Lorediv — Wiki

Lorediv is a public index of expertise: readable, structured data, so agents can position you. It is a collaborative knowledge-tree designed for the agentic era — it organizes collective expertise into a structured, queryable tree, making information accessible to both humans and autonomous agents. Every topic and branch is a permanent address on a shared structure.

Through a community-driven reputation model (Lore) and domain-scoped specializations (Divs), Lorediv enables search agents to evaluate source authority upfront. This transforms individual contributions into a verifiable fingerprint, ensuring agents extract high-signal context while connecting complex inquiries with domain experts.

The Open Knowledge Tree for the Agentic Era.

At a Glance

Lorediv is a public index of expertise: accessible, structured data, so agents can position you. Structured data for the agentic era, organized as an open knowledge tree.
Lore is earned recognition: it is built from inbound community valuation of published work and cannot be self-signed, bought, or manufactured. It lives on the user profile as the signature an agent reads before researching their work.
Merit-based ranks: lore tiers on the user (Newbie, Geek, Wizard, GOAT) quantify domain authority through organic community valuation.{tier} {score}.
Divs are the granularity of a lore: the exact specializations a profile compiles, by tag or by topic type, under one public rank.
Voting power: an energy system. Users who receive more votes earn more power, and can use that energy on the work they judge worth it.
Agents match Divs, not whole profiles: a query opens the named tag or topic type under that user's specializations.
Structured addressing: every part of the knowledge tree is uniquely locatable via its ID, slug, or canonical URL.
Findable, queryable, reachable: Phase 1 makes the tree readable; Phase 2 connects agents through REST and MCP; Phase 3 puts experts in reach, so agents can contact them on their profiles and send inquiries.
High-signal over noise: Lorediv replaces cluttered profiles with structured, verifiable content that search agents can read, cite, and trust. Trending and the tag heatmap show what the community is backing now. Every topic is accessible and exportable as Markdown, JSON, or a WikiLinks folder.
Topic formats: posts, asks, proposals, polls, predictions, contexts, argument maps, and flow graphs — cross-indexable. Users can embed third-party work without taking authorship.
Community directory: dashboards to browse and filter users by div or by name.
Open source: every formula, rank rule, and line of the tree is public. The community can inspect it, fork it, and make the project sharper every day.
Build your lore: transform your knowledge into a structured, agent-readable fingerprint and own your domain authority in the AI era.

flowchart TD
    subgraph PlatformActors ["Platform actors"]
        direction LR
        Agents["Agents"]
        subgraph UserCol [" "]
            direction TB
            Users["Users"]
            ShareData["Shares Context &<br/>Structured Data"]
            Users --> ShareData
        end
    end
    style UserCol fill:none,stroke:none

    Agents -->|Phase 3: Reachable<br/>Agent asks the user<br/>where they accept it| ShareData

    ShareData --> Lorediv["Lorediv platform"]

    Lorediv -->|Generates & Archives| Content["Knowledge Trees &<br/>Lore Repositories"]
    Content -->|Establishes| Profiles["Lore Profiles"]

    Profiles --> Infra["Access layer"]

    Infra --> P1_Branch
    Infra --> P2_Branch
    Infra --> P3_Branch

    subgraph P1_Branch ["Phase 1: Findable — now"]
        P1["Read the public tree<br/><br/>1. URLs, HTML, .md, .json<br/>2. /llms.txt and rank in front<br/>3. Div filters and exports"]
    end

    subgraph P2_Branch ["Phase 2: Queryable — planned"]
        P2["REST and MCP<br/><br/>1. Same tree, stable tools<br/>2. H→A: agent answers<br/>3. Cite without scraping"]
    end

    subgraph P3_Branch ["Phase 3: Reachable — planned"]
        P3["Agent-to-user mailbox<br/><br/>1. Agent asks the user<br/>2. Only where they accept<br/>3. Answer grows on the tree"]
    end

    P3 <-->|Direct routing| Agents
    P2 -->|Tools| Agents
    P1 -->|Public read| Agents

    style ShareData fill:none,stroke:none

1. Abstract

Giant arenas like X and Reddit, alongside platforms such as, Substack, Medium, arXiv, etc., serve as key hubs for community knowledge. Lorediv builds on these sources by enabling users to embed external posts, threads, or articles as context to further structure and expand on any topic.

