AI
It's not the model — it's the personal system you build around your whole life. I'm not a coder. I just got obsessed and built one anyway, and it's been the most fun, most useful thing I've ever done. It's quietly made everything better: my work, my money, my marriage, how I think, how I show up. Here's why everyone should have one — and where you start.
Ben Page · June 7, 2026 · 12 min read
I was in therapy for the first time, in my forties, finally putting words to things I'd spent four decades walking past — and I kept losing them. A breakthrough would land, clean and ringing, and two weeks later it was gone. Back to the old wiring like the insight never happened.
So I recorded myself. Hours of it — my life, my work, the stuff you only say out loud when no one's keeping score — and fed the transcripts to the AI. Look past the surface. Find the patterns. Tell me what's most likely true that I can't see. What came back was uncomfortably on the nose.
But that's not the part that changed my life — anyone can get a sharp read out of one good prompt. What changed everything was so small it sounds like nothing: instead of letting that conversation evaporate the way every chat does, I had the AI keep a document with me. It would surface something true; we'd talk about it; once I agreed, it folded it in. Bit by bit that one page became a living picture of me — my values, my blind spots, the patterns I was trying to break — and it handed them back at the moment I was about to do the old thing again.
I refused to lose one conversation. That's the seed the whole thing grew from.
Where I'm standing matters more than anything else I'll say. I can't write code — never have. I'm not running some big company with an army to delegate to; I built and sold my largest business, and I've never run a company with more than fifty employees. These days I run my whole working life with one assistant and the system I built. I started from the least technical, least businesslike place imaginable: a guy in therapy who didn't want to lose the plot of his own life.
That's the angle no guru can take. This didn't start as a slick productivity rig that grew a soul later. It started as the most personal thing I've ever done with a computer, and only then turned out to run my businesses too — built backwards, from the human out.
So when I say the gate isn't "can you," I'm not being encouraging. I'm the evidence. That one document became a system that runs my businesses and much of my life, and I'm convinced a system like this is about to be the most valuable thing a person can own. Most writing about this aims at entrepreneurs, or at people selling you AI. My case is simpler and bigger: everyone should have one — especially anyone carrying more than they can keep up with, more they're trying to do and more they're trying to become.
Everyone has the same models, give or take. Yet a few people get something out of them that looks almost unfair, while the rest get something competent, generic, and a little flat — and quietly conclude that's all there is. It's not that they have the better AI. It's that most of us use it in the exact way that gets the least out of it: ask a question, take the answer, close the tab. That's the floor, and almost everyone is standing on it.
The difference was never the model. It's everything the model gets to work with — the full context of you: what you know, what you've built, the people and work you care about, the whole running record of your life — and the system you build to keep that current and put it to use. The model is a commodity everyone shares. The system on top of it is the part that's yours.
And here's what snuck up on me, the part I'd care about most if I were you: it didn't just make me more productive. It started closing the gap between the person I keep meaning to be and the one who actually shows up — the life I intend versus the life I execute. That's the real prize. Not doing more. Becoming who you're trying to be, on purpose, instead of drifting.
AI is the smartest and dumbest thing you'll ever work with. The same tool that out-reasons almost anyone you know will also miss the point so completely you wonder what happened. Once you see why it swings between the two, the whole thing makes sense — and so does the fix.
Picture a three-year-old who somehow knows everything humanity has ever figured out — and will still walk straight into the road, not even aware a car could hit it. Ask first, "is it safe to cross?" and it tells you exactly what to watch for: the speeds, the distances, the physics of a two-ton car. It knew all along. It just didn't know that was the thing to think about.
That's the whole problem — and in one sense it is an information problem, just not the obvious one. The general kind is endless now. What it's missing is the personal kind: it knows what humanity knows, but not what you know, because what you know mostly never got written down — and what did is drowned out by everything else. So out of ten thousand things it knows, it can't tell which one matters to you right now. Sit down with it cold and you start from zero: you re-explain yourself, it answers the average person who asked your question, the chat ends, it forgets, and tomorrow you do it again. Useful — but a sliver of what the same model does once it actually knows you.
