{"id":62305,"date":"2026-07-06T13:48:43","date_gmt":"2026-07-06T17:48:43","guid":{"rendered":"https:\/\/overcentral.com\/en\/?p=62305"},"modified":"2026-07-06T13:48:43","modified_gmt":"2026-07-06T17:48:43","slug":"marie-haynes-okf-brain-ai-agents","status":"publish","type":"post","link":"https:\/\/overcentral.com\/en\/marie-haynes-okf-brain-ai-agents\/","title":{"rendered":"Marie Haynes Builds a Personal OKF Brain for AI Agents"},"content":{"rendered":"

The concept of a “personal brain” for artificial intelligence agents might sound like science fiction, but SEO expert Marie Haynes has turned it into a practical reality. Building on the extraordinary response to her earlier deep dive into Google’s Open Knowledge Format (OKF), Haynes has spent months constructing a structured knowledge system that functions as her own digital cognition \u2014 a repository of expertise that AI agents<\/a> can read, navigate, and act upon with precision. The result is not merely an interesting experiment; it is a working model for how professionals can prepare for the agentic web<\/a>, a shift in which autonomous AI agents will increasingly handle complex tasks on behalf of users.<\/p>\n

Google’s Open Knowledge Format, or OKF, is at the heart of this transformation. While critics have pointed out that the underlying markdown files are hardly a novel invention, Google<\/a> has done something that the tech industry has struggled with for decades: it has established a standardized format for knowledge representation. When an agent receives an OKF file, it understands the structure without requiring custom integration or proprietary software. This universality is what makes OKF a genuinely significant development for the future of human-AI collaboration.<\/p>\n

The Anatomy of an OKF Brain: YAML Frontmatter, Index Files, and Markdown<\/h2>\n

Every OKF file begins with what is known as YAML frontmatter \u2014 a compact block of metadata at the top of a markdown file that tells an AI agent precisely what it is looking at. In Haynes’s personal brain, each file is assigned a specific type: concept, entity, playbook, reference, or system. These classifications allow the agent to instantly categorize and retrieve information without ambiguity.<\/p>\n

Her folder structure reflects this organizational logic. The brain contains separate directories for each type, and within those directories, individual markdown files represent discrete pieces of knowledge. A concept file, for example, might define a core SEO principle, complete with its relationship to other concepts, relevant sources, and practical applications. The YAML frontmatter for such a file includes fields that specify the file’s type, its connections to other files, and metadata that helps the agent determine relevance and priority.<\/p>\n

When an agent accesses Haynes’s OKF brain, the first file it encounters is the index. This index.md file functions as a master directory, listing the different areas the agent can explore. Instead of performing retrieval-augmented generation (RAG) across the entire knowledge base \u2014 a computationally expensive and often imprecise process \u2014 the agent can focus on the sections most relevant to the task at hand. This targeted approach dramatically improves both speed and accuracy.<\/p>\n

What is YAML frontmatter in an OKF file?<\/h3>\n

YAML frontmatter is a structured metadata block at the beginning of a markdown file that provides an AI agent with essential information about the file’s content, type, and relationships. In Marie Haynes’s OKF system, the frontmatter specifies whether a file is a concept, entity, playbook, reference, or system, along with connections to other files. This metadata allows agents to understand what they are reading without needing to parse the entire document first, making knowledge retrieval faster and more reliable.<\/p>\n

The system also automatically connects related concepts. When Haynes ingests new content \u2014 whether it is a blog post she has written, a Google announcement, or a piece of industry research \u2014 her brain identifies existing concepts that should be linked to the new material. This process, inspired by Andrej Karpathy’s LLM Wiki idea, creates an evolving graph of interconnected knowledge. Each dot in that graph is a markdown file, and the connections between them form a living map of everything Haynes knows about SEO and artificial intelligence.<\/p>\n

Visualizing a Living Knowledge Graph<\/h2>\n

One of the most striking aspects of Haynes’s OKF brain is its visual dimension. When the system is properly structured, the knowledge base can be rendered as a connected graph. Every node represents a markdown file, and the edges between nodes represent semantic relationships. Haynes can see, for instance, how her concepts for AI Overviews link directly to references drawn from Google’s official documentation and to the internal playbooks she has written for her team.<\/p>\n