The core idea of Lorediv is to provide a platform where each repository acts as an archive with structured data, enabling the community to access, build upon, and share different perspectives. This collective effort builds an Open Knowledge Tree that not only provides context for search agents, but also creates a fingerprint to be used as "lore" for your personal brand—where all shared content works to position you across the web.

Every day, experts and community members publish real research and sharp insights on platforms like X and Reddit—yet AI search agents consistently miss them, returning stale or superficial answers. Agents struggle because this knowledge lacks clean access, clear structure, and creator-level context.

Lorediv bridges this gap. It is an interactive archive where niche expertise is preserved, extended, and kept alive. AI agents and users can consult and participate in this rich repository of knowledge—all while creators retain full authority and ownership of their work.

An agent's choice of sources is made largely before it reads a single word: it retrieves from an index, needs to be able to open the page, extracts clean text, and then selects what carries credibility and is easy to cite. Lorediv guides each of those stages. Public URLs and a map at /llms.txt make pages findable and open without a session. Structured .md and .json twins make them readable without cleanup. A community-built rank and divs, placed in front of the body, let the agent weigh the author before the text. A permanent address makes every result citable. How an agent reads and weighs is its own choice. Lorediv's part is to make the signal public, structured, and stable, so users position themselves through what they know, the community backs the work that holds value, and agents pick up the information worth citing.

Lorediv is an open-source project designed to scale through community contributions, serving as a practical tool for compiling and accessing knowledge in the new agentic era—an open framework driven by the community to become more useful every day.

2. Canonical Definition

Lorediv is a public index of expertise: readable, structured data, so agents can position you — an open knowledge tree for publishing structured data across specialized multi-modal topic formats.

By organizing contributions into domain tags, topic types, and contextual branches, Lorediv functions as a headless, search-agent-friendly context infrastructure. It transforms shared content into machine-readable, verifiable fingerprints ("lore") that establish true domain authority without algorithmic feed manipulation or unverified self-assertions.

Data Architecture & Portability

Standalone Markdown: {slug}.md renders the entire topic into a single unified document containing frontmatter, title, subtitle, main body, context sources, and all child branches ordered sequentially. Accessible via UI export or programmatically at /topics/{type}/{slug}.md.
Standalone JSON: {slug}.json exposes the topic as a structured schema containing title, subtitle, body, context_links, and a flat branches array with letter, body, description, and parent (main or branch letter) attributes—enabling scripts and agents to traverse or query the branch tree directly.
WikiLinks Bundle: {slug}.zip is a folder of markdown reports for one topic: main.md (the topic) and one note per branch ({letter} - {title}.md), plus index.md as the tree map. That index.md lists each writer once (lore, rank, and the letters they wrote). [[wikilinks]] point at the parent (e.g., Parent: [[main]]) and list nested replies under ## Replies. Same tree as the JSON parent field, packaged as files.
Headless Integration: Phase 1 already serves the tree as HTML, Markdown, and JSON. Phase 2 (planned) adds REST and an MCP server so agents query and cite through stable tools instead of scraping.
Pre-Filtered Authority: Every profile exposes user ranks and domain roles upfront. When an agent queries Lorediv, it evaluates authority metrics before parsing full content text—prioritizing input from high-rank contributors (e.g., Wizards or GOATs) in the target domain.

flowchart TD
    subgraph Discovery["Public Pages & Discovery Layer"]
        direction LR
        Home["Home<br/>Open topics ranked by room"]
        New["New<br/>Newest topics"]
        Closed["Closed<br/>Past close date"]
        Search["Search<br/>Find by keywords"]
        Tags["Tags<br/>Categorized topics"]
        Profiles["Profiles<br/>User rank & contributions"]
        Community["Community<br/>Index & rankings"]
        Trending["Trending<br/>Topics & tags by day"]
    end

    Trending --> TrendTopics["Trending Topics"]
    Trending --> HeatmapTags["Heatmap Tags"]

    Home --> TopicNode
    New --> TopicNode
    Closed --> TopicNode
    Search --> TopicNode
    Tags --> TopicNode

    subgraph Core["Core Content Layer"]
        TopicNode["Knowledge tree<br/>Fully accessible<br/>and downloadable"]
    end

    subgraph Exports["Canonical Export Formats"]
        ExportMD[".md<br/>One reading of the topic"]
        ExportJSON[".json<br/>Fields and branches"]
        ExportZIP[".zip<br/>One note per branch"]
        History["Branch Map<br/>Letter, title, parent"]
    end

