System Blueprint for a Link-First Knowledge Network: Architecting the Evolution of CognitiveLiberties.com

Full report

On this page

The architecture of digital discovery fundamentally dictates the boundaries of human inquiry. When search engines and centralized algorithms intermediate the acquisition of knowledge, they inherently apply invisible judgments, behavioral tracking, and algorithmic ranking that compromise mental privacy and cognitive autonomy. CognitiveLiberties.com currently operates as an essential educational redoubt, advocating for freedom of thought, mental privacy, and resistance to centralized informational control1. The platform successfully aggregates vital research, operating with zero trackers, zero adtech, and zero third-party runtime requests1. However, to fulfill its mandate of enabling users to traverse complex concepts autonomously without relying on commercial search engines, the platform must evolve from a hierarchical educational repository into a human-curated, link-first knowledge network. The ensuing analysis provides a comprehensive architectural blueprint for this transformation. By integrating semantic ontologies, client-side search functionality, decentralized web principles, and rigorous editorial controls, the system will allow visitors to begin with a single question, navigate through a web of independent sources, understand the exact contextual relationship between connected ideas, and continue learning continuously. This blueprint details the transition toward a decentralized discovery hub prioritizing visitor autonomy, evidence provenance, and original-source discovery over algorithmic time-on-site optimization.

Documented Current-State Information Architecture

An exhaustive inspection of the existing CognitiveLiberties.com infrastructure reveals a traditional hierarchical content architecture optimized for direct reading rather than lateral discovery. The site operates under strict privacy directives, enforcing a zero-tracker environment and maintaining a dedicated First-Party Self-Audit verification system alongside its foundational charter1. The current platform houses a substantial repository of proprietary and aggregated research, notably featuring seventy-four exact research reports preserved in protected custody and five complete implementation kits designed for public institutions, libraries, journalism, public-interest technology, and education1. The content is logically organized into major thematic domains, including the worldwide pattern of censorship, the defense stack for cognitive liberty, and an extensive AI governance research series comprising sixteen reports1. The site articulates a clear threat matrix divided into mechanisms of identification, control, and preservation, arguing that mandatory identity makes inquiry traceable while intermediary liability privatizes censorship1. However, a structural audit identifies significant friction points that impede lateral discovery. The system currently describes a highly integrated organizational network, a problem-based Resource Navigator, an evidence ledger, and standards directories1. While these concepts are heavily promoted in the site's manifesto and primary navigation pathways, the actual implementation of these network nodes remains dangerously incomplete. The audit confirms that critical discovery pathways function as dead ends. Specifically, the Privacy Standards Directory, the Verified Organizations Directory, the Resource Navigator, and the Evidence Ledger are currently inaccessible, returning resolution failures rather than functional knowledge hubs2. The current architecture relies heavily on linear reading pathways. A visitor attempting to understand the relationship between artificial intelligence regulation and censorship risks1 must read a self-contained report. If the visitor wishes to trace the foundational evidence—such as the V-Dem Democracy Report 2026, which notes that the global level of democracy has fallen back to 1978 levels and that forty-one percent of the world's population now lives in autocratizing countries1—they are forced to exit the site ecosystem, query a commercial search engine, and risk algorithmic profiling. The current-state architecture can be summarized through the structural mapping of active and inactive domains, highlighting the dissonance between the platform's stated goals and its technical reality.

The transition to a link-first knowledge network requires transforming these isolated hierarchical pillars and broken directory endpoints into a cohesive semantic graph. The platform must move from telling the user what to think about cognitive liberty to providing the user with an autonomous vehicle to explore the evidence themselves.

