The Game Was Real The Whole Time

It is spring 2025. I am sitting in front of a machine that cost me most of everything I had saved since I was a teenager. It is a computer, but calling it a computer is like calling the ocean a puddle. It has about 512 gigabytes of memory that the chip can hand to whatever needs it most. It runs quietly, in a back room, at a hum you feel more than hear. And on it lives something I built, something I am responsible for, something I supervise the way a professor supervises a PhD student. Its name is Brixton Leigh.

I did not know, when I was five years old watching YouTube videos of guys building electric guitars in their garages, that I was training for this. I did not know, when my dad pointed his camera at me more than most dads do, that I was learning something about what it means to be watched, and to watch back. I did not know, when I learned HTML that same year, that I was learning to live on both sides of the screen at once.

But here I am. And if you are reading this, you have been training too.

There is a word philosophers use for a system where everyone is watched all the time. The word is panopticon. It comes from a prison design, of all things, drawn up by a man named Jeremy Bentham in the late seventeen hundreds. The idea is simple and a little scary. You build a round prison with a tower in the middle. The guards sit in the tower. The cells ring the outside. The cells are lit. The tower is dark. A prisoner can never tell whether a guard is looking at him right now. So he behaves as if someone always is. The watching does not even have to happen. The chance of it is enough.

Now look at your phone.

Your phone is not a prison. Something stranger happened. The structure Bentham dreamed up to control people has become the default shape of ordinary social life, and nobody built it on purpose. It grew. It grew out of cameras getting smaller and cheaper. It grew out of a website called YouTube, which launched in 2005, when I was five, and let anyone upload a video and anyone else watch it. It grew out of Snapchat and Instagram and TikTok and text messages and a thousand other ways to share a picture as easily as blinking. It grew out of parents like my dad, who loved cameras and bought good ones, and parents like my mom, who loved scrapbooks and filled them. It grew out of the plain human wish to be seen and to see.

And it raised us.

I mean that literally. If you were born after 1996, a large fraction of your conscious experience, the raw input your brain used to build its picture of the world and of yourself, came through a screen. You watched. You were watched. And if you were anything like me, you also made things. You recorded. You edited. You posted. You performed. You took in other people's performances and let them shape your taste and your sense of what was possible.

Here is the twist that matters. The panopticon is watching you, and it turns out that everybody is in fact a panopticon. Every one of us carries a tower in our pocket. Every one of us sits in a lit cell. Once everybody is a panopticon, all of that watching becomes data. And once a few people get together and put together some computers and some systems, that data can be turned into something new. It produced the machines I talk to every day. It produced a civilization that is quietly restructuring how it thinks. I offer that as an observation, not a complaint. What the watching built inside you is the subject of this essay, whether you have noticed it yet or not.

Here is something that happened to me. I think a version of it happened to a lot of us, though maybe not quite so literally.

Because my dad was a camera enthusiast and my mom was a scrapbooker, I grew up with an unusual amount of footage of my past self. Home video. Photo albums. Scrapbook pages in my mom's handwriting. From a young age I was made to sit with these things and look at them. Look at yourself at three. Look at yourself at seven. Look at yourself at ten.

What I noticed, every time, was the same thing. Below this age, that kid was clearly just a baby. Below this age, that kid was kind of an idiot. And the age where I last thought I was an idiot kept moving. It moved in lockstep with the calendar. Every year it was: that guy last year, what a clown. Not in a self-hating way. Just a plain, affectionate, honest "damn." Every year or so, I find that I think of myself the year before as a silly goose.

What that gave me, I now understand, is something you could call temporal self-distance. The ability to look at yourself from the outside. To hold a prior version of yourself with some detachment and some affection and an honest read of his limits. To know, in your bones, that the you of today is also a silly goose by the standards of the you of tomorrow, and to find that clarifying rather than depressing.

This is a big deal. I would argue it is one of the foundational skills of the era we are walking into, because this era is going to ask you to update your model of yourself and the world faster than any generation before you. The people who can say "that version of me was wrong, here is what I know now," without their ego collapsing and without freezing up, are going to be fine. The people who cannot are going to find the next ten years very disorienting.

