This project grabs Hacker News, sorts news items according to the lower bound of a Wilson score confidence interval for a Bernoulli parameter, and spits them out as reformatted HTML. We assume points as upvotes and comments as downvotes.
This project is how I weaned myself off Hacker News.
I've found that the most worthwhile HN articles to read are the ones with high upvote-to-comment ratio. For example, controversial articles e.g. "10 reasons Rails is a fucking joke" tend to have the same number of upvotes and comments. Whereas high-quality analysis and world-changing news generates little discussion but much upvoting.
Reading just the top 5 stories ranked according to this score saves you a ton of time.
Written with Scrapy, found here.