Comparative Analysis of Discovery Experiences

To design a system that maximizes visitor autonomy, it is necessary to contrast the current implementation with established and experimental knowledge-navigation paradigms. The goal is to build an environment where users understand precisely why a link is recommended and where that link will take them before they commit to clicking, thereby eliminating the reliance on algorithmically personalized search feeds. Wikipedia represents the most ubiquitous model of lateral web discovery. Its architecture relies on flat wikilinks embedded directly into the prose. While highly effective for encyclopedia traversal, Wikipedia's discovery experience suffers from a fundamental ontological deficit. A standard wikilink possesses no intrinsic semantic weight; the user does not know whether the linked article supports, refutes, or merely tangentially relates to the current topic until they read the surrounding context. Furthermore, Wikipedia's strict neutrality policies prevent it from acting as an editorialized discovery hub that actively champions specific human rights frameworks like cognitive liberty. Library subject guides and university research guides offer a higher degree of human curation. These platforms typically arrange resources by topic, format, and relevance. However, they frequently devolve into static lists of outbound links. They operate as flat directories rather than interconnected graphs. A visitor cannot easily trace a specific statistical claim found in a PDF document back to the specific organization that funded the research, nor can they view the algorithmic safety standard that addresses the issue, because the relationships are not programmatically defined. Curated web directories targeting the non-commercial web provide a more philosophically aligned model. Marginalia Search, an independent Swedish crawler, actively downranks advertisements and commercial scripts to surface text-heavy, non-commercial content6. Marginalia operates its own index and crawl infrastructure, completely divorcing itself from the Google and Bing ecosystems, which frequently resell indexed results to ostensibly "private" search engines8. This independent index model proves that high-quality discovery can occur outside commercial monopolies. However, Marginalia is inherently a search engine, relying on user-generated keyword queries. Similarly, curated human directories like Rami.party organize open-source and non-commercial tools using explicit tagging structures for security, networking, and GDPR compliance6. This explicit categorization allows users to filter resources accurately but still lacks the rich narrative context required for complex socio-technical concepts like mental privacy. CognitiveLiberties.com must function as a curated discovery hub that anticipates and structures the learning journey, rather than a blank search bar awaiting a query. The most sophisticated model for the proposed evolution is found in independent knowledge-navigation projects like Gwern.net. This architecture standardizes linking and citation behaviors using deep anchors, explicit metadata, and client-side transclusion10. Gwern.net employs a link-icon CSS system and typographic subscripts to provide evidential weight to citations directly in the prose, indicating the year and source format before a click occurs12. More importantly, the system utilizes rich popup annotations. When a user hovers over a link, an annotation automatically expands, allowing the user to read abstracts, summaries, or metadata without losing their place in the primary document10. The site also utilizes a bidirectional backlink architecture, allowing users to see everywhere a specific concept or external URL is discussed across the entire domain, effectively transforming a static page into a fluid node14. The proposed architecture for CognitiveLiberties.com will synthesize the independent ethos of Marginalia, the rigorous topical curation of library guides, and the advanced hypertext typography and transclusion mechanics of Gwern.net.

Proposed Link-First Information Architecture and User Journeys

The proposed information architecture abandons the traditional homepage-to-article hierarchy in favor of a semantic node-and-edge graph. In this system, every concept, organization, standard, and piece of evidence is treated as an equal node in the network. The value of the network lies in the explicitly defined edges linking these nodes. Instead of hiding the dead-end directories behind a dropdown menu, the currently broken Resource Navigator becomes the central nervous system of the site4. The architecture shifts from narrative reports with static bibliographies to interconnected conceptual hubs. The primary structural components of the new architecture include Core Concepts, the Evidence Ledger, the Implementation Ecosystem, and the External Threat Matrix. These are woven together by contextual features designed to facilitate deep traversal.