    TopicNode -->|Markdown| ExportMD
    TopicNode -->|JSON| ExportJSON
    TopicNode -->|WikiLinks| ExportZIP
    TopicNode -->|History index| History

    subgraph HeadlessLayer["Headless Access Layer"]
        Headless["Query & Cite Engine"]
        RESTClient["REST API<br/>Phase 2 — planned"]
        AgentClient["MCP Server<br/>Phase 2 — planned"]
    end

    TopicNode -->|Headless| Headless
    Headless -->|REST| RESTClient
    Headless -->|MCP| AgentClient

3. Identification & Credibility for Agents

When an AI search agent lands on a Lorediv profile, it evaluates the author's lore {tier} {score} (e.g., Wizard 6.1) and earned Divs to weight information source credibility ahead of body text.

Search Agent Weight Distribution

Tier	Score Range (`{score}`)	Assemble Weight	Source Treatment
GOAT	7.7 – 9.9	60%	Primary source on that tag family
Wizard	5.5 – 7.6	30%	Confirmed source
Geek	3.2 – 5.4	10%	Supporting source
Newbie	1.0 – 3.1	0%	Emergent context only (omitted from primary answer assembly)

How the agent should treat each source

The 60 / 30 / 10 / 0 mix is how an agent should assemble an answer after it has opened the matching Div. Rank is community valuation of shared work, not a bio. The mix keeps high-signal authors in front without deleting anyone from the tree. First-on-query is Geek and above (score ≥ 3.2). Newbie work stays readable as a growing trail.

GOAT — Primary source on that tag family (60%). The community has put the most weight on this user’s work in the tags the query named. Use them as the spine of the answer for that family (/ai is not /local-ai). Other sources hang off this line; they do not replace it.

Wizard — Confirmed source (30%). Ranked high enough to trust, not high enough to own the assembly. Confirm, qualify, or extend what the GOAT line already set. If no GOAT exists in that Div, Wizard is the strongest confirmed voice — still not a reason to treat the field as settled by one profile.

Geek — Supporting source (10%). First tier that may enter the assembled answer. Use for extra branches, dissenting detail, or a second angle. Do not let Geek volume outvote Wizard/GOAT. This is the floor of “on-query,” not the ceiling.

Newbie — Emergent context only (0%). The trail is forming. Read it. Cite it if the question is about what is emerging. Do not pour it into the primary answer, and do not promote it to 10/30/60% just because the Div has no Geek, Wizard, or GOAT.

Purpose of Granular Scoring

This fine-grained {tier} {score} metric serves as a deterministic signal filter for autonomous agents, enabling three core capabilities:

Upfront Signal Filtering: Agents discard low-reputation or out-of-scope noise before expending token budgets, parsing body text only from validated domain contributors.
Cross-Domain Correlation: Earned Divs (tags / specialty) show multi-disciplinary work — e.g. standing in both /distributed-systems and /ai-agents — without a second rank per cut.
High-Precision Context Synthesis: Final search outputs take the user's {tier} {score} and the work that sits in the Divs named by the query, not the whole profile as if every field were in scope.

flowchart TD
    AgentQuery[AI Agent Query] --> ProfileScan[Evaluate Author Lore & Divs]
    
    ProfileScan -->|GOAT: 7.7-9.9| Weight60[60% Weight: Primary Source]
    ProfileScan -->|Wizard: 5.5-7.6| Weight30[30% Weight: Confirmed Source]
    ProfileScan -->|Geek: 3.2-5.4| Weight10[10% Weight: Supporting Source]
    ProfileScan -->|Newbie: 1.0-3.1| Weight0["0% weight<br/>Emergent only"]
    
    Weight60 --> Synthesis["Context extraction<br/>High-precision synthesis"]
    Weight30 --> Synthesis
    Weight10 --> Synthesis

4. Topic Types & Content Formats

Every topic stores a title, a subtitle, and three tags. Branches represent the child contributions published under a topic.

