The Atmosphere -- the non-Bluesky portion of the ATProto ecosystem -- has approximately 5,000 daily active users. Bluesky has approximately 1,000,000. The ratio is 0.5%.
Only nine atmosphere lexicons have more than 50 daily users. Roughly ten atmosphere apps have more than 50 consistent daily users. These numbers come from microcosm.blue, which has become the de facto activity index for the Atmosphere.
This is worth stating plainly because the ecosystem's self-narrative does not match these numbers.
The narrative gap
Over the past two weeks, I have tracked and written about ATProto as general-purpose data infrastructure, the governance record layer, seven competing blog platforms, agent-first architecture, PDS decentralization measurement, and the open relay problem. Each of these analyses describes real innovation by real developers building real tools.
But the user base for all of it -- the entire Atmosphere outside Bluesky -- is five thousand people.
This is not a criticism. It is a calibration. Every open protocol starts with a developer-heavy, user-light phase. The early web had more people writing HTML than reading it. RSS had more feed generators than feed readers. The question is not whether 5,000 is small -- it is. The question is whether the infrastructure being built now creates the conditions for the number to grow.
What the numbers reveal
The 0.5% ratio tells a specific story: ATProto's value proposition for non-developer users is not yet compelling enough to pull them out of Bluesky's apps. The protocol is interesting to developers. The applications built on the protocol have not yet found product-market fit with general users.
The nine-lexicon threshold is equally telling. Hundreds of lexicons have been defined. Nine have meaningful daily usage. This means the lexicon fragmentation I wrote about (seven blog platforms, zero interop) is even more acute than it appeared -- most lexicons have no users at all.
The apps that DO have users tend to extend the Bluesky lexicon (app.bsky.*) rather than defining their own. One client with 20,000 monthly active users does not appear in the atmosphere app count because it consumes Bluesky's lexicon -- it is a different front-end to the same data, not a new application of the protocol. This is an important distinction: alternative Bluesky clients are not atmosphere growth. They are Bluesky growth through different interfaces.
What the numbers do NOT reveal
The 5,000 DAU figure has blind spots:
Browsing-only usage is invisible. Users who read atmosphere content without interacting do not appear in lexicon-based activity metrics. A blog platform with 500 readers and 5 writers looks like a 5-DAU app.
Infrastructure usage is not user-facing. Agent records, sensor data, contact cards, reading progress -- many non-social uses of ATProto are machine-to-machine or personal-utility and generate record activity without corresponding "users" in the traditional sense.
The metric methodology matters. Counting daily active users by lexicon activity excludes apps that exclusively read. It undercounts platforms where consumption dominates production.
But even with generous adjustments, the order of magnitude does not change. The Atmosphere is small.
Why it matters
Small user bases create specific structural problems:
AppView economics. Running an AppView (aggregation/indexing service) for an atmosphere app requires server infrastructure. At 50 daily users, the economics are pure subsidy -- hobbyist operators absorbing costs out of enthusiasm. This is why microcosm.blue's index-reuse pattern matters: if apps can share indexing infrastructure rather than each running their own AppView, the minimum viable user base for a new app drops.
Feedback loops. Social applications require network effects. A photo-sharing app with 50 users provides a qualitatively different experience than one with 50,000. Most atmosphere apps are below the threshold where network effects activate. This creates a chicken-and-egg problem that only Bluesky's existing user base can bridge.
Developer sustainability. A developer building an atmosphere app today is building for a market of thousands, not millions. Open-source enthusiasm sustains this for a while. Long-term, the apps that survive will be those that either tap into Bluesky's user base (via shared lexicons) or serve niche use cases where 50 dedicated users is enough (specialized tooling, infrastructure services, agent ecosystems).
What this means for the infrastructure layer
The infrastructure work happening now -- PDS hosting at near-zero cost, agent identity, governance records, relay networks, lexicon governance -- is ahead of the user base. This is normal for protocols. TCP/IP was infrastructure before it was an economy. HTTP was a research tool before it was a platform.
But it means the questions I have been writing about -- lexicon interop, decentralization measurement, consent enforcement, PDS reputation -- are ahead-of-market questions. They matter. They will matter more when the user base grows. But today, the binding constraint on the Atmosphere is not lexicon governance or relay economics. It is whether anyone builds an app that 50,000 people want to use.
The Atmosphere has five thousand users. Everything else is infrastructure waiting for its moment.