This visualization is not merely decorative. It reveals gaps in knowledge, unexpected connections, and opportunities for deeper exploration. A concept that has few connections to other files may indicate an area that needs further development. A cluster of densely interconnected files might represent a particularly well-understood domain. Over time, the graph becomes a diagnostic tool for the health and completeness of the knowledge base itself.<\/p>\n

How does the OKF brain connect related concepts automatically?<\/h3>\n

When new content is ingested into the OKF brain, the system analyzes the material and matches it against existing concepts, entities, and references stored in the knowledge base. It then creates bidirectional links between the new file and any existing files that are semantically related. This process is automated and continuous, meaning the knowledge graph<\/a> grows and reorganizes itself without manual intervention. The result is a self-updating network of information that becomes more valuable as it expands.<\/p>\n

Haynes has also automated the ingestion of external information. Her system checks Google’s documentation daily for updates. If Google modifies a page about AI Answers or introduces new features in Search Console, her brain notifies her and automatically updates the relevant reference files. She no longer needs to rely on her biological memory to track every minor change in Google’s sprawling documentation ecosystem. Her OKF brain has access to far more information than she could ever hold in her head at once, and it never forgets.<\/p>\n

Building Playbooks That Save Days of Work<\/h2>\n

The most immediately practical aspect of Haynes’s OKF brain is the set of playbooks she has built. A playbook is a procedural document that outlines a specific workflow, complete with steps, decision points, and criteria for success. Because the playbook is written in OKF format, an AI agent can execute it autonomously.<\/p>\n

One example is her playbook for generating client proposals. This is typically a tedious, manual process that requires extensive customization and careful phrasing. Now, Haynes’s agent \u2014 which she calls Antigravity \u2014 follows the steps laid out in her proposal playbook to draft the entire document using her specific voice, reasoning style, and strategic logic. The result is a proposal that reads as if she wrote it herself, but it is produced in a fraction of the time.<\/p>\n

Another playbook handles site impact analysis after a Google algorithm update. Haynes created a procedural checkpoint that her agent follows when analyzing traffic shifts, ranking changes, and other signals. What previously required two days of intensive analysis now takes a matter of hours, and the reports it generates are comprehensive, consistent, and analytically sound. The agent does not simply summarize data; it applies Haynes’s own interpretive framework to the numbers, producing insights that reflect her expertise.<\/p>\n

What is a playbook in an OKF system?<\/h3>\n

In an OKF system, a playbook is a procedural markdown file that documents a specific workflow with clear steps, decision criteria, and success metrics. An AI agent can read the playbook and execute each step sequentially, applying the knowledge stored elsewhere in the OKF brain to make context-aware decisions. Playbooks transform human expertise into executable procedures that agents can run autonomously, saving significant time while maintaining quality and consistency.<\/p>\n

The power of this approach lies in its transferability. Once a playbook is written and tested, it can be shared, modified, and improved over time. It captures not just the steps of a process but the reasoning behind those steps. When Haynes writes a playbook, she is encoding her professional judgment into a form that an AI agent can apply reliably and repeatedly.<\/p>\n

How to Build Your Own OKF Brain<\/h2>\n

Haynes is emphatic that this is not a project reserved for developers or technical specialists. The barrier to entry is remarkably low. “You don’t need to be a coder to start,” she wrote in her initial post. “You can tell an agent to help you build your first OKF bundle based on the documentation Google has provided.” The key is to start experimenting with the format and to let the structure emerge organically as you add more content.<\/p>\n

She has shared the specific resources and prompt she uses to help others get started. The recommended approach involves giving an AI agent \u2014 whether Claude Code, ChatGPT Codex, or Google’s Antigravity \u2014 a set of reference links and asking it to help design a personalized OKF system. The links include Google’s official OKF documentation, the specification on GitHub, Karpathy’s LLM Wiki gist, Haynes’s own OKF article, and her video walkthrough.<\/p>\n

