Strategic Implementation Plan: Launching a Link-First Discovery Network

Full report

On this page

Executive Summary and Institutional Context

Modern information discovery relies heavily on centralized architecture. Search engines, algorithmic social feeds, and generative artificial intelligence (AI) systems have established a paradigm where intermediary platforms act as a governance layer over human inquiry. These systems interpret search intent, retain user context, shape attention, and utilize original independent research as training data while obscuring the primary sources from the end user1. For independent publishers, this paradigm extracts the value of original work while actively diminishing direct website traffic, severing the vital relationship between the researcher and the reader. CognitiveLiberties.com operates as a foundational public-interest node, publishing educational research on cognitive liberty, AI governance, private inquiry, and internet resilience1. The site’s technical ethos strictly prohibits behavioral advertising, third-party tracking, and external runtime requests, reflecting a commitment to the principle that lawful queries should not be transformed into hidden risk scores or ideological profiles1. Its existing infrastructure encompasses a 74-report research library, a Resource Navigator, and an extensive reviewed network of organizations1. However, the continued reliance on centralized search engines for user discovery contradicts the site’s mandate to build an internet with practical exit and censorship resistance. To address this systemic vulnerability, this report details a comprehensive, realistic, and evidence-driven implementation plan to establish a decentralized, direct website-to-website information discovery network. The objective is to treat web discovery as public-interest infrastructure, designing a reproducible framework that facilitates voluntary cooperation among independent publishers without compromising user privacy or relying on algorithmic gatekeepers.

Deliverable 1: Comparative Decision Memorandum

Transitioning to a link-first discovery model necessitates a rigorous evaluation of technical feasibility, editorial burden, and the socio-technical incentives required to foster a thriving ecosystem of independent publishers. Three competing strategies have been developed, ranging from foundational HTML semantics to advanced local AI protocols.

Strategy A: Low-Complexity Link Renaissance

Strategy A focuses on maximizing the utility of the foundational open web: hyperlinks, static directories, and traditional syndication formats. It prioritizes contextual outbound links, curated reading trails, and standard feed availability. Technical Architecture, Costs, and Maintenance: This strategy requires virtually zero active server-side processing, leveraging the existing static or cached infrastructure of CognitiveLiberties.com. Directories are maintained using the Outline Processor Markup Language (OPML) 2.0 specification, a highly adopted XML format designed specifically for hierarchical lists and the syndication of RSS/Atom feeds2. Furthermore, this strategy integrates Microformats2—specifically h-entry for individual publications and h-feed for collections—enabling machine-readable metadata (p-name, p-author, u-url) to be embedded directly into HTML without requiring complex Application Programming Interfaces (APIs)4. Development effort is minimal, operating costs are negligible, and the maintenance burden is almost exclusively editorial, as there are no dynamic endpoints to secure against intrusion. Editorial Governance and Safeguards: Source selection relies entirely on the editorial team at CognitiveLiberties.com to identify, verify, and link to external human rights bodies, digital privacy groups, and research centers1. Verification is conducted against the site's existing Cognitive Liberty Audit Protocol1. Abuse prevention is structurally guaranteed because the network is entirely "pull-based" and curated by the host; third parties cannot unilaterally inject links. Conflict-of-interest safeguards are maintained through a transparent editorial policy published alongside the OPML directory. Publisher Incentives and Funding: Publisher incentives under Strategy A are modest. Independent publishers benefit from referential traffic and the prestige of inclusion in a vetted resource directory. However, because the system relies on manual, unidirectional curation, it lacks the dynamic reciprocity that drives compounding network effects. Funding alternatives for this static approach are limited to general institutional grants, as it does not introduce novel software development. Compatibility and Privacy: Compatibility with Search Engine Optimization (SEO) is exceptionally high, as standard contextual links and Microformats2 are natively understood by web crawlers. Accessibility is inherently strong, relying on semantic HTML. User privacy is absolute, as no behavioral data is exchanged during navigation. Adoption Barriers and Discontinuation Conditions: The primary practical adoption barrier is scale; maintaining an exhaustive, manually updated directory across a global network becomes unsustainable as the ecosystem expands. The pilot should be discontinued or redesigned if the editorial team spends more than ten hours per month verifying broken links, or if voluntary link-following by users falls below a statistically significant threshold during usability testing.

