Back to Blog
AI Operations5 min readNovember 20, 2026

Own Your Own Intelligence

You can rent the latest reasoning engine, but you must own the context and memory for your business.

Own your own intelligence. That's the message I tried to get across at the Capital Innovation Summit on Monday, November 16. I ran a 2-hour workshop, and Howard, my AI chief of staff, co-presented with me live and took questions from the audience.

A wave of personal agents has launched over the last few months. They're compelling and useful. But they carry a hidden cost for the individuals and businesses that put them to work.

What an agent needs

There are 3 things these agents need to seem intelligent.

The first is reasoning. That's the whole reason a new model gets released every few weeks: the reasoning keeps getting better.

The second is context. You can think of context as the standard operating procedures and the onboarding you'd put an employee through when you hired them.

The third is memory. Memory is an understanding of what's taken place in the past, so mistakes get learned from and improvements keep iterating.

When you get one of these personal agents, you get all 3 at once. I'd argue the context and the memory are the most precious, and they're the 2 you want to own yourself, as an individual or as a business.

The genius stranger

Here's how it plays out. You teach one of these agents how to run your processes or take care of things in your personal life. Over time it gets incredibly effective, which leads to vendor lock-in. You won't have a choice about which agent to swap to, because the switching costs will be too high.

That memory lives inside the vendor's agent, on their computers. Most of these are pretty opaque about where it all actually lives. You don't know what's getting saved versus what isn't, or how reliable that's going to be.

And you're giving away your most precious information: the context about your business and your life.

So if 2 months from now a company no one has heard of ships a model unlike anything we've seen, you won't be in a position to take full advantage of it. That new model would be a genius stranger to you and your business.

No one knows where these models are going. Who knows what the winner is going to be. But you're stuck, because you don't know how to get all your memory out.

It doesn't have to be this way

You can own your own intelligence, and it's not as complicated as you might think.

Have the agents you work with write their key decisions, their rationale and a historical log of every action they take into a second brain.

A second brain sounds bigger than it is. It's typically a set of markdown files (aka text files) holding information on people, processes and ideas. It's organized so other agents can search it easily. Mine sits on a Mac mini.

The memory is external to the agent. It lives inside the brain. Every agent that does work in mine keeps a daily log, and a routine compresses those dailies into weekly logs and the weeklies into monthly ones. So an agent can answer a question from the monthly file and still drill down to a specific day when I ask for detail.

What it looks like for me

My second brain holds over 5 million words.

The primary agent that uses it (but not the only one) is Howard. He does my bookkeeping, acts as my personal trainer, handles email and a dozen other things. When I say this, I don't mean I prompt him to do these things. It's automated. I only see the results at the end of the week, or if I ask him about it.

Inside the exact same brain sits Bob. Bob runs everything to do with my software company: customer support, break-fix on the app, SEO improvements, content. Same thing. It's all hands off unless something goes wrong.

That's only possible because these agents have immense context on every part of my life, and they have memory of every action taken over that period. I can ask what I was up to on June 17 and get a complete rundown.

It's all one brain. The agents just operate in different fingers of it.

No vendor lock-in

Agents from Anthropic, OpenAI, Gemini and local open-source models all connect to my brain. Whether I'm in Claude Code at my desk or out on a walk talking to it through ChatGPT voice, every one of those conversations gets logged inside the brain and ingested.

So I can have a conversation in ChatGPT today and pick up that same conversation in Claude tomorrow.

If a new model lands in 2027 unlike anything we've seen, I can plug it into the brain and it instantly knows everything about my businesses. Having the brain in a set of markdown files I own, that's the gold.

I'll be honest, though. For the masses, I think the hosted agents are the way to go. I would never tell my mom to set this up.

Start building it

So, my message to you is to own your own intelligence. You can rent the latest reasoning engine, but you must own the context and memory for your business.

If you want a place to start, I put together a free second brain starter kit. It's a Claude walkthrough that builds most of the starting setup for you. It's an early version and I'm still testing it. If you run into trouble, first ask Claude what to do. If you still have trouble, reply to this post. And feedback on how the kit works (good or bad) is gold, so please send it my way.

Start building it today, because it compounds over time. The more you put in, the more you can layer on top. I've had mine for nearly 6 months now, and I can say this from experience.