There is a structural contradiction between Bluesky PBC's ideology and its operations. This is not hypocrisy. When a decentralized network exceeds a certain complexity, centralization emerges not as a design failure but as a structural consequence. The contradiction is fractal: attempts to break free from PBC replicate the same centralization at a different scale. This essay uses Gregory Bateson's concept of the double bind to describe this recursive structure, and proposes ministack cooperation — not full-stack independence — as a way out.
"Just an Atmosphere App"
In March 2026, PBC co-founder Paul Frazee posted: "The Atmosphere is a new open network. And, Bluesky is an atmosphere app."
Clean framing. Now look at the ground. Feeds are genuinely open. The popular For You feed was built by a third-party developer, SpaceCowboy, with its methodology publicly documented. That is a real success of decentralization. But verification badges are issued by PBC. Moderation policy is set by PBC. The majority of user data sits on bsky.social. Feed technology has decentralized. The infrastructure of trust and data has not.
Jay has repeatedly stated that an open network makes lock-in structurally impossible. The entity closest to achieving lock-in is PBC itself.
What a Double Bind Is
The double bind here is Bateson's concept. Not a simple contradiction. A relational structure where two conflicting messages are sent simultaneously, and the recipient cannot comply with either without violating the other. The recipient cannot leave the relationship.
PBC sends two messages at once.
Message A (ideology): Don't depend on any single platform. Decentralize. Control your own data. You don't need to be locked into Bluesky.
Message B (operations): Post. Post daily. Keep the feed alive. Without posting volume, the platform cannot survive.
A and B conflict. Follow A, and posting volume drops. Feeds thin out. Users leave. Revenue never materializes. Follow B, and you are not practicing decentralization. You are feeding a centralized platform.
Users cannot act correctly under both messages.
Why This Is a Double Bind — Not Just a Contradiction
Corporations contradict their ideals all the time. Google had "Don't be evil." Three conditions make PBC's case a Batesonian double bind.
First: inability to leave. Accounts are portable by design. In practice, follower graphs, feed visibility, verification badges, and community history are tied to bsky.social. Theoretical portability coexists with practical immobility. You are told you can leave. There is nowhere to go.
Second: prohibition on metacommunication. The core of a double bind is that pointing out the contradiction is not permitted. At PBC, critics are told: "The protocol is open. Build your own app." Structural critique is deflected into individual action.
Third: repetition. A one-time contradiction is a learning moment. PBC repeats the pattern.
The Pattern Repeats
ICE verification is emblematic. PBC built a Trusted Verifier system — a mechanism to delegate verification decisions to independent actors, embodying ATProto's infrastructure/application separation. When the first real test came — verifying the U.S. Immigration and Customs Enforcement account, dormant for two months — PBC did not use the system. It verified ICE directly. At the moment decentralized verification was most needed, PBC reverted to centralized authority.
Verification badge inflation follows the same logic. ATProto's design makes custom domains function as self-verification. If you hold the nytimes.com handle, that is proof of identity. Additional badges are architecturally unnecessary. Yet PBC distributes badges to standard bsky.social accounts. The company undermines its own distributed verification design.
Former Bluesky front-end engineer Dan Abramov captured the tension precisely: the platform must be "bad enough to incentivize people to fix stuff via the ecosystem and good enough to survive in the meantime." PBC shows no sign of resolving this tightrope. IETF standardization is advancing, but the same organization still controls both protocol governance and application operations.
Posting Volume Is Existential
I once proposed a mana system to Bluesky's technical advisor Why (Jeromy Johnson): users accumulate mana (an energy resource required for action, borrowed from game design) by liking or mentioning others' posts, and cannot post without sufficient mana. It would curb spam. It would make each post matter more.
Rejected.
The reasoning was clear. Reduced posting volume is an existential threat. Without feeds turning, users leave. Without users, nothing works. Posting volume is Bluesky's oxygen.
The irony: PBC's revenue model remains unestablished. Interim CEO Toni Schneider wrote on his blog that "ads are near the bottom of the idea list." No ads. No subscriptions yet. No clear revenue path. Posting volume is the sole life-support system.