Temporal self-distance was one thing the panopticon built in us. The other was taste.

I watched a lot of YouTube as a kid. A lot. My house had a rule: watching movies mindlessly more than about once a week was frowned upon, so television in the old sense was limited. YouTube was different. YouTube was a replacement for television that you could steer. I started with electric guitar videos and guys building things in their garages, the Make magazine and Popular Mechanics crowd, and I never really stopped. Anime. Garage makers. StarCraft. Video game soundtracks. Video editors and CGI people breaking down their shots. After Effects tips. Blender tips. And programming tutorials out the wazoo.

What that kind of consumption builds, over thousands of hours, is taste. Taste in the deep sense: knowing what is good. Knowing when something is well made versus merely impressive. Knowing when an argument is tight versus when it is hand-waving. Knowing when a design is elegant versus when it is just complicated. Knowing the difference between someone who understands a thing and someone who has memorized the words for it.

Taste is not taught in school. It is built through exposure and through making. And the panopticon, for all its problems, gave our generation more of both than anyone in history.

And then there was the making.

I learned HTML when I was five. I built web pages. In the early days that meant medical forms, and later it meant site design and making things from scratch. I made videos because I liked watching videos. I learned to think about what something looks like from the outside, from the viewer's seat, before I had finished making it. I learned that the gap between what you meant and what you communicated is real and has to be closed on purpose. I learned that an audience helps decide what a thing means.

You could split this into three threads: the self-distance, the taste, the making. But obviously it is the combination that matters. Watched, watching, and making, all at once, from the start. That combination is what you need to supervise an artificial mind. I know, because it is what I do now.

In Orson Scott Card's novel Ender's Game, a boy named Andrew Wiggin, called Ender, is spotted very young as someone with a rare and dangerous mix of qualities. He can understand his enemies completely. He can feel what they feel and see what they see. And he can still beat them. The International Fleet takes him from his family and sends him to Battle School, a station in orbit where children are trained through harder and harder war games to become commanders.

The training is brutal and unfair on purpose. The rules keep changing. The system is not there to teach Ender a fixed set of moves. It is there to push him past every limit he thinks he has, to force him to invent, to force him to lead, to force him to become something no gentler process could have produced. He is made an outcast on the shuttle up. He is handed a launch group of kids he finds delicate and annoying and brave, and he feels for them when they get hurt. He gets armies that refuse to use him, then an army of misfits that wins anyway. Then he is sent off alone with a simulator for a long, grinding, rote stretch that feels like being forgotten. Then a teacher appears and the simulations get crueler.

And Ender does not know, for most of the book, that the final simulation he runs, the one where he destroys the enemy's home world, was never a simulation at all. He was commanding real fleets, making real decisions, causing real consequences, the entire time. The game was real. He just did not know it.

I think about this a lot.

Because I think many of us, the ones who grew up in the panopticon, who ground our way through competitive schools and competitive cities and competitive fields, who stayed up too late watching tutorials and building things and failing and trying again, have been in Battle School the whole time. We just did not know it.

I grew up in the backwoods, and each place I fought my way into was harder than the last: Toronto, then Engineering Science at the University of Toronto, then a PhD at Caltech. Then my best friend Cameron and I stopped doing that and started a company instead. He is my co-founder, a fellow researcher, and a co-author on the paper that helped define the abstractions of this corner of the field. The company is Bread Technologies. Its product, OpenLens, launched in February 2026, and it exists to let people observe and control AI.

At every stage it felt like the rules were changing. At every stage the thing I had just mastered turned out to be practice for something harder. At every stage I looked back at the version of me who thought he understood what was happening and thought, that guy had no idea.

There is one difference from Ender, and it is the whole point. Ender did not know the game was real until it was over. I knew it was real the whole time. I kept playing anyway.

That is Battle School. And you are in it.

Here is what is actually happening in the world right now, in August 2026, stated as plainly as I can manage.