Strategy B: Cooperative Discovery Network

Strategy B advances the link-first concept by establishing a bidirectional, voluntary cooperative network among independent publishers. This strategy introduces open protocols to create a decentralized webring, allowing participating sites to notify each other of references and share decentralized identity verification. Technical Architecture, Costs, and Maintenance: This architecture introduces moderate technical complexity. It requires implementing Webmention, a World Wide Web Consortium (W3C) recommendation for cross-site notifications6. When a participating site links to CognitiveLiberties.com, its server sends a Webmention HTTP POST request. To function securely, the receiving server must verify the source link, which necessitates Server-Side Request Forgery (SSRF) protections, rate limiting, and spam filtering8. Additionally, this strategy incorporates IndieAuth, an open standard decentralized authentication protocol built on OAuth 2.0 that allows publishers to verify their identity using their resolvable profile URLs10. To distribute the directory, OPML 2.0's inclusion feature (type="include") is utilized, allowing multiple sites to dynamically syndicate portions of the network directory without manual duplication2. Operating costs increase moderately due to the need for a lightweight backend service (e.g., Node.js or Python) to process Webmentions. Editorial Governance and Safeguards: Governance shifts from centralized curation to federated quality standards. The cooperative network relies on a shared charter dictating inclusion criteria. Verification occurs dynamically via IndieAuth, proving domain ownership, while Webmention endpoints require strict allow-lists to prevent abuse9. Abuse prevention transitions from a passive to an active posture, requiring cryptographic verification or domain reputation scoring to discard malicious pings. Publisher Incentives and Funding: Incentives are significantly stronger in this model. By utilizing Webmention, publishers receive visible acknowledgment (such as automated trackbacks or "mentioned by" labels) when they contribute to the discourse, incentivizing cross-referencing and collaboration. Funding alternatives include decentralized grants targeting open-source web infrastructure, such as the NLnet Foundation, which exclusively funds privacy-preserving, open-source technology14. Compatibility and Privacy: SEO impact is neutral-to-positive, as organic, bidirectional linking naturally establishes domain authority. Accessibility remains dependent on the underlying HTML implementation of the Webmention displays. Privacy remains robust, as Webmention and IndieAuth operate server-to-server, shielding the end-user's browser from third-party tracking. Adoption Barriers and Discontinuation Conditions: The primary barrier is the technical friction required for independent publishers to implement Webmention and IndieAuth endpoints. The pilot should be discontinued if it fails to recruit a minimum of five active institutional partners within six months, or if malicious spam processing overwhelms the backend infrastructure despite mitigation efforts.

Strategy C: AI-Assisted Link Discovery

Strategy C proposes augmenting the link-first network with privacy-preserving, local artificial intelligence to help users navigate the independent web. Rather than relying on centralized AI that obscures sources, this approach deploys an AI research navigator that explicitly guides users to original, independently hosted materials based on a curated vector database. Technical Architecture, Costs, and Maintenance: The technical architecture is highly complex. To adhere to CognitiveLiberties.com's strict zero-tracking policy1, the AI cannot rely on centralized API calls that log user queries. Instead, the architecture must utilize local, in-browser inference models using technologies like WebLLM, WebGPU, or ONNX Runtime Web15. To orchestrate the interface, the Agent-User Interaction (AG-UI) protocol is required. AG-UI is an open, lightweight, event-based specification that standardizes real-time, bidirectional communication between AI agents and user interfaces over protocols like Server-Sent Events (SSE) or WebSockets17. It streams JSON-encoded events, allowing the frontend to render token-by-token text generation (message.delta), execute tool calls (tool\_call), and patch UI states (state.patch)18. Furthermore, publishers across the network would implement the llms.txt standard—a markdown file placed at the root directory that provides structured, noise-free context to AI agents20. Maintenance burdens are severe, requiring constant updates to the local Retrieval-Augmented Generation (RAG) vector database and ongoing management of AG-UI state synchronization. Editorial Governance and Safeguards: Source selection is technically enforced by the RAG vector database, which is pre-compiled and served statically to the browser, restricting the LLM to only query and summarize sources within the verified OPML directory22. Conflict-of-interest safeguards require absolute transparency regarding the model weights, the vector database contents, and the absence of algorithmic bias in retrieval. Publisher Incentives and Funding: Publishers are incentivized by inclusion in a cutting-edge discovery tool. By simply hosting an llms.txt file, independent publishers ensure their content is accurately mapped for agentic workflows24. Funding would likely require substantial academic or institutional backing due to the high research and development costs of in-browser AI optimization. Compatibility and Privacy: Compatibility with ethical Answer Engine Optimization (AEO) is a core feature, heavily leveraging llms.txt. However, industry data indicates that currently, major AI models rarely crawl llms.txt for discovery (only 0.1% of AI bot traffic in recent studies), meaning this standard is primarily useful for explicit agentic integrations rather than general visibility24. Accessibility is a major concern, as highly dynamic, token-streaming AG-UI interfaces present significant challenges for screen readers. Privacy is structurally guaranteed because inference occurs entirely client-side. Adoption Barriers and Discontinuation Conditions: The adoption barriers are massive: client-side AI requires significant user device resources (GPU/RAM), potentially isolating users on older hardware or limited bandwidth, directly contradicting the goal of public-interest accessibility15. The pilot should be rejected if the user interface suffers from severe latency, if hardware exclusion metrics are too high, or if the local model hallucinates sources not present in the verified directory.

