The Living Content Graph: From Content Calendar to Mission Control

The team-of-one system that turns your company's knowledge into content - and heals itself when the facts change.

This newsletter is sponsored by Cello - referrals on autopilot.

Partnerships are one of the most profitable GTM motions, and most teams are still sleeping on it. Usually because building a program from scratch means a week of research before the first email.

Cello built a Claude skill that does it in a few minutes. Give Partner Radar your website URL and it returns a segmented partnership strategy plus 20-25 qualified partners with LinkedIn profiles, audience size, and a personalized hook for each.

---

Dear GTM Strategist,

Every AEO conversation I am in lately stops at the advice level. Nobody opens the hood to show me the actual wiring of a content operation.

So I was thrilled to get an email from Adina Timar offering to lay out her complete content system.

Three times a head of content (incl. Userpilot and PhantomBuster), she’s now running AEO at Weflow and building Sighted as a co-founder.

She operates her whole content function alone, based on what she calls the Living Content Graph, and says the output beats what she used to get from a team of twenty.

Fair warning: this is a long one, and it is a system, not a checklist. Read it with your own content operation in mind and steal the parts that fit.

Over to Adina.

---

The internet is full of frameworks and copy-paste playbooks. But stealing plays from everyone else is not the answer.

AI slop existed before we called it AI slop. It was just slop, because everybody was running the same play: research keywords, look at the pages that already rank, copy them, and maybe add your own input if you weren’t lazy.

That’s how we ended up with a ton of sameness, which we’re now labeling AI slop.

I always knew there was a better way. And because I’m a systems person, this is the system I designed to fight that sameness: content made for the sake of good content, that actually says something, and that addresses both humans and agents.

Here’s what you’ll walk away with:

• The old playbook and why it manufactured sameness (guilty of this too).

• The new three-layer living system I switched to - narrative, graph, engine - and what goes into each.

• A database structure you can copy: six schemas that keep your knowledge structured, clean, and retrievable.

• The engine loop: agents that mine, plan, write, and heal - with a human on every judgment call.

• Mission Control: the one UI where you run it all - Pulse, Brain, Launchpad.

• A minimum version you can start this week with a few Notion databases and a spreadsheet.

The Old Playbook (How We Got Here)

I’m guilty of the sameness. For years, I ran the exact playbook everybody ran:

• Spy on competitors with Ahrefs or Semrush

• Steal all their keywords

• Build a long monthly content plan

• Distribute it across 5 editors and 15 freelancers

• Get back ten different voices, more or less in-depth depending on each writer’s research, and have the editors try to make them great

And everything that’s relevant for AI right now (and was always relevant for SEO) - proper interlinking, building a knowledge graph, truth alignment - was an afterthought. All good SEO was an afterthought, even with a large team. We didn’t have the bandwidth for it, because the game was elsewhere:

• Linking was a bias, not a strategy: Links went in based on who inserted them and how much time they had. Freshness always won: if you had one new article, everything kept linking to it until the next one came along. Lazy writer plus lazy editor meant an article shipped with almost zero links.

• Refreshes were a guess: You never knew what to refresh or when, and you didn’t have time for it anyway.

• Inventory wasn’t a thing: No system tracked what you already had, so nothing stopped you from putting out slop or cannibalizing your own pages.

It was a game of who can publish more, faster. Velocity was everything (it still is, but in a different way), and it made me feel like a hamster on a spinning wheel.

And, worst of all, all it led to was more sameness. As long as it drove traffic and conversions, no one really cared.

Honestly? I’m glad AI came. Because now it feels like everybody’s finally making an effort to build something better (partially because the old playbooks don’t work anymore).

The narrative flipped from “we need more content, fast” to “we need more content, fast, but it needs to be unique” - content that actually says something, teaches something, and helps both humans and agents understand your business. Or how Google likes to call it: non-commoditized content.

That plays into my strengths. I’ve always known there’s a better way, and thinking in systems is how I approach everything. A system like this was hard to build with the tools from two or three years ago. Now, with AI, I run it as one person - and produce higher-quality output than I did two years ago with 5 editors and 15 freelancers.

The better way is a system that always knows what’s in your inventory and across which clusters, what you should build next, and what knowledge you’re actually authoritative on - so you’re intentional about what gets published, what gets refreshed, and what gets interlinked. At scale, without a huge team.

This is my team-of-one content system. Here’s how I operate it.

The Living Content System: Narrative → Graph → Engine

The living content system is a complete circle. It collects information, feeds that information into everything you produce, and heals itself when the information changes.

It stands on three layers that work together - like an onion, with your brand truth at the core:

• The narrative. One short document that represents your brand truth. You don’t guess it - you reverse-engineer it from founder voice and ICP voice. It plays the north star role: it never changes, and every layer around it exists to enforce it.

• The graph. The layer that sits on top of the narrative, derived from it. It’s the taxonomy and connections - pillars, clusters, subtopics, entities, themes - that structure everything you produce, so knowledge is easy to store, retrieve, and check. The brain of the operation.