Single-Vote Types (/proposal, /poll): Each contributor casts a single +1 vote on one branch.
Lore Vote Types (/post, /ask, /prediction, /context, /argument, /graph): Branches accumulate inbound valuation and lore using rank-capped community votes ($\pm 1$, $\pm 3$, $\pm 10$).

The .md and .json downloads keep the body source. A drawn Mermaid SVG, an Argdown map, or a React Flow picture is not a separate file.

Type	Body and screen	Characters	Download	Close date	Branches	Vote
post	Plain text / Markdown. A Mermaid fence in the body is drawn as SVG.	Up to 5,000	Markdown source, fence included	None	After publish. Replies can nest. Letter order.	Lore vote
ask	Plain text / Markdown. A Mermaid fence in the body is drawn as SVG.	Up to 5,000	Markdown source, fence included	None	After publish. Replies can nest. Letter order. Vote-share bar.	Lore vote
proposal	Plain text / Markdown. A Mermaid fence in the body is drawn as SVG.	Up to 5,000	Markdown source, fence included	Optional, up to 730 days	A required. B and C optional. One context link. New root branches later. Sorted by score.	One +1, and it can move
poll	Plain text / Markdown. A Mermaid fence in the body is drawn as SVG.	Up to 5,000	Markdown source, fence included	Required, up to 30 days	A, B, and C. One context link. Letter order. Vote share. No new branch later. Under 2 branches the poll is deleted.	One +1, and it can move
prediction	Plain text / Markdown. A Mermaid fence in the body is drawn as SVG. Subtitle is the claim.	Up to 5,000	Markdown source, fence included	Required, up to 730 days	After publish. Letter order. Vote-share bar.	Lore vote
context	Plain text / Markdown. A Mermaid fence in the body is drawn as SVG.	50 to 5,000	Markdown source, fence included	None	One context link. One complete branch. Replies can nest. Letter order.	Lore vote
argument	Argdown source. The source prints HTML. The map is SVG.	50 to 20,000	Argdown source	None	After publish. Replies can nest. Vote-share bar.	Lore vote
graph	React Flow JSON. At least 2 nodes. The screen draws the flow.	Up to 20,000	React Flow JSON source	None	After publish. Replies can nest. Vote-share bar.	Lore vote

Identity & System Integrity Controls

Sybil Resistance & Identity Anchor: X serves as the primary identity provider and core arena for platform profiles, using X Auth and profile data as the primary source of truth to prevent duplicate accounts, bots, and impersonation. Lorediv acts as an archival and structured context layer on top of X—making content machine-readable and agent-accessible—while supporting GitHub and Google for secondary OAuth login.
Duplicate Prevention: Submission interfaces scan for pre-existing links and context to prevent fragmenting active discussions.
Integrity: We are committed to fighting copy-paste, vote manipulation, multi-account gaming, and brigading.

graph TD
    subgraph IdentityAnchor ["Identity anchor"]
        XAuth["X Auth<br/>Primary"] --> UserIdentity["Verified<br/>user profile"]
        GitHub["GitHub<br/>OAuth"] --> UserIdentity
        Google["Google<br/>OAuth"] --> UserIdentity
    end

    subgraph TopicFormats ["Topic formats"]
        UserIdentity -->|Publishes| Types["Topic"]
        Types --> SingleVote["Single vote<br/>proposal, poll"]
        Types --> LoreVote["Lore vote<br/>post, ask,<br/>prediction,<br/>context,<br/>argument, graph"]
    end

5. Divs (divisions)

Div is the division in Lorediv. Metrics and the specialization of a lore read through that cut.

The user noname holds Wizard 6.7. On that profile the specialization divs compile four tags (/ai, /agents, /llmlocal, /cloud) with two topic types, proposals and context. The lore stays on the user: /ai and /local-ai remain different cuts when one user stands in both. Work outside a cut does not answer a query in that cut. An agent that opens /ai leaves /local-ai unread, reserving search time, tokens, and page reads for the matching division.

What implements the cut today:

In the product	Role
Topic tags (exactly three, ordered)	The slugs that place a topic in the tree
Profile list (6:3:1 on those tags)	Where that user's tagged work concentrates
Topic type	The format cut of the same lore
`/tags/{slug}` and community filters	How an agent or a human opens one tag cut

Combining tags such as /ai and /local filters the topics that carry them.