Synthesis and Strategic Recommendation

Table 1 provides a comparative matrix of the three strategies based on critical implementation vectors. Table 1: Comparative Analysis of Discovery Strategies

The analysis dictates that a phased integration of Strategy A and Strategy B represents the most sustainable, evidence-driven approach. Strategy A establishes the immediate semantic groundwork (Microformats2, OPML), while Strategy B builds the necessary social and technical infrastructure for a decentralized network (Webmention, IndieAuth). Strategy C, while theoretically aligned with privacy through local inference, introduces disproportionate technical risk, massive development overhead, and unacceptable hardware exclusion at this stage. Therefore, the implementation plan will proceed with a hybrid Strategy A/B model, establishing the Cooperative Discovery Network.

Deliverable 2: Minimum Viable Pilot Specification

The Minimum Viable Pilot (MVP) establishes a functioning, bidirectional discovery network between CognitiveLiberties.com and an initial cohort of independent publishers. The pilot strictly adheres to the site's zero-tracker, zero-adtech mandate, utilizing no third-party analytics or client-side telemetry1.

Architecture and Technical Configuration

The MVP relies on a static-first infrastructure augmented by isolated microservices for dynamic protocols.

1. Distributed Directory (OPML 2.0): An OPML 2.0 file is hosted at the root of CognitiveLiberties.com, containing the verified network of organizations, standards bodies, and research centers1. Participating sites will use the type="include" attribute to dynamically parse and render this directory on their domains, ensuring updates made by the central node propagate seamlessly2.

2. Semantic Markup (Microformats2): The existing Resource Navigator and 74-report library will be retrofitted with Microformats2 (h-entry, h-feed, u-url, p-author) to guarantee that all outbound links and publications are machine-readable for syndication and authorship attribution4.

3. Federated Engagement (Webmention & IndieAuth): A self-hosted Webmention receiver will be deployed. To mitigate SSRF vulnerabilities, this receiver will operate strictly on an allow-list basis, only accepting incoming mentions from domains verified within the OPML directory8. IndieAuth will be configured to allow publishers to authenticate to the network using their domain profiles, linking the Authorization Endpoint to their root URL11.

4. Agentic Guidance (llms.txt): A standard llms.txt file will be published at the root domain to explicitly guide ethical, opt-in AI agents toward the OPML directory and highlight the site's privacy doctrine, serving as a low-cost experiment in AEO20.

Baseline Audit Methodology

Because the site prohibits client-side analytics1, traditional browser-based tracking is impossible. The baseline audit will rely entirely on privacy-respecting server log analysis. Prior to deployment, a 30-day log audit will establish the baseline ratio of direct traffic, referring domains, and search-engine queries, relying exclusively on HTTP Referer headers and aggregated request volumes.