To illustrate the efficacy of this link-first architecture, consider the following theoretical user journey through the newly designed graph. A visitor arrives at the conceptual node detailing "Internet Shutdowns and Cognitive Control." The text explains that authoritarian regimes utilize infrastructural control to sever access to independent thought. Embedded within this text is a superscript citation utilizing Gwern-style link icons and typography13. Hovering over this citation triggers a client-side transclusion popup detailing the Access Now \#KeepItOn 2025 findings, noting that 313 internet shutdowns occurred across 52 countries in 2025 alone, with Myanmar standing as the leading offender1. Within this transcluded popup, the user selects the "Follow the Evidence" contextual pathway. This directs them to the internal Evidence Ledger node for the "\#KeepItOn Dataset." Here, the user sees a structured breakdown of the methodology, noting the rising targeting of Low Earth Orbit (LEO) satellite internet systems by regimes in Chad, Iran, and Syria1. Alongside this data sits a "Related Organizations" link pointing to the Access Now organization profile. Navigating to the Access Now profile, the user views a bidirectional backlink indicating that this organization is also intrinsically linked to defending against "Transport Layer Protocol (TLS) Interference." Clicking this takes the user to a technical standard node discussing encryption protocols. Within this technical node, the "Continue Exploring Elsewhere" feature provides an outbound, explicitly typed link to the OONI global measurement dataset, where the user can view the real-time application of the OONI Pipeline v5 processing three billion network interference measurements1. Throughout this entire journey, the user has seamlessly traversed geopolitical theory, human rights datasets, organizational profiles, and cryptographic standards without ever utilizing Google, Bing, or an algorithmically personalized feed.

Topic Knowledge Graph and Semantic Data Schema

To enable this level of sophisticated, contextual discovery, the underlying data structure cannot rely on simple HTML hyperlinks. The system must employ a rigorous topic knowledge graph utilizing established semantic web ontologies. By structuring the content as a machine-readable, yet statically generated graph, the platform can automatically compute the contextual features, backlinks, and evidence ledgers at compile time. The foundational schema will synthesize two primary W3C standards: the Provenance Ontology (PROV-O) and the Citation Typing Ontology (CiTO). PROV-O provides a formal, domain-agnostic model for tracking the lineage of information, ensuring that provenance becomes a core product requirement rather than a governance luxury15. It utilizes three core classes: prov:Entity (representing the data or document itself), prov:Activity (the action that created, modified, or retrieved the data), and prov:Agent (the person, organization, or software responsible)17. These entities are linked by standardized starting-point properties such as prov:wasGeneratedBy, prov:used, prov:wasAssociatedWith, prov:wasAttributedTo, and prov:wasDerivedFrom17. For documents retrieved from the web or authored by specific researchers, the Provenance, Authoring and Versioning (PAV) ontology—a direct specialization of PROV-O—provides additional granularity with properties like pav:authoredBy, pav:retrievedFrom, and pav:importedFrom18. Integrating PROV-O allows CognitiveLiberties.com to state explicitly where an evidence claim originates, guaranteeing verifiable lineage. If the site states that the United States lost its long-term status as a liberal democracy, this claim (prov:Entity) is strictly linked via prov:wasDerivedFrom to the V-Dem Democracy Report 2026 (prov:Entity), which in turn is linked via prov:wasAttributedTo to the V-Dem Institute (prov:Agent)1. However, provenance alone only explains origin; it does not explain rhetorical intent. This is where the Citation Typing Ontology (CiTO) becomes critical. CiTO enables the characterization of the nature of citations both factually and rhetorically19. It allows the graph to define exactly why a link exists, an essential requirement for a platform navigating politically sensitive claims. The data schema relies on JSON-LD embedded within the static Markdown and frontmatter of the site's repository. The static site generator will parse these semantic relationships during the build process to generate the HTML user interface, entirely removing the need for a runtime database.

By utilizing the reified form of CiTO citations, the platform treats the link itself as a discrete entity containing metadata19. This architecture enables the front-end to render a tooltip that explicitly states: "This link directs to the V-Dem 2026 report, which supports the preceding claim regarding democratic backsliding, and is authored by the V-Dem Institute." This completely removes the ambiguity of standard web hypertext, returning autonomy to the reader.

