The Decentralized Fourth Estate

Thanks to Eric Zhang for discussions about several topics related to media decentralization.

The advance of tech has created a fifth estate. People who control tech control the most scarce resource of our time and a very long time to come. In the realm of media and journalism, consolidation of news corporations and social media created the danger of an authoritarian future, especially when major media and social media platforms share a common view. Although the shift of Twitter’s control changed the dynamics, one fact remains true: the fourth estate is controlled by a small group of people, and there is a centralization risk of it which could potentially cause catastrophic results to the entire human society.

Creating a decentralized alternative can mitigate its negative impact and hedge the risk of centralized control over the fourth estate.

Decentralization was called to change the global monetary system, triggered by the financial crisis in 2008. People who control the global financial system control great power, which can affect everyone’s well-being if misused. Therefore, Bitcoin (and now perhaps the crypto community as a whole) is an alternative of the centralized financial system, currently creating real competitions to the established financial institutions. In the long term, if works as designed, the crypto community can not only become a safeguard of an open financial system, but also provide efficiency and new features the existing financial system could not provide.

Although a decentralized financial system can affect the power of all four estates, creating direct competitions to each single estate is necessary. In the same way, it is necessary to create a decentralized ecosystem for media and journalism to compete with the centralized alternatives.

II. Current Practices and Their Problems

A common practice that seems to address this problem, but actually not, is called “Web3 social” or “Web3 creator economy” initiatives, commonly involving tokenized governance, NFT sales, and DID systems. While these efforts tokenize ownership and allow community members to share a minority of the project’s growth are worthy experiments, they are addressing a different problem.

Web3 platforms create community ownership and community governance, kind of like Uber that gives drivers and users non-dividend voting shares so that they are more aligned in interests with the platform, or like the first version of DAOrayaki plan, creating a DAO-owned media platform so that the DAOrayaki community can collectively decide and review contents, will certainly motivate community participation, but this path will by large not lead to fully censorship resistance platforms, because more people governing something doesn’t mean it’s censorship resistant. So ultimately Web3 approaches are useful in other ways, though not able to provide a truly censorship-free media ecosystem.

III. Decentralize media content on open relays network

Base layer censorship resistance requires an interest-neutral protocol. The protocol should be as simple as possible, because its goal is not to maximally capture value for a protocol token, but rather provide a neutral layer for contents.

Media data should be stored in an open network of servers. The servers can have great diversity and should not be aligned with common interests by design.

Basically, there are two layers of neutrality - server neutrality and token economics neutrality. With these two layers of neutrality, fundamental censorship resistance can be achieved between the dynamics of relays and the media clients (platforms).

For example, we can use Nostr to build up a network of media content relays (the recently merged NIP-23 allows Nostr relays to serve long-form articles). The relays are hosted by a variety of entities and individuals, each freely stores whatever content the relay wants to store. Relays do not need to keep everything on their servers, but they have to keep real posts because all posts can be authenticated by users’ signatures.

The bidirectional selection between clients and relays naturally creates a free market of contents and opinions. The base layer should be not only neutral in technology, but also in economic preferences. A base layer with a token cannot be fully neutral because the protocol and the community will naturally prefer proposals and clients that create economic alignment (and soon ideology alignment) with the token over others that don’t. One way to avoid base layer preference is to leave incentive design to clients. Each client can offer its own incentive plan to the base layer relays, and each relay can choose its most preferred incentive. A base layer that is not bonded with a native token is preferred in the case of a decentralized media ecosystem.

A client-relay network for decentralized media contents.

IV. Add incentives to relays

Because the relay layer is token economics neutral, it is up to the clients to decide how incentives are offered. There will be a free market of client-relay incentives. Different ways of incentives can compete with each other. One client can implement multiple incentive mechanisms.

Ultimately, relay incentives are from user/client payments for desired content (for example, in SATs), whereas clients design their own incentives for relays and create business models or token economics for themselves.


Implementing subscriptions on the client and the relay network is important for DAOrayaki’s sustainability. Users can subscribe to creators, clients, topics, and relays, and the subscription fees will be distributed to the subscribed parties and relays. The distribution of subscription fees will encourage creators, curators and relays to moderate content and maintain a stable network in a sustainable way.

A client-relay incentives for decentralized media contents.

A client can subscribe to a relay (subscribing to all its contents) or specific content fed by some relays. Both will require new NIPs on Nostr. In case (1), the relay keeps a whitelist of clients, and independently implements a payment logic to whitelist clients. In case (2), the clients subscribe to some contents with unique identifiers (e.g., from some authors). The payment should be routed by some protocol to incentivize the corresponding relays and authors.

At some point, content encryption will become useful to avoid censorship and protect intellectual property rights. There are some relevant NIPs in discussion. NIP-42 authorizes viewing long-form content; NIP-38 is building encrypted group information. Beyond that, there is no NIP for encrypting long-form content and we will customize it, which should be explored after subscription is enabled.

V. Content curation with open relays

DAOrayaki’s current platform is a curation-centric client. Each curated topic can effectively serve a certain audience. When Nostr compatibility is added (shortly), content under each curated topic can be stored in a certain set of relays. The choice is bi-directional between relays and curators, and the client can remain neutral in content presentation.

Will DAOrayaki frontend shadow-ban contents? It is possible, but the entire architecture allows new clients to compete with DAOrayaki if censorship is going wild. Since contents are hosted on relays, not DAOrayaki servers. This is a basic layer of protection.

There are two ways a client could shadow-ban a curated topic.

The client frontend could lower the weight of a topic, or it can just hide the topic entirely. Therefore, in order to make an ideal client frontend, topic display (or topic suggestion) algorithms need to be entirely open source. Because all data is hosted on relays, it is relatively easy to verify that the algorithm works as it is than other social media networks like Twitter and other smaller platforms.

Another way to shadow-ban topics is through information filtering. Given that the Nostr protocol or any similar protocols will almost certainly have spam posts, spam filtering is a necessary task for relays and especially clients. However, it is obvious that spam filters can also become a censorship tool, as we well know in Twitter’s case. DAOrayaki spam filtering should be done via competitions between relays as well as competitions between curators, avoiding censorship by platform or community rules.

In addition, the current Nostr protocol lacks some basic infrastructures to enrich functionalities of decentralized media platforms. Some proposals are progressing and need to be supported and further evolve. For DAOrayaki, necessary NIPs include content curation, topic subscription, and payment support. Some existing proposals like NIP-38, NIP-42 and NIP-57 are useful resources and good places to start.