Measurable Experiments and Metrics

The pilot will measure success through the following experimental metrics:

• Voluntary Link-Following Rate: Calculated via server-side exit tracking. Outbound links will temporarily route through an anonymized internal redirect (e.g., /exit?url=...) to count the volume of users actively utilizing the network directory.
• Successful Source Discovery in Usability Tests: Qualitative metric gathered through structured, opt-in usability testing with a cohort of 20 researchers. Success is defined as the user locating a primary source on a partner site within three clicks from the CognitiveLiberties.com directory.
• Returning Readership via Privacy-Respecting Subscriptions: Measured by the server-side request volume to the site's standard RSS/Atom and h-feed endpoints, indicating users choosing direct syndication over social media feeds.
• Active Independent Partner Resources: The absolute number of external domains successfully implementing the OPML inclusion and returning verified Webmentions.
• Broken-Link Decay Rate ([Figure omitted from source export]): Calculated as [Figure omitted from source export]. Measured via an automated weekly cron job scanning the OPML directory.
• Source Diversity: The variance of unique domains effectively referring traffic back to CognitiveLiberties.com, indicating a healthy ecosystem.
• Independence from Search Referral: The percentage decrease in the share of total traffic originating from centralized search engines, shifting toward direct or network referrals.

Resource Estimates

Table 2 outlines realistic resource estimates for the MVP phase. Table 2: Realistic Resource Estimates (MVP Phase)

Stop/Go Decision Criteria

The pilot operates on a 180-day evaluation cycle. The initiative will receive a "Go" for permanent integration if it secures at least ten active participating organizations, and if referral traffic from the federated network accounts for at least 15% of total non-direct discovery. A "Stop" or "Redesign" decision will be triggered if the Webmention spam-to-signal ratio exceeds 20% despite the allow-list, if the engineering maintenance required to monitor the endpoints exceeds ten hours per month, or if usability tests indicate severe user confusion regarding the directory structure.

Deliverable 3: Prioritized Engineering and Editorial Backlog

The realization of the Cooperative Discovery Network requires a disciplined, 365-day roadmap. The backlog bridges technical deployment with editorial outreach, ensuring the infrastructure scales alongside human participation.

365-Day Implementation Roadmap

Milestone 1: 30-Day (Foundation and Auditing)

• Engineering: Implement Microformats2 (h-entry, h-feed) across the 74 public research guides1. Generate the initial OPML 2.0 directory based on the existing reviewed network1. Publish llms.txt at the root domain20.
• Editorial: Finalize network inclusion criteria based on the Cognitive Liberty Audit Protocol1. Conduct the 30-day baseline server log audit.

Milestone 2: 90-Day (Endpoint Deployment and Outreach)

• Engineering: Deploy and secure the self-hosted Webmention receiver, strictly configured to the OPML allow-list to prevent SSRF8. Establish the server-side exit tracking mechanism for the link-following metric.
• Editorial: Initiate targeted outreach to 10 high-alignment digital rights and neurorights organizations1. Publish the public-facing proposal ("Rebuilding the Link-Driven Internet").

Milestone 3: 180-Day (Federation and Evaluation)

• Engineering: Implement IndieAuth Authorization Endpoints for federated login among network publishers11. Enable OPML dynamic inclusion testing across partner domains13.
• Editorial: Execute usability testing cohort. Review MVP metrics against Stop/Go criteria. Resolve broken links and iterate on the shared directory taxonomy.

Milestone 4: 365-Day (Scale and Decentralized Governance)

• Engineering: Optimize infrastructure for scale. Deploy static Webmention syndication generators to limit dynamic server load during traffic spikes.
• Editorial: Expand the network globally. Transition directory governance from a centralized model to a shared steering committee of trusted node operators.

Dependency Diagram and Risk Register

Table 3 models the critical path dependencies and mitigation strategies for the network's deployment. Table 3: Dependency Diagram and Risk Register

Deliverable 4: Sustainable Governance and Funding Proposal

A decentralized network cannot rely on proprietary funding mechanisms or venture capital, as these inevitably demand monetization through user surveillance, data brokering, or behavioral advertising—practices that directly violate the core tenets of cognitive liberty1. Consequently, the network requires a sustainable, open-source governance and funding architecture.

