// BUILD IN PUBLIC
Building in the open.
I run a marketing operation that runs on AI. Here's the system — the frameworks, the infrastructure, and the working receipts. Not theory; the thing I actually operate.
// FRAMEWORKS
Named models, used in production
Reusable frameworks I've published and run — the intellectual capital, not hot takes.
The AI-native maturity ladder
Tools → Assistants → Colleagues → Operating system. Where a team actually sits, and why most stall on rung two.
read →The compounding test
Three questions that separate AI that moves a number from AI that just demos well: does it touch a real system, remember, and improve?
read →Govern the blast radius, not the intelligence
Governance as accelerator. Control what the AI can touch, not what it can think — so you can safely say yes to autonomy.
read →Architecture over headcount
When the marginal cost of codified work approaches zero, leverage stops scaling with people and starts scaling with the operating layer.
read →Operations Intelligence Architecture
The OIA framework for finding what's hidden and fixing what's broken — five pillars, a structured audit.
read →The SCU framework
See. Clarify. Use. Turning operational complexity into measurable improvement.
read →// THE OPERATING SYSTEM
Four layers that mean it runs on AI
The difference between an operation that runs on AI and one with AI bolted on. None of them are a model — all of them are architecture.
Persistent memory
Context compounds across sessions. I'm not briefing it — it's briefing me.
Hands on the real stack
Agents wired into live tools, not stuck in a chat window. Work happens where the systems are.
Codified workflows
Written down once, improving while I sleep. The operating layer is the asset.
Gated autonomy
Unsupervised work writes to a staging surface a human promotes to live. Autonomy real, blast radius zero.
// AI-SEARCH VISIBILITY
Built to be cited, not just ranked
Content and infrastructure architected so frontier LLMs resolve and cite the work. This site is the proof case.
Frameworks as entities
Named models marked up as DefinedTerm + CreativeWork, authored — so models cite the concept, not just the page.
Sourced citations in schema
Every essay's sources are machine-readable in Article.citation[] — attributable, dated, never fabricated.
One entity graph
Person / Organization / WebSite share stable @ids — one canonical entity, not duplicate nodes.
Answer-first + AI-crawler open
Self-contained answers up top, llms.txt index, AI crawlers explicitly welcomed.
// THE CONTENT ENGINE
Write once, distribute everywhere
An owned hub with rented mechanics — the data stays home, the posting is leverage.
Atomisation
One essay → newsletter + LinkedIn + X variants, each in voice, review-gated before anything ships.
Owned audience
Own the list and the attribution; rent only the posting surface. No platform lock-in.
UTM attribution
Every link tagged to source / campaign / variant — measured in analytics, no platform API needed.
The people worth listening to are running the thing they describe.