Site Piece Writing / 003
A third layer. What the machines will never carry for us.
I'm back on the couch. Third time this year, if you're counting, and I am. The brace again. The leg out straight. The room quiet enough that I can hear my son in the next room asking his tablet a question I don't know the answer to.
He says "hey Google" and then he doesn't stop. Fifty questions about dinosaurs. Volcanoes. How fast a train has to go before it stops being safe. I watched his face open the first time he did it, the way a kid's face opens when something finally keeps up with how fast he thinks. He's always been like that. Curious past the point where I have the answer, which sends me to the same search bar he's using, quietly, so he doesn't know his dad had to look it up too.
I think about the cost of that more than people assume. What it takes, somewhere else in the world, every time either of us asks a question we didn't strictly need to ask. I don't have that resolved. I hold it and keep building anyway, because for me the tool has to stay a means to an end. Not the end itself.
Last summer, the first time the ankle went, we spent a lot more time indoors than either of us wanted. I was already deep in building with AI, so instead of keeping that world separate from him, I let him inside it. We built a video game together. Pulled shark characters out of Runway and Midjourney. Cut it all together in CapCut, badly, happily, him narrating the whole thing over my shoulder.
Watching him direct his wildest imagination into something we could actually see and hold felt like handing him a kind of freedom I didn't get at his age. Not freedom from limits. Freedom from waiting on someone else's permission to make the thing in his head real.
And here's what I noticed, sitting there with him. The thing we built together, he will remember. Not because a server kept it. Because he was in the room, and I was in the room, and we made it side by side. That's a different kind of memory than the one his tablet keeps of him. I didn't have language for the difference yet. I do now.
I hear Mark wants us all walking around in Meta glasses all day. I get the pitch. Intelligence, everywhere, in your pocket or on your face, no longer rationed to whoever could afford to employ it. A single parent running a business gets the kind of strategic support that used to require a consulting firm. A kid with an idea gets the kind of technical horsepower that used to require a funded team. I don't think that vision is wrong. I think it's unfinished.
Because I still have to ask: who are we actually talking to in person once that happens? If the cashier is a robot, and the voice on customer service is AI, and the glasses are doing the noticing for us, at what point does connected just start meaning surrounded? Personal intelligence for everyone answers one question. It doesn't answer the one underneath it. Once everybody's individually brilliant, who still knows the whole story?
Picture my son at fifteen, twenty years from now, asking his AI what his grandmother was like at his age. It'll search. It'll find a few photos, a school record, a name in some clipping nobody archived on purpose. It'll give him an answer, and the answer will be thin. Not wrong. Thin. The way a shadow is a true but incomplete account of the thing that cast it.
The gap is memory in the real sense, not the technical one. Not storage. Not a transcript. A record of relationships across time. This decision came out of that earlier failure. This belief replaced an older one, and here's why. That kind of memory used to live in people, the way it lived in my grandmother, the way it'll live in me for my son if I do this right. Machines can log a life in more granular detail than any human relationship in history has managed. What they can't do is remember the people who made that life possible, unless somebody decides, on purpose, that those people are worth keeping.
I've watched this same gap open up in organizations, not just families. I built Legacy Narrative around exactly this problem, teaching partners a method and a handoff process because the alternative, the thing that happens by default, is that the reasoning behind a decision lives in one person's head and leaves when they do. Give that same organization twenty AI assistants, one per person, each one genuinely brilliant at its job, and you don't automatically fix that. You can make it worse. Four confident answers instead of one uncertain one. The finance side still modeling a partnership that fell through in March. Marketing still pointing to the old plan because nobody told its assistant otherwise.
Nobody in that room made a mistake. Everybody got smarter. The organization still went a little bit lost, because intelligence moved to individuals faster than understanding moved across the whole. That's not a new failure. It's an old one, just running at a speed we've never had to survive before.