Governance Architecture

Governance of the Cooperative Discovery Network will initially follow a "Benevolent Dictator for Now" (BDFN) model, operated by the editorial team at CognitiveLiberties.com to ensure rapid iteration during the MVP phase. As the network matures past the 180-day milestone, governance will transition to a federated consensus model. Participating publishers who successfully maintain IndieAuth compliance and adhere to the shared quality standards for six consecutive months will be invited to form a network steering committee. This committee will jointly oversee the master OPML directory repository. To prevent conflicts of interest, no commercial entity or organization reliant on targeted advertising revenues will be permitted voting rights. Any modifications to the core inclusion criteria must pass a cryptographic vote signed by the domain operators using their IndieAuth credentials.

Funding Alternatives

To maintain zero reliance on ad-tech, funding must be sourced through mechanisms dedicated to public-interest technology and digital resilience.

1. The NLnet Foundation: The initiative aligns perfectly with NLnet's mandate to support open, secure, and decentralized internet technologies. Grants from funds such as the NGI Zero Discovery program do not extract commercial benefit and require all funded developments to be published under open-source licenses, making them ideal for covering the engineering costs of developing secure Webmention/IndieAuth endpoints14.

2. Philanthropic Digital Rights Grants: Organizations focused on anti-censorship and internet freedom can provide operational funding. The network directly addresses the "Discoverability" and "Connect" threat vectors highlighted in the Cognitive Liberties framework, wherein centralized systems silently rank publishers and eliminate practical exit options1.

3. Institutional Micro-Grants: Leveraging the existing institutional implementation kits (designed for libraries, education, and journalism)1, academic networks utilizing the curated OPML directories can provide sustaining micro-grants in exchange for prioritized technical support during their integration.

Deliverable 5: Initial Outreach Framework for Independent Publishers

Recruiting independent publishers requires a compelling value proposition that emphasizes mutual sovereignty and resilience over traditional marketing or traffic metrics. The outreach framework targets the existing human-rights bodies, digital-rights groups, neurorights organizations, and standard bodies already referenced within CognitiveLiberties.com's ecosystem1. Target Profiling: Initial outreach will focus on organizations that self-host their content, possess technical competency, and demonstrate an explicit commitment to privacy (e.g., organizations operating without programmatic advertising or hidden tracking architecture). The Value Proposition: The correspondence will frame participation not as an SEO link-building exchange, but as vital civic infrastructure development. The core message focuses on mutual defense: centralized search engines interpret intent and AI assistants synthesize answers, acting as intermediaries that abstract original source material1. By establishing reciprocal links via OPML and Webmention, publishers create a durable, censorship-resistant web of trust that guarantees audiences can find original research even if algorithmic indexing deprioritizes it. Phased Onboarding Process:

1. Phase 1 (Alignment): The partner receives a customized brief detailing how their existing research aligns with the cognitive liberty framework and the specific societal threats of layered censorship1.

2. Phase 2 (Low-Friction Adoption): The partner is provided with a lightweight onboarding kit. This kit requires zero backend coding. It requests the addition of their RSS feed to the shared OPML directory and the placement of standard HTML contextual links to the network on their resources page.

3. Phase 3 (Federated Integration): For advanced partners, technical support is provided to enable IndieAuth and Webmention, unlocking bidirectional engagement metrics and decentralized identity without relying on third-party tracking7.

