Bluesky, the decentralized social media platform that has positioned itself as the anti-algorithm alternative to X, is building an AI assistant. Not to generate posts. Not to summarize threads. To help users construct custom feeds — the very feature that distinguishes Bluesky from virtually every other social network on the market.
The product, announced on Wednesday, represents a fascinating philosophical pivot for a company that has spent its short life championing user control over algorithmic curation. Or maybe it doesn’t. The truth, as usual, is more complicated than the headline suggests.
An AI That Builds Algorithms Instead of Being One
Here’s the core idea: Bluesky already lets users create and subscribe to custom feeds through a system it calls “custom algorithms.” These feeds can surface posts about specific topics, from particular users, or based on engagement patterns. The problem is that building one requires technical knowledge most users don’t have. You need to understand Bluesky’s AT Protocol, write code, or use third-party tools like Skyfeed to piece together filtering logic.
The new AI assistant, as reported by Engadget, is designed to eliminate that barrier entirely. Users will describe what they want to see in plain language — “show me posts about European soccer but not transfer rumors” or “surface independent game developers who post screenshots” — and the AI will translate those instructions into a functioning custom feed. Think of it as a natural-language interface for algorithmic construction.
Bluesky CEO Jay Graber framed the tool as consistent with the platform’s founding principles. The company has long argued that the problem with social media isn’t algorithms themselves but the fact that users have no say in how those algorithms work. Facebook decides what you see. X decides what you see. TikTok’s For You page is a black box optimized for engagement, not satisfaction. Bluesky’s answer has been to make the algorithm layer open and user-directed. The AI assistant simply makes that openness accessible to people who can’t code.
That’s a meaningful distinction, and one the company clearly wants to emphasize.
But it also raises a question that Bluesky hasn’t fully answered: what AI model powers this assistant, and what data does it train on? The company has not disclosed whether it’s building its own model, licensing from a third party like OpenAI or Anthropic, or using an open-source foundation model. For a platform built on transparency and decentralization, those details matter enormously. Users who fled X and Meta platforms over data privacy concerns will want specifics before handing their content preferences to an AI system.
The timing is notable. Bluesky has been on a growth tear since late 2024, when waves of users abandoned X following various controversies involving Elon Musk. The platform reportedly crossed 30 million users earlier this year, a figure that still pales next to X’s claimed 600 million but represents serious momentum for a startup that only opened to the public in February 2024.
Growth brings expectations. And investors who participated in Bluesky’s $15 million Series A round are watching to see how the platform translates user acquisition into something durable. Custom feeds are Bluesky’s signature feature, the thing that makes it genuinely different rather than just “Twitter but nicer.” Making that feature usable by mainstream audiences — not just developers and power users — is arguably the single most important product challenge the company faces.
The Competitive Context: Everyone Wants to Be Your Curator
Bluesky isn’t building this AI assistant in a vacuum. Every major social platform is grappling with the same tension between algorithmic curation and user agency, and most are reaching for AI tools to resolve it.
Meta has been integrating its Meta AI assistant across Facebook, Instagram, and WhatsApp, though its focus has been on conversational AI and content generation rather than feed customization. X launched Grok, its AI chatbot built by xAI, which can summarize trending topics and answer questions but doesn’t offer granular feed-building capabilities. TikTok continues to rely on its recommendation algorithm, widely regarded as the most sophisticated in the industry, but gives users almost no direct control over how it works beyond crude “not interested” buttons.
None of these approaches do what Bluesky is proposing. The closest analog might be Reddit’s community-based structure, where users self-select into topic-specific forums. But Reddit’s model is organized around communities, not personalized algorithmic feeds. Bluesky’s vision is more granular: every user gets to be their own algorithm designer.
The risk, of course, is that most users don’t actually want that responsibility. The history of consumer technology is littered with products that offered maximum customization and lost to simpler, more opinionated alternatives. RSS readers gave users total control over their information diet. Google Reader was beloved by its users and killed by Google because those users numbered in the low millions. TikTok’s entirely passive, zero-effort feed became the most addictive product in social media history.
Bluesky is betting that an AI layer can thread this needle — giving users the power of customization without the burden of configuration. Describe what you want in a sentence. The machine does the rest. It’s an elegant theory.
Whether it works depends on execution details the company hasn’t shared yet. How well does the AI interpret ambiguous requests? Can it handle nuance — distinguishing between, say, political news and political commentary? What happens when a feed generated by AI surfaces content the user finds objectionable? Does the system learn and adapt, or is it a one-shot generator?
These aren’t hypothetical concerns. They’re the exact problems that have plagued algorithmic curation at every platform that’s attempted it.
There’s also a philosophical wrinkle specific to Bluesky’s decentralized architecture. The AT Protocol, which underpins the platform, is designed so that anyone can build services on top of it — including feed generators. Third-party developers have already built dozens of custom feed tools. Introducing a first-party AI assistant could be seen as Bluesky competing with its own developer community, the very people who helped make custom feeds viable in the first place.
Graber and her team will need to handle that dynamic carefully. Developer goodwill is a finite resource, and Bluesky’s open-protocol pitch depends on maintaining it.
What This Signals About Bluesky’s Business Model
The AI assistant announcement also offers indirect clues about Bluesky’s evolving business strategy. The company has been notably vague about how it plans to make money. It has rejected advertising as a primary revenue source, at least in its current form, and has floated ideas about paid subscriptions and premium features.
An AI-powered feed builder fits neatly into a premium tier. Basic custom feeds could remain free; AI-assisted creation could sit behind a subscription. This is speculative — Bluesky hasn’t announced pricing or packaging — but the pattern is familiar across the tech industry. AI features are increasingly the wedge that justifies subscription revenue. Microsoft did it with Copilot. Google did it with Gemini in Workspace. Bluesky could follow the same playbook at a much smaller scale.
And the company needs revenue. Running a social network isn’t cheap, and $15 million doesn’t last forever, especially when you’re scaling infrastructure to support tens of millions of users. Bluesky will need to raise again or generate meaningful income within the next 12 to 18 months. Having a differentiated AI feature that users are willing to pay for would make either path considerably easier.
The broader signal here is that Bluesky is no longer content to be a niche alternative for disaffected Twitter users and protocol enthusiasts. It’s building for a mainstream audience. The AI assistant is the clearest evidence yet that the company understands its custom feed advantage is only an advantage if ordinary people can use it.
That’s the right instinct. Whether the execution matches the ambition is the open question. Social media startups have a long history of promising user empowerment and delivering something less. Bluesky’s credibility rests on getting this right — building an AI tool that genuinely serves the user rather than the platform’s engagement metrics.
So far, the company has earned a degree of trust that its competitors have squandered. This next product will test whether that trust is justified or merely provisional.
WebProNews is an iEntry Publication