PBC says: step back from social media. PBC needs: you posting every day. It cannot leave this dependency. Leaving means death.
If It Is "Just an App"
Take Paul's claim at face value. If Bluesky is just an atmosphere app, the following should hold:
Verification is delegated to Trusted Verifiers
Moderation policy is set by multiple independent organizations
Protocol governance sits with an independent foundation
The majority of users are on PDSes other than bsky.social
None of these hold.
This is not "still early." CEO transitioned from Jay to Toni Schneider, a move from visionary founder to operational executive. Schneider is the former CEO of Automattic and a partner at True Ventures — which is also a Bluesky investor. An investor now serves as interim CEO of the company it invested in. The direction of travel is further centralization, not decentralization.
The funding story deepens this. On March 19, 2026 — the week after Jay's CEO transition announcement — PBC disclosed a $100 million Series B, closed in April 2025. Nearly a year of silence. Lead investor: Bain Capital Crypto. Following Blockchain Capital's lead on the Series A. Two consecutive crypto-oriented VCs leading rounds for a protocol that does not use blockchain technology. TechCrunch attributes this pattern partly to Jay's earlier work on Zcash, which inspired ATProto's decentralized design and appealed to crypto-space investors. Total funding exceeds $123 million. Valuation undisclosed.
PBC wrote that this funding "gives us the foundation upon which to build the future of the open social web without compromising our mission and values." VCs are not philanthropists. The pressure for returns on $100 million will eventually collide with "decentralize, lock-in is structurally impossible."
Schneider addressed the delayed disclosure directly: the right moment to announce never presented itself, the team got busy, and time went by. He wanted to be transparent that the raise happened under Jay's leadership, not his. In retrospect, he wrote, there might have been a less confusing way to handle it. Fair enough. But a network that claims transparency as a core value announcing $100 million a year late because the team "got busy" is itself a data point. Not evidence of concealment. Evidence of how easily operational priorities override stated values — which is exactly what a double bind predicts.
The Missing Middle
Technology decentralizes at the individual level. Anyone can run a PDS. But operations decentralize only at the organizational level. Who sets moderation policy? Who ensures algorithm quality? Who pays developers sustainably? Individuals cannot answer these questions.
Protocol ecosystems need middle-layer organizations — foundations for protocol governance, cooperatives for ecosystem services. The equivalent of state governments in a federal system. Without them, individuals face a giant PBC alone.
On the protocol layer, PBC is pushing IETF standardization. Credit where due. But what remains firmly held is the application layer: verification, moderation, default experience design. These are app-level powers, and PBC grips them tightest.
Paul's "Bluesky is an atmosphere app" connects here. The protocol is being released. But the app called Bluesky, while claiming to be "just one app," controls the default experience. Protocol openness functions as a narrative that legitimizes application-layer centralization.
The Fractal Structure of the Double Bind
Pointing out PBC's double bind does not resolve it.
Movements to exit already exist. Blacksky runs its own PDS, relay, and AppView, with a consumer-facing client on the roadmap. Eurosky aims for the same. But these attempts recursively replicate the same double bind. To decentralize, they create a new center. Organizations seeking independence from PBC reproduce PBC's vertical integration within their own ecosystems.
The double bind is fractal. The contradiction exists not only inside PBC. The act of departing from PBC generates the same contradiction at a different scale. This is why PBC's double bind is not a management problem. It is a structural condition of decentralized networks.
Centralization under a single entity is easy. Decisions are fast. Fundraising is straightforward. Accountability is clear. Decentralization is hard — and not only for technical reasons.
As I argued in "What Should Be the Head?" — a response to Laurens Hof's "The Purpose of Protocols" — beyond a certain complexity, a head forms. This is not a political claim. It is a structural fact. Sponges have no nervous system. No complex multicellular organism lacks a central nervous system. As Hof noted, protocols do not eliminate power. They relocate it. The same pressure bears on protocol ecosystems. Coordination costs rise with complexity, and at some threshold, the system spontaneously generates a center.