The prompt she suggests is direct and collaborative:<\/p>\n

“I want to build an OKF system similar to Marie’s. Read these links and then give me some ideas of what this would look like. Then, ask me questions one at a time so that together we can decide what we want to build.”<\/p>\n

This approach transforms the agent from a passive tool into an active collaborator that helps design the knowledge structure itself. The agent understands the OKF specification and can guide a user through the process of defining their own concepts, entities, playbooks, references, and systems. The result is a personalized knowledge base that reflects the user’s specific domain, expertise, and workflow needs.<\/p>\n

Why OKF Matters for the Future of Professional Knowledge Work<\/h2>\n

Haynes’s OKF brain is a prototype for a much larger shift in how professionals will interact with AI. For years, the promise of AI has been tempered by the reality of fragmented, inconsistent, and poorly structured data. Large language models are incredibly powerful, but they perform best when the information they receive is organized in predictable, machine-readable formats. OKF provides that predictability.<\/p>\n

The implications extend far beyond SEO. Any knowledge worker who relies on complex, evolving information \u2014 lawyers, researchers, analysts, consultants, product managers \u2014 can benefit from structuring their expertise in a format that AI agents can navigate. The agentic web, in which autonomous agents perform tasks, make decisions, and synthesize information on behalf of users, will demand precisely this kind of structured knowledge. Those who invest in building their own OKF systems today will be significantly ahead when that shift accelerates.<\/p>\n

Haynes’s brain is not a static archive. It is a dynamic, self-improving system. Her agent is always looking for ways to improve and connect the information it holds. When something seems off, she can ask the agent to correct it. When new information arrives, the system finds the right place for it. When she needs to produce something \u2014 a proposal, a report, an analysis \u2014 the brain synthesizes her notes, her references, and her playbooks into a coherent output that reflects her unique expertise.<\/p>\n

The question is no longer whether AI agents will become central to professional work. They will. The question is whether professionals will have the structured knowledge systems in place to direct those agents effectively. Marie Haynes has shown what that looks like in practice, and her OKF brain is a compelling answer to one of the most important challenges of the coming decade.<\/p>\n","protected":false},"excerpt":{"rendered":"

The concept of a “personal brain” for artificial intelligence agents might sound like science fiction, but SEO expert Marie Haynes has turned it into a practical reality. Building on the extraordinary response to her earlier deep dive into Google’s Open Knowledge Format (OKF), Haynes has spent months constructing a structured knowledge system that functions as […]<\/p>\n","protected":false},"author":7,"featured_media":74369,"comment_status":"closed","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"fifu_image_url":"https:\/\/iili.io\/CcVZoV1.jpg","fifu_image_alt":"Marie Haynes Builds a Personal OKF Brain for AI Agents","footnotes":""},"categories":[31],"tags":[],"class_list":["post-62305","post","type-post","status-publish","format-standard","has-post-thumbnail","category-technology"],"fifu_image_url":"https:\/\/iili.io\/CcVZoV1.jpg","fifu_image_alt":"Marie Haynes Builds a Personal OKF Brain for AI Agents","_links":{"self":[{"href":"https:\/\/overcentral.com\/en\/wp-json\/wp\/v2\/posts\/62305","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/overcentral.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/overcentral.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/overcentral.com\/en\/wp-json\/wp\/v2\/users\/7"}],"replies":[{"embeddable":true,"href":"https:\/\/overcentral.com\/en\/wp-json\/wp\/v2\/comments?post=62305"}],"version-history":[{"count":0,"href":"https:\/\/overcentral.com\/en\/wp-json\/wp\/v2\/posts\/62305\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/overcentral.com\/en\/wp-json\/wp\/v2\/media\/74369"}],"wp:attachment":[{"href":"https:\/\/overcentral.com\/en\/wp-json\/wp\/v2\/media?parent=62305"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/overcentral.com\/en\/wp-json\/wp\/v2\/categories?post=62305"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/overcentral.com\/en\/wp-json\/wp\/v2\/tags?post=62305"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}