• The engine. The outer layer that does the work: agents that mine, approve, plan, write, place, publish, heal, and measure - on a cadence, with a human in the loop. Its whole job is to build the graph and enforce the narrative, so what comes out is original instead of sameness.

Now let me explain each layer individually and how it works.

The Brand Narrative

The narrative is your brand truth: one short document that you write, ideally once, and that is used as the core truth of the whole system. It’s an expansion of the vision-and-mission document we all used to build years ago. It plays the same role, only now, someone actually reads it and uses it.

Write what you want your brand to be recognized for, and which entity associations you want models and buyers to make when your name comes up. There isn’t one template to rule them all here, but I kept mine simple, with sections for:

• The entity and the canonical sentence. What you are, in one sentence you’ll repeat verbatim everywhere: “X is the [category] platform for [who].”

• The graph around the entity. The category you claim, your products, the topics you want to own, the concepts you want associated with you, and the competitive neighborhood you want to be compared in.

• Who the brand is and why it exists. The problem story, in plain words.

• The narrative itself. The story you tell and the shift it names - why now, and what changes.

• Competitor positioning. One sentence per competitor on how you win against them.

• Positioning rules. The framings you always use, and the ones you ban - the descriptors that would absorb you into someone else’s entity.

When building this, don’t guess, and don’t treat it as a superficial task.

Everything is reverse-engineered from your founder voice and your ICP voice. Ideally, from a large volume of sales calls and transcripts, where you mine the exact vocabulary your founder uses when selling, and your ICP uses when describing their needs and pains.

There’s a simple reason for this: a founder selling explains why the product exists better than any positioning doc. It’s detailed, it’s with passion, and it’s in their own words. This is the best resource you can access to make your brand unique and clear.

If you don’t have sales transcripts, set up a few interviews with the founders and some power users. Ask them how they use the product, what they like about it, and how it’s helping them. Use their exact language, and use AI to build and add to the document until it sounds like you.

The Graph: Interconnecting Knowledge

Think of the graph as a neural network: it’s how information is stored, and how the different nodes communicate with each other in a structured way.

It’s a replica of a brain (or aims to be).

Why structure at all?

Because a structured approach is easier for LLMs to process. Imagine feeding AI one large .md document, 5,000+ words. One might be fine. Feed it five - positioning, voice, features, pricing - and the LLM will act as if it read them. (We all know they’re a bit self-preserving these days.) It reads portions and assumes the rest, and everything you produce with that context comes out superficial, leaning mostly on training data instead of your knowledge.

When knowledge lives in interconnected databases instead, information is stored in small bits, cataloged against a preset taxonomy. That makes it easy to retrieve exactly the right bits at specific moments in your workflows. The agents building your content can access it, ask questions, and learn while they’re writing the article.

Here’s how I structure mine. I use Supabase because it’s very easy to build and control schemas and tables through an LLM - but you can use any database, even a few simple Notion databases. Things don’t have to get complex if you don’t want them to. The most important part is keeping things structured.

Mine is organized around six schemas. Think of a schema as a folder that holds tables with one specific purpose:

• raw - the landing zone. Whatever the outside world sends: analytics, search data, calls. Nothing here is trusted or cleaned yet.

• content - the knowledge base. Everything mined from sources: brand facts, competitor facts, quotes, customer evidence, entities and their connections.

• rules - the judgment calls, stored as rows instead of buried in prompts so you can really enforce them: the taxonomies (pillars, clusters, personas, topics, entities etc), what counts as an AI source, who the competitors are, and the cleanup rules that filter bots and duplicates out of the analytics etc

• stage - the clean, combined mirror of all the data, ready to be used.

• ops - the watchdog. Nightly checks that flag when something breaks or goes stale.

• production - what gets surfaced: the windows into the data needed for reporting and for producing new content.

Multiple schemas keep permissions clean and make everything findable - every agent and every human knows exactly where a piece of knowledge lives, and nothing important is buried in someone’s prompt or memory.

Apart from analytics and mined knowledge, I also store my entire page inventory: every content page, mapped across the pillars, clusters, funnel stages, categories, and personas it’s meant for, and the entities and subtopics it’s connected to.

Each page is linked to a record of the facts used to create it, and to a record of all the internal links between pages. So at any point, I can ask an LLM: how many pages rely on this fact? This is what feeds the healing part of the engine loop - more on that in a minute.

To start small, forget Supabase. Start with a few Notion databases where you store individual brand facts, one at a time: one row is one fact. Keep them on point and simple, and add labels so everything is categorized. The test is simple: when you ask AI “how many brand facts do I have stored about this feature?” - it should be able to pull every single one. That’s what the labels are for.

Here’s the schema I use for brand facts which is a table in my content schema,

and how the visual UI listing those facts looks.

The schema is for the machines and agents, but as a human, you’re the one approving all the knowledge and making edits, and you need a visual way to do it. This is very easy to vibe-code: connect your tables and schema, and ask AI to build it for you. It can live as an artifact in Claude, or as a standalone dashboard on Vercel.