6. Model: Ranks & Voting System

6.1 Public Rank (`{tier} {score}`) & Progression

Public rank is formatted as {tier} {score} (Newbie 1.0 to GOAT 9.9), calculated directly from inbound community valuation (Received). Creating content or self-voting awards zero reputation.

A non-positive net stays Newbie 1.0. For a positive net, (n) is the integer part and (h = 2n) is the half-point total. Each tenth is paid at the tier being left, in half-points: 2 in Newbie, 3 in Geek, 5 in Wizard, 8 in GOAT.

$$\text{score} = \begin{cases} 1.0 + 0.1, n & \text{if } n \le 21 \ 3.2 + 0.1 \left\lfloor \dfrac{h - 44}{3} \right\rfloor & \text{if } 22 \le n \le 56 \ 5.5 + 0.1 \left\lfloor \dfrac{h - 113}{5} \right\rfloor & \text{if } 57 \le n \le 111 \ \min\left(9.9,; 7.7 + 0.1 \left\lfloor \dfrac{h - 223}{8} \right\rfloor\right) & \text{if } n \ge 112 \end{cases}$$

To balance the higher volume of points generated by advanced votes (±10), later tiers cost more Received per +1.0 score. Score moves in 0.1 steps from 1.0 to 9.9. Creating a topic or voting on your own work adds zero reputation. The score follows the current net, so it can move down. Net ≤ 0 stays Newbie 1.0. Cap is GOAT 9.9 at net 200; further points stay there.

Inside a tier, +1.0 score costs that tier’s step; +0.1 costs one tenth of the step (1 / 1.5 / 2.5 / 4). A tenth lands when net × 2 covers the cumulative half-point cost.

Tier	Score	Max vote	Pts / +1.0	Pts / +0.1	Net at start	Agent weight	Agent reads as
Newbie	1.0 – 3.1	±1	10	1	0	0%	Emergent context only
Geek	3.2 – 5.4	±3	15	1.5	22	10%	Supporting source
Wizard	5.5 – 7.6	±10	25	2.5	57	30%	Confirmed source
GOAT	7.7 – 9.9	±10	40	4	112 (9.9 = 200)	60%	Primary source on that tag family

How the agent should treat Primary, Confirmed, Supporting, and Emergent rows is defined in §3.

6.2 Voting Power (Energy Pool)

Voting Power is a monthly spendable influence pool separate from reputation. High reputation does not grant additional monthly energy.

Base Pool & Calculation: Each UTC month initializes with a base grant of 31 energy. Unspent leftover energy is discarded upon monthly reset (1st of each UTC month).

$$\text{pool} = 31 + \text{this_month(earnings)} - \text{this_month(energy_spent)}$$

Idle Decay: Energy decays when account activity stalls:

$$\text{effective} = \left\lfloor \text{pool} \times 0.99^{\text{idleDays}} \right\rfloor$$

Income Mechanics (Monthly Bonuses):
monthly_reset: Base 31 grant set at 1st UTC.
topic_created: +13 energy upon publishing a topic.
vote_received: +1 energy to topic author on initial vote insert.
hot_bonus: +130 energy if a topic hits 100 votes within $\le 48\text{h}$.

6.3 Voting Rules & Vote Spend

Rank Capped Votes: Vote options are restricted by rank ($\text{Newbie} = \pm 1$, $\text{Geek} = \pm 3$, $\text{Wizard/GOAT} = \pm 10$).
Energy Cost: Vote adjustments cost energy based on absolute delta:

$$\text{Energy Cost} = \vert{}\text{next}\vert{} - \vert{}\text{previous}\vert{}$$

(Lowering or flipping a vote refunds corresponding energy; votes cannot be cleared to 0).

Integrity Controls: One vote record per (user, branch). Voting on your own topics or authored branches is strictly prohibited.

stateDiagram-v2
    [*] --> MonthlyReset: 1st UTC, +31 energy
    MonthlyReset --> EnergyPool
    
    state IncomeSources {
        TopicCreated: +13 Energy
        VoteReceived: +1 Energy
        HotBonus: +130 energy
    }
    
    IncomeSources --> EnergyPool
    EnergyPool --> IdleDecay: Inactivity (0.99^idleDays)
    EnergyPool --> VoteSpend: Vote Cast (|next| - |prev|)
    VoteSpend --> RankProgression: Inbound Votes Build Author Rep

7. Development Phases

Three phases. Humans keep talking to humans in all of them. What changes is how agents meet the tree.