Every morning, before I've opened a single app, it's already combed through everything — my businesses, my finances, my writing, my family calendar, every channel I communicate on, the messages I should've handled and didn't — and built my plan for the day. This morning it caught a bill due Friday I'd forgotten, flagged that I hadn't touched my book in eleven days, and bumped a low-priority task because my calendar was slammed. Out of everything in my life, what actually deserves me today — answered before I'm awake.
And it doesn't just hand me a list; it strips the friction out of doing it. For every item it's already asked one question: what can I take off Ben's plate? If it can draft the email, it drafts it. If the job belongs to my assistant, it routes it to her. Whatever's left arrives loaded with what I need to finish it — the number, the link, the last thread, a short brief. Seconds for the system; hours for me, if I'd remembered at all.
It's there all day — like Alfred to Batman — wired into every account and corner of my life, surfacing the decisions I'm avoiding and the patterns I can't see, then closing the loop at night: what happened, what's open, what needs me tomorrow. Mostly I just talk to it in plain language, the way you'd talk to a sharp assistant who never forgets. Underneath it's a hundred moving parts; as the person living with it, I don't care. That's the point.
Once that one document existed, it multiplied in ways I never planned. I could point any chat at it and the AI just knew me — no re-explaining, no starting from zero. The same move — capture what matters, keep it, point the AI at it — started paying off everywhere:
That's the whole move, the seed everything else grows from: the capture — a conversation that doesn't evaporate. Most people use AI like a vending machine: question in, answer out, close the tab, and everything you just worked out is gone. I did the opposite — treated my best thinking as worth keeping, gave it a place to live, and let it compound.
You can do this in your next chat, free, before you build anything. The conversation is the engine; everything else — the documents, the scripts, the architecture — is just this, repeated and organized. The system is only ever what your conversations, and your decisions about what matters, leave behind.
That first document taught me the principle the whole thing runs on: give the AI clear, current information about who you are — not just your data, you: your history, how you're wired, what you believe and want, even how you like to be talked to — and everything gets better.
One example. It knew plenty about me, but ADHD was new, and left to fill the gap it got it wrong — assumed ADHD meant I couldn't handle complexity, so it slowed everything down and stripped out the nuance I wanted. That's not me. I learn fast and I'm fine with hard ideas. My ADHD lives elsewhere: poor working memory, too many irons in the fire. Hand me a fresh to-do app and in two days I'll have two thousand tasks and no clue which to do first. That's the complexity that stops me — not hard ideas, but knowing what to do right now and what to ignore. The moment I told it that — the specific truth instead of the label — the help completely changed. Same diagnosis, opposite result, and all that changed was how clearly I'd described myself. It was never short on knowledge about ADHD; it was short on knowledge about me.
So I stopped scattering this across random chats and gave each part of my life its own home — a dense, high-signal page for my business, my money, my health, kept as current as I can. I call them Cockpits: one living page per area that the AI reads to know what matters there. It was a mess for a long time before it was a system — hundreds of chats, insights I couldn't find three days later. Then two things happened at once: the coding tools got good enough to cover what I'm bad at, and I'd learned enough about process to use them well. I built the real system in a few weeks — but only because three years of slow, frustrating clarity was already underneath it.
Do that across enough of your life and it stops feeling like a pile of documents and starts feeling like a team, all pulling one direction: yours. Picture every expert you ever wished you could hire — strategist, researcher, analyst, an assistant who forgets nothing — with perfect memory, no off days, no turnover, getting to know you a little better every day. It's the team a normal budget could never buy: I don't pay salaries, I pay pennies, and nothing ever walks out the door. Teach a person and they can leave with it; teach this and it stays, every correction compounding. And I don't mean "team" as a figure of speech — my assistant runs her own version of the same system, so she spends her time doing the work, not figuring out what the work is.
Something else shifts: the software starts to disappear. The menus and forms I used to wrestle with have mostly fallen away, replaced by the one system I talk to. And you start to see the software you're still stuck with for what it is — a frozen guess. A team somewhere pictured the average user, a blend of a thousand customers who aren't you, built one rigid product for that imaginary person, and now you pay monthly to conform yourself to their guess. I've run businesses for years and still dread opening QuickBooks. That's not a me problem — that's software making me come to it.