Here's the part that worries me more than the org chart version. Two people can watch the exact same thing happen and ask their own AI to help them understand it, and each one, doing exactly what it was built to do, will explain that event using that person's own history, their own beliefs, their own goals. Not lying to either of them. Just doing the ordinary, useful work of making sense of something in terms that fit.
And the two of them can walk away from the identical event with understandings so far apart they can't really talk about it anymore. Not because either one is unreasonable. Because "reasonable" got built, quietly and helpfully, on two different foundations.
You've already lived a version of this. You've watched two people argue past each other with total confidence, and you've probably wondered how they were even watching the same thing. That's not new. What's new is a machine in each of their pockets whose whole job is to make that gap feel like clarity instead of distance.
Same thing shows up in what gets made, not just what gets understood. Somebody with no training can make something today that looks and sounds finished. That's close to an unambiguous good. Making things is one of the more dignified things a person gets to do, and I want the barrier lower, not higher. My son proved that to me on the couch with a shark and a bad CapCut edit that he was prouder of than almost anything else he's made.
But culture was never just the pile of things made. Think about "you the birthday." It didn't start on a platform. It started the way real things start, at a kitchen counter, three friends, no plan to go viral, carrying forward from chicken alfredo music, carrying forward from basic, a whole lineage of language that came from somewhere specific before it got picked up and relabeled and sent back to us with someone else's name on the box. Or think about Walter. Mr. Tendernism. What made him matter wasn't the barbecue plate by itself. It was everybody who was in the room for it, everybody who could tell you why it mattered, the community that noticed the plate had his name missing from it and did something about that.
A song, a phrase, a plate of barbecue, means more because of what surrounds it. Who made it and why. What it was responding to. What it meant to the first people who felt it and passed it along. Strip that away and you're left with the artifact alone, technically finished, sometimes genuinely good, and strangely light. Creation is becoming close to unlimited. Meaning never scales at the same rate, because meaning was never the artifact's property. It belonged to the people around it.
I don't think the answer is less intelligence, spread less widely. I think Mark is right that a monopoly on it is dangerous, and I think he's stopped one question short of the real one. Even in a world where everybody has personal access to something powerful, somebody still controls the infrastructure underneath it. Somebody decides what counts as context. Somebody decides what an AI treats as true when two sources disagree, what carries over and what gets dropped, who owns the accumulated history between a person and the system that's come to know them that well.
That's not a future problem. That's already how it works. Can you take what you've built with a tool somewhere else if you switch. Can a family keep its own memory independent of whichever platform happened to be popular the years that memory accumulated. Can a community that spent a decade building something together keep that thing if the tool disappears or changes hands.
Call it context sovereignty, for lack of a cleaner word. People, families, communities, ought to actually own the accumulated context that lets anything, artificial or otherwise, understand them. Not just access to something smart. Ownership of the memory that builds up over time, in a form that survives the tool that built it.
The first wave of this whole conversation was about who has the smartest model. That's already mattering less, because smart is becoming available from every direction at once. The next wave is about who owns the memory. Models will keep changing. Companies will keep changing. What a family or a community actually built, the decisions, the relationships, the lessons that cost something real to learn, deserves to be treated as something that outlasts all of that. Not something that quietly evaporates every time the technology underneath it gets replaced.
I'm not just writing about this anymore. I'm already building the version of it I wish somebody had handed me, something that holds context the way a person would, that doesn't forget you between sessions, that treats memory as structure instead of a feature. Quietly, for now. More on that when it's ready.
My son will grow up extraordinarily well remembered by machines. That part's already decided. What isn't decided yet is whether he grows up ordinarily connected to the people who made him possible, or whether that connection gets built on purpose, the way I'm trying to build it now, on this couch, with this brace on, telling him the parts I know before I run out of time to tell him.
The question was never going to be whether the machines would know us. It's whether, with their help, we still know each other. And I think you might already know which side of that you want to be building toward.