Tony Zhou

Bluesky gives users the power to choose their algorithms, but how does this actually work in practice? In our new preprint, we explore the ecosystem of Bluesky custom feeds and the feed creators behind them! 🧵 arxiv.org/abs/2609.12958

We contend that meaningful implementation of middleware for feed recommendation requires reducing these power and cost gaps. We propose new design implications and more in the paper!

Bluesky gives users the power to choose their algorithms, but how does this actually work in practice? In our new preprint, we explore the ecosystem of Bluesky custom feeds and the feed creators behind them! 🧵 arxiv.org/abs/2609.12958

Despite Bluesky's open architecture, we find significant power asymmetries. The platform retains privileged access to granular user interaction data, and only a handful of creators hold the technical expertise to build powerful algorithms without relying on third-party tools.

Bluesky gives users the power to choose their algorithms, but how does this actually work in practice? In our new preprint, we explore the ecosystem of Bluesky custom feeds and the feed creators behind them! 🧵 arxiv.org/abs/2609.12958

As feeds grow, keyword matching isn't enough. Creators evolve their custom feeds by implementing vetted allowlists (to control exactly who can post) and even custom labelers (to filter out spam or AI-generated imagery).

Bluesky gives users the power to choose their algorithms, but how does this actually work in practice? In our new preprint, we explore the ecosystem of Bluesky custom feeds and the feed creators behind them! 🧵 arxiv.org/abs/2609.12958

We identify two key orientations of feed creators: ⚙️ utility-providing and 👥 community-building. ⚙️ Utility-providing creators focus on passive content delivery ("set it and forget it"), while 👥 community-building creators act as active stewards tending to shared social spaces.

Bluesky gives users the power to choose their algorithms, but how does this actually work in practice? In our new preprint, we explore the ecosystem of Bluesky custom feeds and the feed creators behind them! 🧵 arxiv.org/abs/2609.12958

The potential benefits of middleware are substantial, offering users algorithmic choice and decentralizing platform power. But it introduces new questions, such as: who takes on the labor of building these algorithms, and who pays to host this new infrastructure?

Bluesky gives users the power to choose their algorithms, but how does this actually work in practice? In our new preprint, we explore the ecosystem of Bluesky custom feeds and the feed creators behind them! 🧵 arxiv.org/abs/2609.12958

These days, many large social media platforms (Twitter/X, YouTube, TikTok) use centralized, top-down algorithms that maximize engagement through behavioral modeling, leaving users with little control over what they actually see.

📣 Do you create custom feeds on Bluesky? We’re researchers at the @uofwa.bsky.social @socialfutureslab.bsky.social looking to understand feed creator experiences! Interview Details: 1 hour, $20 gift card Sign up: forms.gle/UMNgeoVbUx3e... Info: social.cs.washington.edu/project-page...