How it works

The Idea

Every day, millions of conversations happen across the internet. We built this site to track how people feel about key topics over time by analyzing discussions on Reddit and Bluesky. Why those two? Their APIs were free.

The Methodology

We start with a curated list of 100 terms that we want to track. Every day, around midnight MST, we search our platforms for new posts mentioning these terms. Each post is analyzed using an aspect-based sentiment model, which evaluates how the author feels specifically about the tracked term rather than the overall tone of the post.

For each term, we calculate an average sentiment score for Reddit and Bluesky posts from that day. These platform-level scores are then combined using a weighted average based on the number of posts collected from each platform, ensuring that sources with more discussion contribute proportionally to the final score.

The resulting sentiment values are scaled to create scores ranging from -1 (very negative) to 1 (very positive).

Notes

• Sentiment scores are weighted by the number of posts collected from Reddit and Bluesky, but not by likes, upvotes, reposts, or other engagement metrics.
• We use a plaintext search, so we do not analyze related terms, synonyms, or implied references—only posts containing the tracked term itself.
• The scoring model we're using performs well, but language is nuanced and automated sentiment analysis remains inherently imperfect.