Introduction

Many social platforms analyze user data to deliver targeted recommendations or advertisements. As this functionality becomes more purpose-driven, users have gradually shifted from initial excitement to frustration, emphasizing personal data protection. However, as long as social platforms remain under centralized management, users are stuck in a “black box,” unknowingly surrendering their usage preferences. This situation has sparked the emergence of a new model combining decentralization with social networking — Bluesky stands as a leading example.

Funding Background

In July 2023, Bluesky raised $8 million in seed funding. Later, in October 2024, the company completed its Series A funding round led by Blockchain Capital, with SevenX Ventures, True Ventures, and Alumni Ventures participating as co-investors.

Team

Jay Graber serves as the CEO of Bluesky and previously worked as a software engineer for Zcash and the blockchain company Skuchain.

One of the co-founders, Jack Dorsey, is the co-founder and former CEO of Twitter. He is also the CEO and Chairman of Square and Block. Jack Dorsey has been recognized as one of Time magazine’s 100 most influential people and named an outstanding innovator under 35 by MIT’s Technology Review.

Project Overview

Bluesky is a social application similar to X and Facebook. Launched in February 2023 as an invite-only beta, it rapidly grew to over 10 million registered users in just 20 months. Bluesky Social PBC (Public Benefit Corporation) develops the official client applications and manages core services. The platform embraces open-source principles—both its client and multiple server-side components use MIT and Apache 2.0 licenses, while its protocol follows open standards. Independent third parties develop and operate various system components, including feed generators and alternative clients.

Bluesky aims to build a new foundation for social networks, enabling creators to remain independent of platforms, developers to innovate freely, and users to make autonomous choices in their experiences. Its “official” client application supports iOS, Android, and web versions. Users can interact with posts by replying, reposting, or liking and following other users. By default, users have two feeds: one displaying posts from followed accounts in reverse chronological order, and another recommending trending content across the network. Users can also select alternative feeds showcasing content on specific themes.

Core Feature – AT Protocol

Bluesky is built on the AT Protocol, which empowers users with data sovereignty by consolidating all their social information into a unified structure. It leverages non-black-box algorithms to avoid irrelevant recommendations and advertisements.

AT Protocol Projects

For social protocol developers, the AT Protocol provides an open-source framework for modular social applications, allowing for easier and more professional deployment. The AT Protocol is designed to support various social models, not limited to Bluesky. For instance, beyond a Twitter-style microblogging app, AT can facilitate Reddit-style forums, long-form blogging with comments, or specialized social apps like link sharing or book reviews. All these applications can share the same user identity, social graph, and user data storage servers.

Based on the AT protocol, here are Bluesky’s main features:

• Custom Feeds and Algorithm Choice

Cartoon illustration of custom feeds and algorithm choice

In daily use of social platforms, recommendation algorithms are often a love-hate feature. Many platforms rely on advertising for revenue, using proprietary black-box algorithms to push trending or ad-related content.

Some decentralized social networks counter this by offering a reverse chronological timeline from followed accounts, avoiding the opaque recommendation algorithms of mainstream networks. For instance, Mastodon claims it has “no algorithms or ads to waste your time.”

Bluesky, however, argues that the issue isn’t algorithms themselves but the centralized, opaque ones that reduce user autonomy, prioritizing engagement at the expense of other factors—such as promoting controversial posts to capture attention. A good recommendation algorithm can help users discover meaningful content and find new accounts to follow. This is especially crucial for new users who haven’t yet built a following. Recommendation algorithms also help highlight content on specific topics, whereas following a user means seeing all their posts, which may span various themes and not appeal to all followers. Bluesky aims to create an open and diverse algorithm marketplace, enabling the community to tailor systems to their needs and giving users more control over how they allocate their time and attention.

Bluesky Social PBC provides several official feed algorithms while allowing anyone to create their own feed generator. To date, tens of thousands of custom feeds have been created. Feed generators can select content using any criteria, and with the introduction of custom feeds and algorithm choices, users can personalize content recommendation models to suit their preferences.

• User Data Sovereignty

Cartoon illustration of user data sovereignty

A key advantage of Bluesky over traditional social platforms is that it gives users complete control over their data. Users—not platforms or third parties—decide how their creations can be used commercially. This data sovereignty provides greater freedom to all users, particularly creators, by protecting their intellectual property rights.

According to the white paper, all data a user wishes to publish is added to their repository, which stores a set of records. Each user action (such as posting, liking a post, or following someone) generates a corresponding record in their repository. These records are encoded using DAG-CBOR, a compact binary data structure in a constrained CBOR format. The schema for records is defined by dictionaries (Lexicon), allowing repositories to include records from multiple dictionaries representing user behaviors across different social modes. Media files, such as images, are stored outside the repository but are referenced in records via CIDs (essentially cryptographic hash values). Similarly, references to records in other repositories (e.g., identifying a liked post) include its CID.

Each user account has a repository containing all their actions, except for records the user explicitly deletes. A Personal Data Server (PDS) hosts the user’s repository and exposes it as a network service.

Users update only their own repositories. For instance, if User A follows User B, this generates a following record in User A’s repository but does not alter User B’s repository. To find everyone following User B, one would need to index the contents of all repositories. This design is similar to hyperlinks on the web: while it’s easy to find all outbound links from a webpage, finding all inbound links to the page requires indexing the entire web, a task typically handled by search engines.

• Personal Data Server (PDS)

Relationship between primary service providers and data streams

The Personal Data Server (PDS) is responsible for storing a user’s data repository and related media files, allowing anyone to query the hosted data through an HTTP API. Additionally, the PDS provides a real-time update stream via WebSocket, enabling low-latency notifications for newly added or deleted records (such as posts, likes, or follows) in the repository.

The computational resources required to host a PDS for a small number of users are minimal, even for users with large followings. As a result, users who wish to self-host their PDS can opt to run it on low-cost virtual machines in the cloud or even on devices connected to home routers.

PDS operators typically perform basic content moderation, removing any illegal content hosted on their servers. However, in the AT Protocol system, content moderation at the PDS level is less critical, as the primary responsibility for moderation is handled by other roles in the system—such as labeling services and feed generators. This separation allows different entities to provide hosting and moderation services independently.

• Labeling Services and Feed Generators

The AT Protocol separates the “opinionated” parts of the system into independent services: Labeling Services (Labelers) and Feed Generators. Labeling Services produce streams of content assessments (e.g., “This post is spam”), while Feed Generators return lists of post IDs selected for inclusion in custom content feeds. Users can choose their preferred feeds and labeling services within the client application.

Conclusion

Unlike traditional centralized social platforms like X and Facebook, Bluesky offers users and developers more flexible tools for social applications. As the traditional social world increasingly emphasizes user sovereignty and data control, decentralized platforms like Bluesky and the AT Protocol—focused on protecting user data, minimizing intrusive ads, and aggregating social data—are poised to replace traditional social protocols gradually.