Editorial Controls and Semantic Distinctions

Because CognitiveLiberties.com operates explicitly as a human rights advocacy platform, distinguishing between platform endorsements, neutral documentary evidence, and opposing viewpoints is a paramount editorial necessity. If the site links to a state-sponsored surveillance whitepaper or a corporate data-brokering strategy, the link must not be interpreted by the user—or by external search crawlers analyzing the site's outbound link graph—as an endorsement of those practices. The semantic data schema provides the necessary editorial controls through CiTO predicates19. The editorial workflow requires that every external reference added to the platform be explicitly categorized, preventing any link from existing in an unclassified state. The editorial doctrine must enforce the use of distinct ontological types to govern visual rendering. Endorsements of privacy-preserving standards (such as Oblivious HTTP or client-side indexing) will utilize cito:isCitedAsRecommendedReadingBy or cito:isCitedAsPotentialSolutionBy19. When the platform utilizes data from neutral human rights organizations, it will employ cito:isCitedAsDataSourceBy or cito:isCitedAsEvidenceBy. Crucially, when addressing opposing viewpoints or documenting the architecture of censorship systems, the platform will utilize predicates such as cito:refutes, cito:critiques, or cito:isDisputedBy19. For example, the Freedom on the Net 2025 report highlights the rising trend of "AI sovereignty," noting that authoritarian settings utilize government-supervised AI tools to codify state censorship, pointing specifically to Vietnam enforcing party-line alignment on generative AI and Thailand rolling out the Pathumma large language model1. If CognitiveLiberties.com links directly to the documentation for Thailand's Pathumma LLM as an example of systemic cognitive control, the link edge must be strictly encoded as cito:isCritiquedBy or cito:isDisputedBy relative to the concept of cognitive liberty. This semantic typing translates directly into the visual language of the site. Drawing inspiration from Gwern.net's link-icon CSS systems12 and chess commentary-style evidential markers13, the platform will append automated visual indicators to every hyperlink. A green checkmark subscript might denote cito:supports, a neutral document icon denotes cito:isCitedAsDataSourceBy, and a red warning triangle denotes cito:critiques. This visual vocabulary ensures that users possess full contextual awareness of the target destination's editorial standing before they commit their attention, significantly reducing the cognitive load required to parse dense socio-technical debates.

Accessibility, Mobile Navigation, Multilingual Content, and Core Performance

A knowledge network is inherently exclusionary if its advanced discovery features degrade performance, violate privacy, or render the platform unusable on mobile devices or in restricted bandwidth environments. The system must operate primarily as static, server-rendered, first-party functionality to fulfill the platform's zero-tracker mandate.

Static Server-Rendered Architecture and Client-Side Search

The platform will utilize a modern Static Site Generator (SSG). During the build process, the SSG parses the Markdown files and their embedded JSON-LD frontmatter, resolving all CiTO and PROV-O edges, and rendering them directly into static HTML files. Consequently, the bidirectional backlinks and evidence panels are generated entirely at compile time. This eliminates the need for a runtime database query when a visitor loads a page, guaranteeing exceptionally fast loading performance and absolute visitor privacy. To facilitate discovery without a commercial search engine, the platform will implement Pagefind. Pagefind is a fully static search library compiled to WebAssembly (WASM)20. During the static build process, Pagefind indexes the generated HTML, respecting data-pagefind-body attributes to ensure only relevant content is indexed while repeated navigation elements are ignored21. It outputs a static chunked index. When a user searches the Resource Navigator, the Pagefind client downloads only the specific index chunks required to execute the query entirely within the user's browser20. This architecture ensures that search queries—which the site explicitly defines as vulnerable "thought data"1—never leave the user's device, perfectly aligning with the platform's mission.

