The Rise of Bluesky

The Internet has been hosting social conversations around discussion topics and common interests since at least the inception of Usenet News. Later, social networks such as Twitter (now X) added a layer of social networking that brought some structure to the content presentation. It now depended on which content creators were followed by each content consumer, with individual feeds merged into a common timeline for each consumer. A final step was to substitute the timeline with algorithms that sought to maximize engagement. A step that had dire consequences.  

Imagine you regularly go to your favorite coffee shop and have quiet conversations with your friends. Occasionally, a new friend of friends is introduced and joins the conversation. Many tables are doing the same in an open space, and you can sometimes hear conversations nearby and maybe engage with these neighbors. This is a simplified analogy to what happened with non-algorithmic social networks. Now imagine that a table has a particularly polarized discussion, maybe arguing about politics, and the persons at that table start talking louder and more emotionally, now all the tables are hearing that heated conversation as it disrupts and replaces the quiet exchanges at each table. In the physical world, these disruptions occur, but rarely last long and normality resumes. With the algorithmic selection of content, permanently optimizing for emotion and engagement, these heated conversations will become the norm, and the quiet coffee shop can become a stressful and alienating place.

Platforms such as X and Threads might provide some sort of chronological feed while aggressively defaulting to the algorithmic feed that maximizes engagement. Having that default reduces the incentive for users to curate who they follow to have a high-quality chronological feed experience. Alternatives that provide a purely chronological timeline of who you follow are present in Mastodon and Bluesky. Both support a more open approach to social conversations based on open protocols, even opening the possibility for user-side management on how content is selected for the user feed. Mastodon is more federated; users can select which server to join; posts are propagated among servers that agree to interconnect. However, this creates some barriers to presenting a unified view of what happens across all servers. Bluesky offers a more user-friendly presentation that mimics the old Twitter user experience from the 2010s.

Until very recently, both these platforms lacked a critical mass of users to be more than a niche alternative to the incumbents. In November 2024, a change started to become visible. Bluesky experienced very sharp growth in users and activity and even exceeded that of Threads in the U.S. Contributing factors might have been geopolitical, with the temporary ban of X on Brazil and the U.S. elections, but there are also technological reasons.

Pure timelines depend on careful curation of who you follow, and require time for users to look for other users, check if they like their content, and make the decision to follow them. In August 2024, Bluesky introduced the concept of Starter Packs. Any user can curate a starter pack on a topic (e.g., Probabilistic Data Structures) and link to users active on the topic. Other users can look at the starter packs, follow all or individual members and preview the aggregated timeline of the starter pack users to get a taste of the content. At the time of writing, more than 70k starter packs exist, and there are searchable directories that help find them, such as the BlueskyDirectory.

This combination of technological and political factors, plus reaching a certain critical mass threshold, might explain the big migration that seems to be underway. The scientific Twitter community has been longing for a long time to find a successor. Some decisions taken on X, like bringing less algorithmic visibility to posts with links, made it very difficult to share links to papers and systems to discuss them. A recent text stated that Bluesky is emerging as the new platform for science. It does seem that a migration is occurring, but more time is needed to confirm the trend.    

Acknowledgements

I want to thank Sérgio Nunes and Salvatore Sanfilippo for their comments on improving this text.

Carlos Baquero is a professor in the Department of Informatics Engineering within the Faculty of Engineering at Portugal’s Porto University and is also affiliated with INESC TEC. His research is focused on distributed systems and algorithms.