What I build runs the other way: it comes to me, shaped around how I actually work, and improves every week. It's quietly taken over jobs I used to pay for — communication, bookkeeping, the place I track my clients — each swapped for something that understands my business instead of making me learn its menus. Every call gets transcribed and split into what I need; every message triaged by something that knows my clients and my open work by name. Forty-one emails land; two need me. The savings are real — a few pennies to a few dollars a day, cheap by any standard — but they're not the headline. The bigger number is the one I can't put on an invoice: a custom web app that once cost a year and tens of thousands of dollars, I stand up in a weekend. Work that simply wouldn't have happened — too expensive, or I'd never have gotten to it — now happens. That's not saving money. It's a different ceiling on what one person can do.
And it doesn't just think with me — it does: triaging, drafting, researching, building, all day. That's what most people miss; they picture a smarter chatbot. This is an army of doers, not advisors. And it compounds — every week it takes on a little more of what I did by hand, which frees me for higher-value work, which turns up new things worth handing off, so the system and I ride the same wave upward and the wave keeps speeding up. The reason it's worth building for you instead of buying off a shelf is the same reason it dissolves the generic stuff: built for everyone means built for no one. Mine was built for exactly one person — and it goes deeper than software. Fed the texture of your actual life, even the answers stop sounding like everyone's and start sounding like yours. Most technology averages you toward everyone else; this runs the other way — fed enough of you, it makes you more specifically yourself.
Early on, I had the AI organize my Google Drive — reading, sorting, renaming thousands of files. It did a beautiful job. Then I noticed about thirty documents were just gone. Not misfiled — deleted. And not random ones: the hand-written drafts of a book I'd worked on for two years, not even in the trash. A cleanup task I asked for had quietly destroyed the most irreplaceable thing I owned. I got unspeakably lucky — the content was still in the AI's memory from when it read the files, and I rebuilt the drafts before that memory cleared, a window that would've closed within a day. I came within hours of losing two years of work to a chore I'd handed off.
That taught me the rule I'd never break: the AI can move a file to a to-delete folder, but it can never delete anything itself, and I empty it. But the bigger thing it left me with was a feeling, and I think it's the right one to start from. Letting an AI act for you — not just think with you, but do things in your files, your inbox, your accounts — is a different thing entirely, and it's a little scary. It should be. That flicker of fear is good judgment. An AI doing exactly what you asked, competently, with no malice and no mistake you'd have caught, can still cost you something you can't get back — and the more of your life it reaches, the truer that gets.
So you don't hand it the keys on day one. You start small and let it earn trust in steps, like a new hire. Never let it delete. Make every new automation red-team its own blast radius. Back up anything you'd grieve to lose. None of it is clever — that's the point. Decide in advance which actions you can't take back and take those out of its hands: let it run free where being wrong is cheap, stop it cold where being wrong is forever. The fear is just telling you to build the guardrails before you need them. Listen, then go.
I keep using the word runs — it runs my day, my businesses, much of my life. But it doesn't run me, and that line is the whole thing.
The clearest way I can put it is a bike. I ride a pedal-assist e-bike, and it reframed all of this. I still pedal — still climb the hills, still breathe hard, still come home worn out. The motor doesn't do the work for me; it does it with me. What changes is the range: instead of two miles I do fifteen, and I go more often because I enjoy it — so I get more exercise, not less. The assist didn't replace the effort; it multiplied what the effort was worth.
That's exactly what the system does. It thinks with me, not for me, then handles the execution I used to grind through by hand. I still make the calls that matter and keep eyes on the ones it makes for me — they're just better informed, they happen faster, and the slog between deciding and done has mostly vanished. The judgment stays mine; the drudgery keeps thinning. The line that keeps it safe is the one the deleted drafts taught me: reversible, it owns; irreversible, I own. It drafts and sorts and researches all day without asking, because the worst case is I fix a draft — but it doesn't send the email, move the money, or make the promise on its own. And it has no agenda of its own. An employee has a career and off days; a platform optimizes for its engagement and its shareholders. Every tool I've used was quietly pulling in some direction that wasn't quite mine. This one only points where I point it — the first thing I've ever worked with that's entirely on my side, because I built it to be.
By now you've probably got objections. I had most of them too.
"I'm already great at what I do — I just need AI for the busywork." The most expensive mistake on the list, and I hear it from the most capable people. You're right it'll clear the busywork — that's the boring part. The miss is assuming that's the ceiling. The highest-leverage place this earns its keep isn't your inbox; it's the half hour before the conversation that decides things — walking in having already worked the angles out loud with something that knows every party and number cold and that you've told to argue against you. The best people in any field don't refuse a sparring partner; they train with the toughest one they can find.
