Introduction to the Problem In a typical centralized social network, if two users follow the same set of accounts, their experience will be very similar, filtered by the same algorithm. In a decentralized network like Nostr, this does not happen. Two observers can look at the same “social graph” and see radically different things. This is not a bug, but an intrinsic feature of the design.

1. The First Level: Fragmented Technical Infrastructure (The “Where”) The protocol is based on two components: clients (the apps) and relays (the servers). The user chooses which client and which relays to use. This choice defines the first and fundamental level of inequality.

• The Client: This is the software that decides how to display content. Different clients have different interfaces, functionalities, and, crucially, different filters (e.g., anti-spam, display rules).
• The Relay: These are autonomous servers that store and forward messages. They decide what is available. Each relay can have different data acceptance and retention policies.

Practical Consequence: If User A publishes a message only on Relay X, and Observer 1 is connected to that relay while Observer 2 is not, the message will exist for the first but not for the second, even though both formally follow User A. The network is a collection of partially overlapping “information neighborhoods.”

2. The Second Level: Decentralized Social Dynamics (The “Who” and the “How”) “Moderation” and “visibility” are functions delegated to the peripheral layers of the network, not managed by a center.

• A. Client-Level Moderation and Filters: Each client can implement custom filters. A common filter is “show only interactions from users followed by my follows.” This means that:

  • A common user’s replies to an influencer might be invisible to most of the audience, filtered out by other people’s clients.
  • The influencer’s replies, being the central node, are by default “relevant” to a much wider network.
  • Result: The common user may experience a form of “social shadow-ban”: their content exists, but the filters of other clients render it invisible to the broader public.

• B. The Weight of Reputation and Attention: The network does not treat all nodes equally.

  • A high-influence node attracts infrastructural attention. The clients of many users connect actively and frequently to the relays that node uses, so as not to miss its content.
  • A common node does not generate this “pull pressure.” Its content is received passively only by those who share the same relays.
  • Result: An influencer’s content is actively “sought out” by the network. The same content from a common user is only passively “awaited.” This creates a fundamental disparity in distribution guarantees.

3. Synthesis: Why the Experience is Unequal Inequality arises from the sum of these two layers:

Conclusion The system replaces the opacity of centralized control with the transparency (and complexity) of individual responsibility and technical choice. The user is not a passive consumer of an algorithm, but an active architect of their own information infrastructure, with all the advantages (control, censorship resistance) and burdens (complexity, fragmentation) that this entails.