The Engine: Agents on a Cadence

The engine is a set of AI agents and workflows, each with one job, running on a cadence or triggered manually.

The mining is where the moat is, because every source gets you different assets.

Everybody mines sales calls for pains and objections - but you should also mine them for brand facts, because a founder selling articulates why the product exists better than any document. Store all of it as knowledge to use. I use 6 different tables to store and catalog knowledge, but you can design your own system of record.

Sales transcripts give you ICP needs that are fresh, brand facts coming directly from your founders or reps doing the demos, and topic knowledge as opinionated facts about the industry always surface here. Expert conversations give you how the people who build the category actually think and talk about it - I mine our podcast for those. RFP documents give you the technical nitty-gritty, plus what buyers literally asked for, in their words.

Every time I need ideas for my backlog, all of that knowledge tells me what’s relevant, what reinforces the entity I’m building, and what’s actually useful for the reader.

Everything is automated and grounded in real data - building the knowledge base, writing the articles, flagging what’s changed or gone stale.

Judgment stays human: approving facts, resolving contradictions, prioritizing what gets built.

Here’s the loop in short:

• Mine. Agents extract facts, quotes, pains, and topic knowledge from transcripts, RFPs, podcasts, and docs. New data gets checked against old - contradictions are flagged.

• Approve. A human gives every fact a go or no-go before anything can use it.

• Plan. A backlog agent proposes topics based on knowledge coverage and analytics - and only ones with enough approved facts behind them. No evidence, no article.

• Prioritize. A human picks what gets built this week.

• Write. A writer agent fetches the evidence, builds the brief, writes, and QAs against the facts in the database.

• Place + publish. Links, images, and embeds come from inventoried tables, not an editor’s memory. Publishing writes back which facts and links each page used.

• Heal. Retire a fact, and every page built on it gets flagged - down to the section. The refresh queue builds itself.

I run this on Supabase, AirOps, and Webflow, but the same loop runs on Notion, a spreadsheet, and any CMS. Minimum version: one mining pass a week over sales calls, one approval board, one backlog sheet, one writing workflow that’s only allowed to use approved facts.

Mission Control

This living system is human plus machine. But humans aren’t built to read databases with a million rows and fully understand what’s in there - so you need a UI to control everything.

I stole the Mission Control term from NASA, and I think it fits: the one place where all the information comes together, and where you control the engines and everything they do.

One note before the tour: for this to work across every computer - and as you scale your team - everything is synced in a Git repository. Every change to the architecture, every agent or skill you add, gets reflected there. One source of truth. Think of it as the documentation layer any software has, except it documents your content system.

Mission Control has three sections:

1. The Pulse - where you take the system’s vitals. The cleaned data from Google Analytics, Search Console, your CRM, and the AEO metrics from your visibility tool gets blended and transformed, so you can calculate your own metrics and build the tables, charts, and graphs on top. This becomes your reporting: what’s working, what’s not. It’s also where you visualize how your knowledge graph expands - where you have coverage, and where the gaps are.

2. The Brain - where you control what the system knows. All your stored facts and knowledge, and where you approve them.

3. The Launchpad - your production pipeline. Backlog, content plan, refresh queue: one view of what’s being assembled and prepped for launch, and what’s on the roadmap. So you always stay in control.

The Takeaway

Think of this system as a systematic approach to feeding and training your AI as your intern, one bit at a time (yes, pun intended). The more it knows about your business, the less it has to invent - and invention is where slop comes from.

It’s the same way humans learn.

A writer gets better the longer they’re at a company: they pick up brand facts, hear new ICP objections, absorb how the founder talks. The knowledge compounds, and the writing sharpens with it. This whole system mimics exactly that - except the knowledge never leaves when someone does.

And it’s built to evolve, not to be rebuilt. AI will make half of today’s tactics obsolete within a year; here, the layers stay and the workers get swapped. You evolve the content AND the system.

The content calendar told you what to publish. Mission Control tells you what’s true, what’s stale, and what’s missing. That’s the difference between running a content schedule and running a content system.

Adina designs and implements systems like this as head of AEO at Weflow, and builds them for B2B companies through Sighted, together with John. Want an audit of your own content system - or this one built for you? Reach out to hello@sightedagency.com

---

✅ Need ready-to-use GTM assets and AI prompts? Get the 100-Step GTM Checklist with proven website templates, sales decks, landing pages, outbound sequences, LinkedIn post frameworks, email sequences, and 20+ workshops you can immediately run with your team.

📘 New to GTM? Learn fundamentals. Get my best-selling GTM Strategist book that helped 9,500+ companies to go to market with confidence - frameworks and online course included.

📈 My latest course: AI-Powered LinkedIn Growth System teaches the exact system I use to generate 7M+ impressions a year and 70% of my B2B pipeline.

🏅 Are you in charge of GTM and responsible for leading others? Grab the GTM Masterclass (6 hours of training, end-to-end GTM explained on examples, guided workshops) to get your team up and running in no time.

🤝 Want to work together? ⏩ Check out the options and let me know how we can join forces.