Deliverable 6: Rebuilding the Link-Driven Internet

The following text serves as the public-facing proposal, designed to be published on CognitiveLiberties.com and distributed across partner networks. Title: Rebuilding the Link-Driven Internet: A Manifesto for Private Inquiry and Sovereign Discovery The architecture of information discovery is quietly failing the independent web. For two decades, the internet relied on a decentralized web of hyperlinks, where publishers curated pathways for readers, and discovery was an act of human editorial intent. Today, that ecosystem has been eclipsed. Search engines parse intent through opaque algorithms, social feeds optimize for behavioral extraction, and generative AI systems synthesize our original work into homogenized answers, severing the vital connection between the researcher and the reader. At CognitiveLiberties.com, our core thesis is that the mind is a private space. Curiosity is not a confession, and a lawful query should not become a permanent judgment stored in a corporate server1. Yet, the centralized systems we rely on to navigate the web do exactly that. They infer who is asking, and they judge who is publishing. This centralization is not just an economic threat to independent publishers; it is an architectural vulnerability in global civil liberties, contributing to the expanding "censorship stack" that restricts access to vital information1. It is time to build an internet with practical exit. We are launching a Cooperative Discovery Network—a voluntary, decentralized architecture designed to revive direct website-to-website navigation. We are moving away from dependency on algorithmic intermediaries. Instead, we are utilizing open, public-interest protocols. Through OPML directories, we are sharing our curated lists of trusted human rights, digital privacy, and research organizations. Through Microformats and Webmentions, we are enabling sites to converse directly, recognizing when independent research informs new ideas, without requiring a centralized platform to mediate the exchange. This network operates with zero behavioral advertising, zero third-party tracking, and zero hidden telemetry. It is built on transparency, shared quality standards, and the fundamental right to read without being monitored. We invite independent publishers, academic centers, and digital rights organizations to join this architecture. By linking directly to one another, we bypass the algorithmic gatekeepers. We ensure that our work remains discoverable, resilient, and deeply human. Success in the modern internet is not measured by manufacturing search-engine signals; it is measured by helping visitors reach useful, original sources in a space where their cognitive liberty is entirely protected.

Deliverable 7: Reproducible Framework for Independent Adoption

A cornerstone of resilient, censorship-resistant infrastructure is that it must be permissionless1. Other independent websites must be able to adopt this model and instantiate their own networks without requiring authorization from CognitiveLiberties.com. The following reproducible framework outlines the technical steps for any sovereign publisher to establish a link-first discovery node.

Step 1: Establish Semantic Foundations with Microformats2

Before joining or creating a network, a site's content must be inherently machine-readable. Adopt Microformats2 by adding specific class attributes to standard HTML elements.

• Wrap the main article or publication in class="h-entry".
• Define the title with class="p-name".
• Ensure the author is marked with class="p-author h-card".
• Identify the permalink with class="u-url".
• For a collection of reports, wrap them in class="h-feed". These simple CSS classes allow any decentralized parser to extract metadata accurately without relying on proprietary graph protocols or complex APIs4.

Step 2: Publish and Syndicate a Curated OPML Directory

To share trusted resources, generate an OPML 2.0 file. This XML document lists the RSS/Atom feeds and HTML URLs of partner sites3. Place this file at the root of the domain (e.g., example.com/network.opml).

• Ensure the document utilizes type="rss" and provides the xmlUrl for direct feed links.
• Critically, use the type="include" attribute to seamlessly pull in OPML directories curated by other trusted nodes. This allows for dynamic syndication, creating a decentralized web of references that updates automatically when a trusted peer updates their list2.

Step 3: Implement Webmention for Cross-Site Awareness

To allow sites to communicate and acknowledge references without centralized comment systems, advertise a Webmention endpoint in the site’s HTML \<head\> using \<link rel="webmention" href="https\://example.com/webmention"\>7.

• For statically hosted sites, leverage open-source, self-hosted Webmention receivers.
• Ensure the receiver verifies incoming requests. The receiver must fetch the source URL (using an HTTP GET or HEAD request) to confirm it actually contains a link to the target URL7.
• Implement strict security measures to mitigate SSRF by refusing to process internal IP addresses or loopbacks, and enforcing rate limiting8.

Step 4: Establish Decentralized Identity with IndieAuth

To authenticate across the network without relying on massive tech conglomerates for Single Sign-On (SSO), implement IndieAuth10.

• By adding a \<link rel="me" href="..."\> tag to a personal or organizational domain that points back to the user's profile, the domain itself becomes the verifiable cryptographic identity.
• This extension of OAuth 2.0 allows publishers to utilize Authorization Endpoints and Token Endpoints entirely under their own control, ensuring that their identity remains fully sovereign and portable11.

Step 5: Guide Autonomous Agents with LLMs.txt

Recognize that automated AI agents will traverse the web, regardless of human navigation patterns. Provide clear, structured boundaries by placing an llms.txt file at the root of the server20.