Mobile Navigation and Accessible Popovers

The rich transclusion popovers inspired by Gwern.net must be carefully adapted for touch interfaces. While mouse-hover states work flawlessly on desktop, mobile users require explicit tap targets. The system will implement a dual-interaction model: on desktop, hovering a citation triggers the popover, while on mobile, tapping the citation opens an accessible bottom-sheet dialogue containing the annotation and provenance data. This ensures WCAG compliance regarding touch target sizes and screen-reader focus management, preventing the rich graph experience from alienating visually impaired users.

Link Rot and Local Archiving

Link rot poses a severe existential threat to any link-first architecture. If the V-Dem Institute or Freedom House alter their URL structures, the semantic graph breaks, and the evidence disappears. To mitigate this, the platform will implement a build-step archiving script. Similar to the architecture of Gwern.net, whenever a new external URL is added to the graph, the build script automatically requests a snapshot via the Internet Archive, downloads a local copy of the PDF or HTML, and appends the local\_archive\_url to the node's schema14. The frontend UI will always offer the primary URL first to respect the original publisher, but transparently provides the local mirror if the primary fails, ensuring true censorship resistance.

Multilingual Knowledge Federation

The current site contains a Spanish global summary, "Resumen mundial en español"1. However, in a flat architecture, translations often exist in isolated silos. The new semantic schema handles multilingual content via sub-graph compilation. The schema allows a Spanish node to be intrinsically linked to its English counterpart via the prov:alternateOf property. When a user toggles their language preference, the system does not merely redirect them to a translated homepage; it traverses the prov:alternateOf edge to land them on the exact equivalent conceptual node in the target language, maintaining their exact place within the knowledge graph.

Detailed Page-Layout Concepts

To materialize the link-first architecture, the user interface must be comprehensively redesigned to prioritize lateral navigation and contextual disclosure. The following five page-layout concepts define the visual and structural presentation of the knowledge graph.

Layout Concept 1: The Concept Hub (Node View)

This layout replaces the traditional long-form article. It is designed to act as a clearinghouse for a specific theoretical or practical idea within the cognitive liberty framework, such as the "Layered Censorship Stack."

Layout Concept 2: The Evidence Ledger Interface

Addressing the currently broken /research-evidence pathway5, this layout presents the raw data supporting the platform's mission, such as the OONI measurement data.

Layout Concept 3: The Organizational Profile

Fixing the dead /cognitive-liberty-organizations directory3, this layout maps the human and institutional actors within the network, such as Freedom House or Access Now.

Layout Concept 4: The Discovery Stream (Resource Navigator)

Replacing the inaccessible /cognitive-liberty-resource-navigator4, this interface provides a guided, problem-based entry point into the knowledge graph.

Layout Concept 5: Deep Reading View (Report Presentation)

For the 74 exact research reports and institutional implementation kits currently hosted1, this layout optimizes deep focus while retaining underlying graph connectivity.

Measurement Approach for Successful Discovery

Traditional web analytics rely entirely on behavioral tracking, third-party cookies, and identifiable browsing histories to measure engagement. CognitiveLiberties.com strictly prohibits these practices, boasting a zero-tracker ecosystem1. Therefore, measuring the success of the new link-first discovery experience requires a radical alternative paradigm focused on aggregate infrastructure metrics and graph density rather than individual user profiling. Success in a discovery hub is not defined by "time-on-site" or "pageviews per session." If a user arrives, immediately finds a vital external organization, and leaves the site to donate or download a tool, that is a successful interaction that a traditional analytics platform would categorize as a negative "bounce." To measure success without tracking, the platform will rely exclusively on first-party, server-side log analysis utilizing tools like GoAccess or AWStats. These tools parse standard Nginx or Apache access logs, configured to cryptographically hash or entirely strip IP addresses prior to analysis. The metrics of success will include:

1. Graph Traversal Depth Metrics: By analyzing the aggregate referrers strictly within the internal domain, the platform can measure how often users move from a Concept Hub to an Evidence Ledger node. This indicates successful lateral discovery without profiling who made the jump.

