Haz Stonks is a retail investor sentiment tracker built to answer one question: can the signal buried in social media stock discussions be quantified before it moves prices? The idea took hold after January 2021, when retail investors drove GameStop from $20 to nearly $500 and proved that collective online sentiment is a real, measurable market force. I wanted to know whether a social media stock sentiment tool could surface that kind of momentum early enough to matter.
The result is an experiment, not a product — an automated pipeline that ingests public financial discussions twice daily, scores them with sentiment analysis, and blends the output with a machine learning model trained on price history and technical indicators.
Twice daily an automated pipeline scrapes social posts from Bluesky and Mastodon, financial news from RSS feeds (CNBC, Seeking Alpha, Yahoo Finance, MarketWatch), and price history from Yahoo Finance. Each piece of content is scored with VADER sentiment analysis — a model tuned specifically for social media text — and enriched with finance-specific vocabulary (words like "mooning", "bagholder", and "short squeeze" carry their actual market meaning).
The sentiment scores are combined into a hype signal that accounts for both the direction and volume of discussion. Separately, an XGBoost model trained on historical data evaluates technical indicators — RSI, MACD, Bollinger Bands, volume anomalies, and price momentum — alongside institutional ownership and short interest data from SEC filings.
The final signal blends both: 70% technical model + 30% hype signal. This gives primacy to quantifiable price action while still letting crowd sentiment push the needle when it's strong enough to matter.
Automated sources
social
social
news
price + news
+ Community
moderated · low weight
hype score
RSI · MACD · Volume
stored in Cloudflare D1 · served here
Most retail sentiment tools focus on Reddit or X. Haz Stonks targets Bluesky and Mastodon — open, federated platforms that have grown significantly among financially engaged users since 2023. Bluesky stock sentiment in particular tends to be more discursive and less meme-driven than Reddit, which means VADER picks up cleaner directional signals with less noise to filter out. It also gives the tool a differentiated data layer that most existing services don't cover. For a deeper look at how to interpret the scores, see the Signal Interpretation Guide.
Community members can submit short tips about individual stocks via the Tipline tab on any ticker's detail page. Tips are screened for harmful content by an automated moderation system before being published, and carry low weight in the hype calculation to prevent manipulation. Anyone can report a tip that violates community standards.
This is a research experiment to explore whether retail noise contains exploitable signal — and to make that exploration visible. The signals shown are not buy or sell recommendations. Past sentiment patterns are not a reliable predictor of future price movements. Nothing on this site is financial advice.
The site is fully automated, runs on a free-tier serverless stack (Cloudflare Workers + D1), and updates twice daily. Data coverage spans the most actively discussed retail stocks on major English-language platforms.