Attention earned, never purchased.
Discovery, linking, and attention-direction for the mesh.eco distributed ecosystem. Fully P2P. No central server. No ads.
Traditional platforms answer this with advertising. mesh.link answers it with three questions, in order of priority.
Content with consistently positive community feedback, services with high completion rates, actors with sustained contributions. Past performance, verified by the community, is the strongest signal.
Alignment with what you are engaged with right now, tracked locally by mesh.observer. Your interest data never leaves your device. The network never learns what you are interested in.
Opinion similarity, topic neighborhoods, behavioral co-occurrence, and predicted demand identify content you have not yet encountered but are likely to find valuable.
These three questions — proven value, current interest, predicted interest — replace the single question that advertising-driven platforms ask: “who paid the most?”
Every connection in the ecosystem is one of seven link types, each arising from a different signal and carrying different weight.
Automatically detected references to known entities, topics, and actors. Trie-based name index, topic recognizer, and semantic phrase detection. Authors retain full control.
Deliberately created connections, weighted by the author's domain reputation. Links to external URLs carry trust classification from mesh.trust.
Discovered through the six-stage matrix pipeline. Topics structurally similar based on how the community evaluates them. Four weighted dimensions.
References from other actors' content, credibility-weighted. MIA performs sentiment classification to distinguish positive, neutral, or critical mentions.
Reputation-backed recommendations. Direct referrals stake domain reputation; quest-based referrals add CC compensation. Always attributed, never anonymous.
The highest-commitment link type. Stake contribution credits (earned, never purchased) to boost visibility. Symmetric risk/reward: 2x return or total forfeiture.
Personal attestation of an actor's work quality in a specific domain. Targets actors (not content), stakes credibility (not CC), non-transferable, 90-day active period.
All seven link types are combined through a ranking model that enforces the three discovery principles.
No volume advantage. No recency advantage without quality. No coordination advantage. No cross-domain manipulation. Reputation does not transfer across domains.
These walls exist to preserve the integrity of attention in the ecosystem. No configuration, governance decision, or future development can override them.
No mechanism exists or will be created for converting money into discovery ranking. Promotion requires contribution credits, earned through demonstrated contribution, never purchased. No premium tier, no enterprise exception, no “sponsored content.”
Actors can be endorsed by peers who stake their credibility. They cannot be promoted through CC-staked visibility boosts. This prevents personality cults and influencer dynamics.
Volume of content, frequency of promotion, number of referrals, and social network size provide no ranking advantage. The model rewards quality, accuracy, and relevance.
Every non-implicit link displays its type and source. No anonymous promotion, no hidden endorsement, no disguised referral. Full transparency at every level.
Detailed personal data never leaves your device. Only anonymized, aggregated signals enter the network.
mesh.link operates entirely within the P2P network. Each peer maintains local indexes that serve their own discovery needs and contribute to the collective.
Trie-based lookup of entity names, aliases, and abbreviations for implicit link detection.
Per-topic Flexsearch indexes, replicated as TopicIndex CRDT documents via Peerbit.
Local vector index of MIA-generated content embeddings for semantic similarity search.
Local cache of mat3 topic similarity from the six-stage pipeline, updated incrementally.
Enter the passphrase to unlock the content.
This mesh.link could not be resolved. It may have expired or been revoked by the author.