — and never will. Nothing here is a recommendation to buy or sell any security. Directional hypotheses are informational only; do your own research and consult a licensed professional before making investment decisions. See the Terms of Service and Privacy Policy.