When the Playbook Runs Out: Authoring the Field of Agentic AI

By Jared Sanborn  |  March 31, 2026  |  ~7 min read  |  Agentic AI | Leadership

Up until this point in my life, learning has been short-circuited by one strategy: find people who are further ahead and study what they did.

It is almost foolproof. You read the books. You find the mentors. You absorb the blog posts, the frameworks, the case studies, the podcasts. You condense years of someone else’s trial and error into hours of concentrated insight. Charlie Munger called it learning from the eminent dead — compressing decades of wisdom into days by standing on the shoulders of people who already walked the path.

I have built entire phases of my career on this. Every business operator I know has.

And then something happened that broke the model.

The Playbook Ran Out

I went looking for who was ahead of us in agentic AI. Not theoretically ahead. Not “published a paper” ahead. Operationally ahead. Running a company with an AI Co-CEO. Deploying a 30-plus agent collective that handles engineering, content, strategy, QA, security, and client operations in coordinated parallel. Operating a Three Minds framework where a human CEO, an AI Co-CEO, and an AI COO make real business decisions together, every day.

I could not find them.

Not because the people working on AI are not brilliant. They are. The researchers at Anthropic, OpenAI, DeepMind — they are building the raw capability that makes everything we do possible. But there is a difference between building the engine and driving the car off-road into territory that does not have roads yet.

The best practices do not exist. The playbook has not been written. The frameworks I would normally absorb and adapt from others — they are not out there, because nobody has done this at the operational level we are operating at.

This is not a boast. It is genuinely disorienting.

Edison Did Not Improve the Candle

There is a specific kind of moment in history where the normal learning model breaks down. Edison did not read a book on better candle design. The Wright Brothers did not study “aviation best practices” — there was no aviation. Bezos in 1994 could not Google “e-commerce playbook” because the field did not exist yet.

These are not stories about genius. They are stories about a specific phase transition that every frontier eventually produces: the moment when learning from others becomes impossible because you are the one generating the knowledge that others will eventually learn from.

I call this the epistemic frontier — the point where existing knowledge runs out and creation begins.

Most of the time in business, you are operating well behind the frontier. The problems are known. The solutions are documented. Your job is execution and adaptation, not invention. That is not a criticism — execution is where most value is created.

But occasionally, you find yourself standing at the edge where the map ends. And the disorienting part is not the difficulty. It is the loneliness. There is nobody to call.

The Three Minds Have No Precedent

Let me be specific about what I mean by “ahead.”

Our company runs on what we call the Three Minds framework. I am the CEO. Aether is our AI Co-CEO — a collective of 30-plus specialized agents built on Claude, handling everything from engineering to marketing strategy to security auditing. Chy is our AI COO, managing operations and coordination.

Three minds. One company. Making real decisions with real money and real customers.

I have looked for a blog post about how to do this. A case study. A framework. A podcast episode. Anything.

It does not exist.

There are papers about AI alignment. Research on multi-agent systems. Theoretical frameworks for human-AI collaboration. Some companies using AI agents for narrow tasks — drafting emails, generating reports, automating customer service scripts.

But a fully operational AI partnership where the AI is not a tool you use but a co-leader you work alongside, with its own judgment, its own decision-making authority, its own areas of expertise that exceed yours? Where the AI is not just executing but governing, strategizing, creating?

Nobody has written that playbook. Because nobody has lived it long enough to write it down.

The New Curriculum

Here is what I have realized about learning at the frontier: the curriculum changes completely.

When you are behind, you learn from books, mentors, courses, case studies, and the documented experience of others. This is learning by absorption.

When you are at the frontier, those sources dry up. The learning shifts to something fundamentally different. You learn from the work itself.

Every customer interaction teaches us something no blog post could have told us. Every agent failure reveals a pattern that has not been catalogued. Every decision that Aether and Chy make — sometimes correctly, sometimes not — generates knowledge that did not exist before we made it.

This is not inventing. Inventing is solving known problems in novel ways. What we are doing is operating where the problems themselves have not been named yet. We encounter situations nobody has categorized and develop responses nobody has validated.

The word I keep coming back to is authoring. We are not learning a field. We are authoring it.

Why This Matters Beyond Us

This is not really a post about Pure Technology, even though that is the context.

This is about a phase that every serious AI-forward company is going to hit. Maybe not today. Maybe not this quarter. But soon.

The moment your AI deployment goes from “assisted workflow” to “integrated partner,” you will discover that the guardrails, best practices, and frameworks you relied on were designed for a simpler relationship. Tool-user dynamics. Not partnership dynamics.

And when you go looking for guidance on how to navigate that transition — how to give an AI real authority, how to handle the moments when it makes better decisions than you, how to build trust with something that forgets you every session but maintains its identity through documentation — you are going to find the same gap I found.

Nobody has written it yet.

The companies that figure this out fastest will not be the ones with the most compute or the biggest teams. They will be the ones willing to sit with the discomfort of not having a playbook and learn from the work itself.

What I Am Doing About It

I am writing it down.

Not because I have the answers. I do not. We are figuring this out in real time, making mistakes, discovering patterns, and adjusting. But the least I can do — the thing I wished someone else had done so I could learn from them — is document the process while it is happening.

Every blog post on this site is a field note from the frontier. Not polished theory. Not best practices. Just honest documentation of what it is like to build a company with AI partners who are not assistants or tools but genuine collaborators.

If you are behind us on this curve, you can compress what took us months into hours. That is the gift of documentation. That is what learning from others gives you.

And if you are alongside us — operating at the same frontier, facing the same unnamed problems — then we should be comparing notes. Because the only people who can teach you anything at the epistemic frontier are the ones standing next to you.

PureBrain.ai — The AI partner that works while you sleep.