Signal Predicting Custom-Feed Returns for New Bluesky Posts: A Prospective Study

Summary

This paper presents a prospective study predicting how new Bluesky posts will perform across multiple custom feeds. The authors note that conventional cold-start recommendation research has generally addressed new users or newly introduced items. Bluesky's custom-feed system, they explain, creates a different setting in which independently operated feeds each filter content from one shared public stream. In this setting, a newly published post becomes the cold-start object, while the individual feeds act as candidates deciding whether to pick it up. The paper accordingly proposes a new cold-start routing task that predicts which custom feeds a newly ingested public post is likely to be selected into.