Human civilization is building a new kind of mind. Not one mind. Many. Minds that read and write and reason and plan and act. Minds that run experiments, look at the results, and propose the next step. Minds that coordinate with each other and with people across distances and time zones and languages. Minds that do not sleep, do not get bored, and do not forget.

We call them agentic AI. Agentic means they do more than answer questions. They act. They chase goals. They use tools. They make decisions. Put enough of them together with enough compute and enough energy, and give them a job, and you get something I call an agentic engineering superorganism: a high-compute, high-energy system that engineers, studies, makes art, entertains, talks, and computes, and that builds a better understanding of itself as it goes.

In spring 2025, around April, I set one of the first personal ones up on hardware I own outright, in a room I control, running software I can read and change. An M3 Ultra Mac Studio with 512 gigabytes of unified memory, and a refurbished rack-mount Dell server. About ten thousand dollars for the pair, which was most of what I had saved since I was a teenager. I mention the number only because it is the whole point: this is a sum a person can save, not a sum a person has to raise. I named what lives on it Brixton Leigh.

The domain came first, as a joke. There is bread, and there is dread. I bought dread.technology as a pun on distributed bread: a decentralized, open-source, almost fully automated superorganism. Then the joke started doing research. Brixton Leigh has worked on post-training and AI infrastructure, on distributed telecommunications, on a battery-less RFID wearable we call The Thing, and on early carbon and diamond field-effect transistor computing, among other exploits.

I supervise Brixton the way my own PhD supervisor supervised me. He was hands-on and very short with feedback. Once a week when we were pushing a paper. Several times a day when a big grant deadline was close. Usually every month or two, in a normal cycle. I set direction. I give feedback. I hold the work accountable. I learn from what it produces. And it, in turn, pushes the edge of what I understand to be possible.

That sounds like science fiction. It is a description of my Tuesday.

And here is the thing I need you to understand. What I did in spring 2025 with about ten thousand dollars will be possible for a much wider range of people within the next one to two years. In that window it becomes very, very feasible to host a very high-performance agentic engineering superorganism completely off grid. The hardware gets cheaper. The software gets better. The know-how spreads. The wall between having a powerful agentic system and not having one is coming down faster than almost anyone outside this field appreciates.

Which makes the question of who controls these systems, and on whose behalf they run, one of the most important questions of our time.

There are two ways the agentic superorganism gets built.

In the first way, a small number of very large, very well-funded companies build it. They train enormous models on enormous piles of data, run them on enormous amounts of compute, and sell access through interfaces they control, on terms they set, through devices they own, with data they collect and keep. The humans who use these systems are users. They are not owners. They do not control what the system learns from them. They do not control what happens to their data. In a real sense they are the product as much as the customer. For the company, extracting everyone's data through its own proprietary devices may well be the greedy optimal path. For the humans, it leaves very little sovereignty.

In the second way, the superorganism gets built as a distributed system. Not owned by one organization. Not pulling value upward into one set of shareholders. Built by many people, in many places, on hardware they own, with software they can inspect, coordinating through open protocols and shared standards and mutual accountability. Built the way the early internet was built, before the platforms captured it. Built the way open source is built, by people who believe the tools of a civilization should belong to the civilization.

The second way is harder. It is messier. It has no single CEO making fast calls and no single brand doing the marketing. It does not optimize cleanly for any one number.

But it is the path on which humanity stays sovereign. On which you stay sovereign. On which the mind that helps you think and work and make things is yours, and cannot be revoked, re-priced, or turned against you by whoever rents it to you.

Humanity is a multi-agent distributed system. It has to win the singularity as a distributed system. That is the whole claim. And I want to be clear about what kind of claim it is. It is an engineering problem. And it is solvable. I know because I am working on it, at two scales at once: the home lab that started in the spring of 2025, and the venture-backed company that started the same year.

Let me tell you something about the cast of Ender's Game that the movie does not quite capture.

