With $100M in the bank, Bluesky’s Attie hands AI feed control to users
Bluesky has rolled out Attie, an AI-driven app that lets individual users craft their own feeds with natural‑language prompts. Built on the open atproto protocol and using Anthropic’s Claude model, Attie is intended to shift algorithmic control away from platform-optimized engagement and into the hands of account holders — a move that comes as Bluesky secures $100 million in fresh funding and extends its runway for further decentralized infrastructure work.
Launch context: Attie, Anthropic, and a three‑year runway
Attie was developed by Bluesky’s innovation team, led by Jay Graber, and released as a standalone app that connects to the wider atproto ecosystem — users sign in with their Bluesky credentials and can run customized feed rules that other atproto apps can also surface. Technically the app routes user prompts to Anthropic’s Claude AI to translate natural language instructions into filters and ranking behavior, no coding required.
Bluesky announced the $100 million funding round to extend operations for more than three years and prioritize protocol-level privacy and monetization options beyond advertising. The company has explicitly ruled out integrating crypto payments or tokens despite backing from crypto investors, framing that decision as a deliberate risk-avoidance step while it builds moderation and privacy primitives. Bluesky reported growth to roughly 22 million users by late 2024, which sets a scale test for whether Attie’s user-centric approach can attract mainstream adoption.
How Attie changes the agency model for social feeds
Instead of an opaque platform algorithm tuning for time-on-site, Attie lets a user write or speak commands (for example: “only show tech explainers with fewer than 300 words and at least one external source link”) and produces a personalized ranking and filtering layer for that account. Because Attie operates on atproto, those personalized feeds can be read by other apps in the network, making the customization portable rather than locked to the Attie UI.
This design flips two practical levers: where the model runs and who controls the criteria. The AI (Claude) interprets user intent to produce feed behavior, but the user decides the prompts and can export or host account data via Bluesky’s personal data server (PDS) options and custom domain handles. That portability is the concrete distinction from mainstream platforms that centralize both the model and the signals it optimizes.
Governance trade-offs and a quick comparison
Bluesky pairs Attie with a hybrid moderation system: automated detection plus human review and user tools (reply filters, blocking, muting, content labels). That mix aims to prevent abuse without hiding moderation behind a single opaque policy engine, but it also raises operational questions about consistency and scale if personalized AI feeds proliferate.
| Feature | Typical centralized platform | Bluesky + Attie |
|---|---|---|
| Who controls the algorithm | Platform-owned, tuned for engagement metrics | User-defined via natural language; Attie interprets with Anthropic’s Claude |
| Feed transparency | Mostly opaque | Explicit prompts and exportable rules |
| Data ownership & hosting | Platform-held | User can self-host via PDS, use custom domains |
| Moderation model | Centralized policy enforcement | Hybrid automated + human review, user controls for replies/blocking |
| Monetization focus | Ad-driven or data-driven | Exploring subscriptions and hosting services; $100M funding to develop alternatives |
| Crypto involvement | Varies | Explicitly excluded from product plan |
Adoption thresholds and what to watch next
The immediate practical questions are measurable. Does a meaningful share of Bluesky’s user base adopt Attie-style feeds (look for single-digit to mid‑teens percentage points of active users in the first 12 months)? Do third‑party apps read and preserve Attie filters via atproto, showing interoperability beyond a single interface? And can Bluesky maintain moderation response times and appeal processes as personalized feeds create new edge cases?
Common questions
Will Attie just boost engagement for Bluesky? No — the app is explicitly designed to let users craft their own ranking rules rather than let the platform tune a single ranking model for engagement. The test is whether users prefer this agency enough to use it.
Can I host my own data and still use Attie? Yes. Bluesky supports personal data servers (PDS) and custom domains; self-hosting increases privacy but requires more technical setup and maintenance.
When will we know if this model scales? Expect clear signals in the next 12–36 months: adoption rates of Attie, third‑party app interoperability on atproto, and whether Bluesky’s non‑ad monetization experiments start to generate sustainable revenue before the runway wanes.
