The Trust Problem

Every open network eventually hits that wall. Anyone can post, so eventually someone posts badly. Email solved this with centralized spam filters. Twitter solved it with employees, machine-learning models, and a willingness to ban accounts. Nostr can do neither, there’s no central server to filter, and there are no employees to enforce policy. The system was built to empower the user, but… A group of Nostr developers is boldly looking for ways to filter the content to what they believe is the best for you… The irony, we want freedom to choose what we consume, that is one of the reasons we are here and if we want viral content, entertaiment, news and celebrities? I guess in Nostr we are out of luck.

So how does a decentralized protocol stop a feed from drowning in noise?

The short answer is Web of Trust: the social graph itself becomes the filter. But the longer answer is more interesting, because Web of Trust is now a layered system spanning multiple NIPs, several reference implementations, and a set of real, unsolved trade-offs. And at the center of it all sits NIP-85.

NIP-85 represents something different. It’s not just another technical specification. It’s an admission that some problems are too big for any single client to solve alone.

What NIP-85 Actually Does

NIP-85, merged into the Nostr protocol in January 2026 through PR #1534, it defines what the community calls “Trusted Assertions”. In plain English: it lets specialized service providers perform heavy computational work, like calculating Web of Trust scores, follower counts, engagement metrics, and spam rankings, and publish signed results that any client can consume.

Web of Trust calculations require crawling many relays, processing large volumes of events, and running complex algorithms. Doing this on a mobile phone is practically impossible. Doing it consistently across thousands of clients is even harder.

NIP-85 accepts that reality. Instead of pretending every client can compute the whole network locally, it standardizes the output format, not the algorithm. Two providers can both publish a rank tag for the same public key while using completely different Web of Trust formulas, mute handling, relay coverage, or anti-spam heuristics. Clients stay interoperable because the result format matches, even when the computation doesn’t.

Nostr isn’t trying to pick one winner…

The Trust Model

One thing I really appreciate about NIP-85 is how honest it is about what it can and cannot do.

Service providers must sign their own assertion events, and the spec recommends different service keys for different algorithms or user-specific viewpoints. That keeps a provider from collapsing unrelated ranking systems into one opaque identity. A score without a source becomes platform power by another name. A score with a signed provider key lets users compare views, switch providers, and understand why two clients disagree.

Users declare which providers they trust through kind 10040 events. Those provider lists can be public in tags or encrypted when a user doesn’t want to publish their trust inputs openly. The trust stays local. Signed output proves which provider published a score, not that the score is correct. Clients need policy around which provider keys to use, which relays to fetch from, and how to handle conflicting assertions.

On Adoption

Amethyst, shipped NIP-85 Trusted Assertions for Web of Trust scoring already. Coracle has had WoT moderation for a while. Primal is integrating trust signals.

Then there’s nostr-veil, which won the top score at WoTathon 2026 across all six judging metrics. It uses ring signatures to let groups collectively score someone’s trustworthiness while keeping individual contributions anonymous. Anyone can verify the result, but nobody can tell which group members actually contributed. That’s a step forward for sensitive use cases like abuse reporting, whistleblowing, and anonymous peer review.

The Debate

Not everyone is convinced. Vertex published a direct critique titled “Why We Don’t Use NIP-85,” arguing for DVM-style approaches in its own Web of Trust work. Some argue that trusted assertions are hard to use when you don’t already know the public keys of the people you’re looking for. Processing a hundred thousand assertion events client-side is computationally expensive, especially on mobile devices.

These are real limitations. NIP-85 is better suited for verification than recommendation. If you already have a pubkey and want to know how trustworthy it is, querying pre-computed assertions works well. If you’re discovering new content from scratch, the picture gets more complicated.

The Bigger Picture

NIP-85 extends beyond people and posts. Kind 30385 lets providers score NIP-73 external identifiers, things like books, websites, hashtags, and locations. That creates a path for interoperable reputation and engagement data around subjects outside Nostr itself.

The danger, of course, is that one provider becomes invisible infrastructure and starts deciding what clients show. NIP-85 is healthiest when provider identity, algorithm scope, and user choice remain transparent. The protocol gives us the tools. It’s up to the community to use them wisely.

The Editor Two Satoshis

For Nostr Magazine position on the WoT and the Nostr purist angle, please read our article: “The Silicon Valley Lesson Nostr Refuses to Learn”, we honestly believe we need to shift to a more inclusive approach, one in which celebrities and corporations are welcome, embraced and distinctively tagged, we need to let users decide what content they want to see and we believe the solution is not complex.