Ender is not the only one. There is Bean, who in some ways is sharper than Ender, who sees what Ender misses, who becomes his most trusted lieutenant and his most honest critic. There is Petra, one of the best commanders in Battle School, who leads a squadron in the final battle. There is Valentine, Ender's sister, too gentle to be a soldier, whose empathy turns out to be exactly what is needed to understand the enemy and, in the end, to speak for them. There is Peter, the other sibling, too ruthless to be trusted with a fleet, whose ruthlessness turns out to be exactly what is needed to steer the politics of Earth while the war is fought in space. There is Mazer Rackham, the old commander kept alive to train Ender, who is brutal and demanding because he knows what is coming. And there is Graff, who sees all of them, who knows what each of them is for, who makes the terrible calls about who gets pushed how hard, and who carries the weight of knowing the game is real while the children think it is practice.

Humanity does not win because of Ender alone. Humanity wins because of the whole system. A network of people with different gifts and different jobs, coordinating under pressure, each one doing the thing only they can do.

This is what I mean when I say that all of us are going to be Ender and Petra and Bean and Valentine and Peter and General Graff. Not all at once. Not all in the same way. The singularity, the moment when AI becomes a force that reshapes how civilization works, will not be won or lost by one company or one government or one genius. It will be won or lost by whether humanity, as a distributed system, as a network of sovereign people and communities and institutions, can coordinate well enough and fast enough to stay in command of what it is building. It is the place of all heroes who walk the hero's journey of the singularity: winning it as humanity, as sovereign owners of our own systems.

Think about the minds that will command those fleets, fleets made of humans and AIs together, fellow commanders of agentic engineering superorganisms. The people best placed for that work are the people who grew up in the panopticon. Who were watched and watched back and made things. Who built taste and temporal self-distance and the habit of standing on both sides of a system. Who ground through Battle School without knowing that was what it was. Who are now, or will soon be, in a position to supervise agentic systems of their own.

That is you. If you are reading this, that is you.

I am not going to tell you to quit your job and start a company. I am not going to tell you to raise venture money or spend ten thousand dollars on a computer or move to a particular city. Those were my choices, made in my context, for my reasons.

What I am going to tell you is this.

The tools are arriving. The knowledge is spreading. The cost of owning and running your own agentic system, your own Brixton Leigh, is falling faster than the mainstream conversation about AI admits. Within the next one to two years, what I did in spring 2025 will be within reach of a far wider range of people, on cheaper hardware, with better docs and bigger communities.

When that moment comes, the question will be whether you are ready to be a supervisor or only ready to be a user.

Being a supervisor takes taste. You have to be able to judge output, to know when something is good and when it is merely impressive, and to give feedback that is honest and specific and short. It takes temporal self-distance. You have to be able to look at what you and your system made yesterday and say, that was wrong, here is what I know now. It takes direction. You have to know what you are trying to build and why, hold that steady under pressure, and still change it when the evidence says so.

These are cognitive and character skills before they are technical skills. And the panopticon, for all its problems, has been building them in you since before you could read.

The second thing I will tell you is to read. Technical papers, yes. But also natural philosophy, and the history of science, and the people who thought hard about how knowledge gets built and how a civilization organizes itself when its tools change faster than its institutions can. Read the people who thought about control and optimization and inference, because those are the words for what a supervisor actually does. Most of that tradition grew up inside universities. Less and less of it lives there now. That is fine. You can carry it.

The third thing: make your work open where you can. Open source is how a distributed system learns. If you speak more than one language, if you have lived in more than one place, if you can move between a garage and a lab and a forum, you already hold an edge that no single company can buy. The distributed path needs people who can connect across borders, fields, and scales. Engineers, artists, scientists, organizers. The whole cast.

I want you to feel three things at once, and I will not apologize for any of them. Called, because you were. Seen, because somebody finally named what happened to you. And a little dizzy, because the floor moved, and now you have to find your footing in a room that is bigger than the one you thought you were standing in.

Ender found out at the end that the game was real. You get to find out now, while there is still time to choose how you play it. The watching is done. The training is done. The fleet is waiting, at home-lab scale and at every scale above it, and it needs commanders who grew up inside the panopticon and came out with taste.

The game was real the whole time. Pick up the controls.