• This plain-text markdown file should concisely explain the site's purpose and explicitly define the terms of cognitive liberty and privacy24.
• While current data shows minimal traditional SEO impact from this file26, it is highly effective for agentic workflows25. Direct these agents toward the OPML directory to find verified external sources, ensuring that when AI systems attempt to answer queries about your domain, they rely on your curated, evidence-driven network rather than hallucinated context.

Works cited

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

2. OPML 2.0 spec, https\://opml.org/spec2.opml

3. What is an OPML file? How to move your RSS feeds | Knowledge base, https\://feeder.co/knowledge-base/rss-basics/opml-file/

4. h-entry \- IndieWeb, https\://indieweb.org/h-entry

5. Using microformats in HTML \- MDN Web Docs, https\://developer.mozilla.org/en-US/docs/Web/HTML/Guides/Microformats

6. 40031 (Consider Adding Web Annotations to WordPress), https\://core.trac.wordpress.org/ticket/40031

7. Socialwg/2016-06-07-minutes \- W3C Wiki, https\://www\.w3.org/wiki/Socialwg/2016-06-07-minutes

8. HTTP client \- Lib.rs, https\://lib.rs/web-programming/http-client

9. Guides \- Big Iron, https\://www\.bigiron.cc/guides/

10. IndieAuth \- HandWiki, https\://handwiki.org/wiki/IndieAuth

11. IndieAuth \- W3C, https\://www\.w3.org/TR/indieauth/

12. IndieWeb Living Standard 12 February 2022 \- IndieAuth, https\://indieauth.spec.indieweb.org/20220212/

13. Archive: myPublicFeeds.opml (RSS 2.0 at Harvard Law), https\://cyber.harvard.edu/rss/myPublicFeedsOpml.html

14. NLnet Foundation: Supporting Open, Secure, and Decentralised, https\://donors.fundsforngos.org/information-technology/nlnet-foundation-supporting-open-secure-and-decentralised-internet-technologies/

15. Inside the Web AI Revolution: On-Device ML, WebGPU, and Real, https\://senoritadeveloper.medium.com/inside-the-web-ai-revolution-on-device-ml-webgpu-and-real-world-deployments-c34abbf22fdb

16. Stop Paying for API Calls: Run Powerful LLMs Entirely in Your, https\://drlee.io/stop-paying-for-api-calls-run-powerful-llms-entirely-in-your-browser-using-webgpu-and-local-rag-7e7fbe5057d4

17. AG-UI: How the Agent-User Interaction Protocol Works \- Codecademy, https\://www\.codecademy.com/article/ag-ui-agent-user-interaction-protocol

18. AG-UI: The Interface Protocol for Human-Agent Collaboration, https\://medium.com/@jatingargiitk/ag-ui-the-interface-protocol-for-human-agent-collaboration-a93025ab327c

19. Events \- Agent User Interaction Protocol, https\://docs.ag-ui.com/sdk/js/core/events

20. LLMs.txt: Control How AI Crawlers Use Your Content \- Royal Plugins, https\://royalplugins.com/blog/llms-txt-ai-crawler-control/

21. What Is LLMs.txt? Definition, Format, and How to Implement It, https\://searchatlas.com/blog/what-is-llms-txt/

22. Best vector databases for RAG in 2026 \- Articles \- Braintrust, https\://www\.braintrust.dev/articles/best-vector-databases-for-rag-2026

23. Outline Processor Markup Language (OPML) format \- Visure Solutions, https\://visuresolutions.com/alm-guide/outline-processor-markup-language-opml-format/

24. Squarespace & LLMs.txt Support: What it Means for AIO, https\://sarahmoon.net/blog/squarespace-llmstxt

25. How to Adapt LLMs.txt Files for Agents \- Wix.com, https\://www\.wix.com/studio/ai-search-lab/llms-txt-files-for-agents

26. llms.txt: what it is and why AI still does not read it \- Elevam, https\://elevam.es/en/blog/what-is-llms-txt/

27. OPML \- Wikipedia, https\://en.wikipedia.org/wiki/OPML

28. h-feed \- IndieWeb, https\://indieweb.org/h-feed

29. Do llms.txt files actually improve AI search visibility? \- Contentful, https\://www\.contentful.com/blog/llms-txt-search-visibility/

30. Social Web Working Group \- Publications \- W3C, https\://www\.w3.org/groups/wg/social/publications/