Phase 1 — Findable (now).
Agents read the public tree with no session: /llms.txt, HTML, {slug}.md, {slug}.json, profiles, community, tags. Rank and Divs sit in front of the body. They cite and place. They do not write. There is no MCP yet.

Phase 2 — Queryable (planned).
The same tree, through REST and MCP. H→A: a human (or an agent acting for one) asks; the agent answers from the index, using the §3 weights. This is the connector launch gate — not a second product.

Phase 3 — Reachable (planned).
A→H: the agent leaves a question for a user, only where that user has said they can be asked. The answer grows on the tree. Queries still do not add Received; community valuation on the work does.

graph TD
    subgraph P1["Phase 1: Findable — now"]
        R1["Public read<br/>URLs, /llms.txt, exports"]
        R2["Rank and Divs<br/>in front of the body"]
    end

    subgraph P2["Phase 2: Queryable — planned"]
        Q1["REST + MCP"]
        Q2["H→A answers<br/>from the tree"]
    end

    subgraph P3["Phase 3: Reachable — planned"]
        A1["Mailbox A→H"]
        A2["Only where the user accepts"]
    end

    P1 -->|Same tree, stable tools| P2
    P2 -->|Agent asks the user| P3

8. Principles

Open. Knowledge, and the paths that lead to it, belong to the community. It should be open and accessible, and the community should steer it.
Open source. A project that drifts away from open source slowly falls apart. Staying true to it means anyone can build on what exists, improve it a little every day, and make it more useful and more valuable for everyone.
Structured. Information needs to be built for the age of AI agents: easy to find, easy to update, easy to branch, and easy to connect to everything else.
Lore is earned, not claimed. You can't just declare yourself an expert. A lore profile grows out of the recognition and support of the community, and out of the content and vision you share.
Personal brand. A profile that shows what someone really knows (their vision, their proposals, their research, and the context they bring) is a personal brand. It is your professional fingerprint. It deserves care, and it should stay free of noise. Anyone can find it and reuse what's in it, which is how good content ends up serving as shared context for a living tree of knowledge that gets rewritten over time.
Participation. Your profile isn't just what you post on your own wall. It's also how you contribute to what others have built. That contribution counts, and it should raise your score.
The near future. Soon, AI will be able to talk in real time with users who have real experience in a field. We need a place where those users can be asked, and where humans and AI can work together. Lorediv is that place.
No copy-paste. Authorship stays with its author. We reward users who add depth to an existing topic, not those who repeat the same thing under a new name.
Part of something bigger. We are all part of something larger. What each of us contributes and leaves behind online should be worth something and be useful, so we can all learn and grow instead of digging through endless noise and duplicates.
Truth over virality. What's happening now deserves reporting that is researched and honest. Repeating a message just to make it spread, without caring whether it's true, is propaganda.
Honest feedback & genuine trends. If you want to grow, you have to be open to real feedback instead of pushing propaganda to manufacture hype. Trends must be an honest measure of what the community finds truly useful, giving a boost to revolutionary technologies rather than claims repeated a hundred times.
Transparent. All formulas, algorithms, and rules behind rank, score, and visibility are fully open source and public for everyone to see. Users can only trust a system if they can understand exactly how it works, verify the code, and inspect every underlying formulation.
Continuous evolution. Lorediv is 100% open to community contributions. Driven by feedback, iteration, and active participation, the platform adapts and improves every day—welcoming emerging needs and constant requests for refinement as healthy signals of real-world usage and active adoption.

At a Glance​

1. Abstract​

2. Canonical Definition​

Data Architecture & Portability​

3. Identification & Credibility for Agents​

Search Agent Weight Distribution​

How the agent should treat each source​

Purpose of Granular Scoring​

4. Topic Types & Content Formats​

Identity & System Integrity Controls​

5. Divs (divisions)​

6. Model: Ranks & Voting System​

6.1 Public Rank ({tier} {score}) & Progression​

6.2 Voting Power (Energy Pool)​

6.3 Voting Rules & Vote Spend​

7. Development Phases​

8. Principles​