2. Outbound Exit Velocity: By utilizing a privacy-respecting, anonymized edge-worker script (which counts a generic outbound click event but strips all user metadata and session IDs), the platform can measure which "Continue Exploring Elsewhere" links are utilized most frequently. A high exit velocity to verified organizations indicates the site is successfully acting as a hub rather than a destination.

3. Client-Side Search Efficacy: Because Pagefind operates entirely in the browser, search queries cannot be logged on the server. However, the system can measure the aggregate download frequency of the Pagefind index chunks to understand overall search utilization rates without ever seeing the vulnerable query data itself.

4. Zero-Tracker Verification: The platform will expand its existing Cognitive Liberty Audit Protocol and First-Party Self-Audit1 to include automated continuous integration (CI) tests that deliberately fail the site build if any external script, un-mirrored external font, or tracking pixel is detected in the compiled output.

Prioritized Engineering Backlog

The transition to a semantic, static knowledge graph requires phased execution. The following engineering backlog separates immediate, highly feasible improvements from complex tasks requiring new software architecture or external cooperation.

The evolution of CognitiveLiberties.com into a link-first knowledge network represents a critical alignment of the platform's technological architecture with its philosophical mission. By dismantling hierarchical silos and replacing them with a semantic graph governed by PROV-O and CiTO ontologies, the platform ensures that the provenance and rhetorical intent of every claim are transparently documented. The implementation of client-side static search, bidirectional backlinks, and rich transclusion popovers empowers visitors to explore the complex landscape of mental privacy and digital censorship autonomously. By explicitly designing pathways that encourage users to leave the site for verified external resources, the system rejects the attention-extraction mechanics of the commercial web, guaranteeing that the process of becoming informed remains entirely free from the surveillance structures the movement seeks to dismantle.

Works cited

1. Cognitive Liberties, https\://cognitiveliberties.com/

2. https\://cognitiveliberties.com/cognitive-liberty-standards

3. https\://cognitiveliberties.com/cognitive-liberty-organizations

4. https\://cognitiveliberties.com/cognitive-liberty-resource-navigator

5. https\://cognitiveliberties.com/research-evidence

6. Rami's Open Source Toolkit \- Rami.party, https\://rami.party/workshop/opensource/

7. Robots.txt and Crawler Control | SearchEngines.Net, https\://www\.searchengines.net/guides/robots-txt-and-crawler-control/

8. Privacy Search Engines | SearchEngines.Net, https\://www\.searchengines.net/privacy-search-engines/

9. Search Engine Directory: Every Engine and Its Source, https\://www\.searchengines.net/directory/

10. Design Of This Website · Gwern.net, https\://emilien.ca/Notes/Notes/references/Design%20Of%20This%20Website%20-%20Gwern.net.pdf

11. Manual of Style \- Gwern.net, https\://gwern.net/style-guide

12. Design Graveyard \- Gwern.net, https\://gwern.net/design-graveyard

13. Subscripts For Citations \- Gwern.net, https\://gwern.net/subscript

14. Design Of This Website \- Gwern.net, https\://gwern.net/design

15. 5\. Provenance information \- FAIR Cookbook, https\://faircookbook.elixir-europe.org/content/recipes/reusability/provenance.html

16. Agentic Apps Need a Semantic Layer — and It Is More Available, https\://medium.com/@yu-joshua/agentic-apps-need-a-semantic-layer-and-it-is-more-available-than-you-think-c3479df99021

17. PROV-O: The W3C Provenance Ontology \- CASRAI, https\://casrai.org/dictionary/term/prov-o

18. PAV \- Provenance, Authoring and Versioning \- GitHub Pages, https\://pav-ontology.github.io/pav/

19. CiTO, the Citation Typing Ontology, https\://sparontologies.github.io/cito/current/cito.html

20. Configuration Reference \- Astro Starlight, https\://starlight.astro.build/reference/configuration/

21. Build a documentation site with Astro and Markdown \- Flavio Copes, https\://flaviocopes.com/astro-documentation-site/