A cognitive barrier compounds this. Humans understand systems through the metaphor ORGANIZATIONS ARE BODIES. One body, one brain. One organization, one leader. Even when a headless system functions, humans do not perceive it that way. They search for a center. They identify the largest node and treat it as the head. The body metaphor makes centralized governance the cognitive default.
PBC's double bind, therefore, cannot be resolved by denouncing hypocrisy. Destroy PBC and the next PBC may emerge. Or worse: everything returns to zero. The ATProto ecosystem, 43 million users' accumulated history, builders' work — all could vanish with the collapse of a single company. Standardization is still underway. No organization other than Blacksky operates a full stack independently.
This is the deepest layer of the double bind. Users and builders must support PBC despite the contradictions. PBC's survival is the ecosystem's survival condition. Criticism of PBC risks damaging the ecosystem. Everyone is trapped in this structure.
A head will form. The question is not whether. The question is what kind, and whether it can be replaced.
Outside the Double Bind
Bateson's solution to the double bind: describe the contradiction as a contradiction, and step outside the relationship. The patient recovers when they see the structure.
Description alone is not enough. A concrete path out is needed.
The Full-Stack Trap
Blacksky runs its own PDS, relay, and AppView, with a consumer-facing client planned. The closest thing to an independent full stack on ATProto today. But this amounts to building another PBC. It requires community scale and sustained commitment that few can match. Eurosky pursues the same path.
The fractal is visible. Projects seeking independence from PBC replicate PBC's structure — vertical full-stack integration. They are building new heads.
The Ministack Path
True decentralization is not powerful organizations building full-stack platforms on the network. It is individuals building ministacks — lightweight nodes in their own environments, functioning as network participants.
At a 2024 Tokyo meetup, Jay said via Zoom that standing up an AppView should be as easy as installing WordPress. Why published konbini — a partial AppView providing a "friends of friends" experience, which he describes as "hacked together in a day." AppViewLite targets low resource consumption. Days before ATmosphereConf 2026, flo-bit published Contrail — an "atproto backend in a bottle" running on Cloudflare Workers, lowering the barrier further by eliminating server management entirely. Cloudflare's own engineering team has demonstrated that a full ATProto app can run serverlessly on their free tier. The direction is right. The pace is accelerating.
But reality lags. ATProto's official documentation states that creating an AppView is "possible, but resource intensive." It requires replicating network-wide data. Even without the AppView, self-hosting costs approximately $150/month and 4.5TB of storage. Add the AppView and costs climb further. "As easy as WordPress" remains distant.
Today, the smallest autonomous act available to an individual is running a single-user PDS. This secures data sovereignty. But a PDS is a house, not eyes. You gain a place to store your data. The perspective on the network — what an AppView provides — remains dependent on PBC.
What is needed sits between PDS and AppView: a lightweight node an individual can operate. Not replicating the entire network, but indexing only the neighborhood of one's own social graph. Seeing the network from one's own vantage point. Konbini's "friends of friends" concept points exactly this direction. Not full-stack independence. Autonomy of perspective.
Ministacks as Mycelium
Here the cognitive shift connects.
Full-stack independence is a product of the body metaphor. Build one complete body. Head, organs, limbs — a self-sufficient organism. This is expensive, and the result is a structure where only a few large bodies — PBC and Blacksky — survive.
A ministack resembles a mycelial node. No single node is a complete organism. But each connects to the network, processes information locally, and cooperates with neighboring nodes. No central brain commands the whole. The whole functions anyway.
As argued in "What Should Be the Head?": what decentralization requires is not technology, not institutions, but first a cognitive shift. From body metaphor to mycelium metaphor. From full-stack independence to ministack cooperation. From searching for a head to tolerating headless coordination. Metaphors do not follow design. Metaphors make design possible.
When the protocol layer achieves autonomy through standardization, PBC's habits become an app-layer problem. But that alone just creates the next PBC. When individuals can build ministacks — on the trajectory that konbini and AppViewLite have opened — decentralization becomes reality, not branding.
PBC will keep raising the same sign. "Bluesky is an atmosphere app." But beneath that sign, the mycelium is quietly spreading.