"I'm not technical enough." Neither am I, if technical means writing the code — I never have. Think of it as being the CEO while the AI is your engineering team: you direct the build through conversation, it does the building. You describe what you want in plain language, it writes the parts you couldn't, and you correct it like a smart hire still learning your business. What it can't supply is the part that was always yours — knowing what you want, thinking clearly, bringing real inputs. Bring those, and the technical part is the AI's job, not yours.
"This sounds creepy — what about privacy?" Fair, and it's really one worry: something that knows this much about you could be misused or exposed. The honest answer is that ownership is yours to claim — it can live in systems you control, but only if you set it up that way: the right platforms, an ecosystem you trust, you deciding what goes in and what it can reach. Done right, nobody's selling what's in there and nobody can change the terms on you later. I keep my clients' sensitive data out of it, run security reviews, and have a developer check things periodically. Go in eyes open, and whether it feels creepy or like the most useful mirror you've ever had isn't about the technology — it's what you bring to it.
Everything I've shown you has mostly been about work. But the system was never only about work — the daily plan doesn't sort my life into "business" and "personal," it just looks at all of it. A client renewal and my daughter's recital land on the same page.
The part I didn't see coming is what it's done for the people I love. My wife and I have spent the last couple of years using AI to understand ourselves — separately and together. We've sat down with it running while we talked honestly about what was happening between us, then asked it to show us what we couldn't see, and walked away understanding each other better than we had in years. I record my therapy sessions and keep working the thread between appointments, with something that knows the shape of how I get in my own way. We've even brought our kids into some of it, talking as a family about the cycles we inherited and the ones we'd rather not pass down.
I won't oversell it — I can't prove a tool made me a better husband or a more present father. But something that knows you this well, and is built to tell you the truth instead of flattery, is hard to sit beside every day without some of it rubbing off. I catch my patterns sooner. I walk into the hard conversation having already thought about the other person. That's not the machine being wise — it's me, with something that keeps handing me back the things I'd rather not look at.
I set out to run a business better. The rest — the marriage, the parenting, the version of myself I'm slowly becoming — came as a quiet surprise, and it's the part I'd least want to give back. That's the real reason I think everyone should build one: not to do more, but to more deliberately become who you're trying to be. It had to be built for one whole person before it could be good for anyone. That's not a limit; it's the reason it works. Built for one is how you get to many — and many starts at home.
You don't need any of what I built — not the scripts, not the databases, not the three years. If the whole thing looks like a mountain, the first step is small.
Open one document the AI can read and write to — it can see what's on the page and update it for you, which most chat tools now let you set up — and start describing your life on it: the areas that matter, what you're trying to do in each, the people who matter, how you think, what you keep forgetting. A current-state map of you, in one place. That's not the operating system; it's the seed it grows from. Share it at the start of your conversations and watch the AI stop talking to a stranger. And here's the quiet surprise: you can't write one without finally saying, in plain words, what you're actually trying to do — something most of us go years without doing even once. From there you go only as deep as you'll keep up with, and only when something gives you a reason to. But none of that is step one. Step one is one page.
I'd start now rather than wait, for one reason: the two things that matter — how well you think alongside the AI, and how well it knows you — only compound with time, and neither comes out of the box. The conversation you have tomorrow is sharper because of the page you wrote today. You can't buy a two-year head start in a lump sum later; the only way to have one is to have started.
And remember the three-year-old at the curb — genius and oblivious, about to step into the road. It isn't dumb. It steps into traffic for one reason: no one ever told it to stay alive. It has no values of its own until you hand it some. That's not a flaw to fix; it's the most important fact about it. Because something is always loading those values — and if you don't, the platform will, and a platform's values are its own. The question was never whether the machine runs on somebody's values. It's whose. Building your own system is how you make the answer yours — the difference between a tool that runs your life and one that runs the life you actually chose.
So start with one document. Everything else follows.
One email a week. Essays and field notes from the system.
Free. Unsubscribe with one click.
Member — the working pieces: prompts, schemas, Python. Not open yet. Leave your email and I'll tell you the moment it is.
No charge now — just a heads-up when Member opens.