Not struggling. Not “facing headwinds.” Not pivoting to a new business model.

Dead.

I have a phone in my pocket containing essentially every song ever recorded, every podcast ever made, every radio station on Earth, several lifetimes of audiobooks, and an AI that will presumably hum something original for me if I get desperate enough.

And yet SiriusXM lives.

Not merely lives. It has tens of millions of subscribers and generates billions of dollars in revenue.

Why?

Satellite radio belongs to a very particular technological moment.

There was a time when it was genuinely magical. You could drive from Toronto to Miami listening to the same station. No hunting around the dial. No losing the signal outside Buffalo. No discovering that the only thing available for the next 150 miles was Christian talk radio and something called The Bear.

A satellite followed you across a continent.

Fantastic.

Then we invented the smartphone.

That should have been the meteor.

Instead, SiriusXM somehow became the cockroach of consumer technology.

And I think I know why.

The car saved it.

Cars are weird technological museums. They have enormous development cycles, remain on the road for 15 years, and are operated by people who absolutely do not want to troubleshoot Bluetooth while doing 120 km/h.

Satellite radio has one spectacular feature Spotify will never quite replicate:

You turn the car on and music comes out.

No phone. No app. No pairing. No choosing. No algorithm asking what mood I’m in. No notification that someone else’s AirPods have somehow stolen the connection.

Just press 60s on 6 and go.

We keep assuming the best technology wins.

It doesn’t.

Sometimes the technology that wins is the one that has removed one tiny piece of friction so completely that millions of people keep handing it $15 a month long after its reason for existing has disappeared.

Which is irritating.

Because every few years I look at satellite radio, surrounded by Spotify, Apple Music, YouTube, podcasts, CarPlay and unlimited cellular data, and think:

Surely now.

No.

It’s still there.

And then, this week, I received some wonderful news.

Columbia House is finally dying.

You remember Columbia House.

The people who would give you an apparently impossible number of CDs for a penny, provided you accidentally entered into a commercial relationship from which neither death nor bankruptcy appeared sufficient to release you.

For a certain generation, Columbia House was the original spammer.

Before spam had a folder.

Before email.

Before the web.

They came directly to your house, on paper, over and over and over again.

And somehow they were still alive.

Until now.

After 71 years, Columbia House is finally shutting down.

Which should make me happy.

Instead, it terrifies me.

Because I look at SiriusXM and suddenly realize I may have misunderstood the timescale involved here.

Do I really have to wait another 30 years for this thing to die?

I’m 55.

I don’t have that kind of time.

And if you’re wondering why I have such a hate-on for satellite radio, it’s not just that I’m aesthetically offended by the continued existence of this geosynchronous fax machine.

It’s personal.

I receive maybe ten pieces of actual mail at my house in a typical month.

An astonishing percentage of them seem to be from SiriusXM.

I don’t even want to think about how many emails they’ve sent me.

And every few years, I lease a new car.

The dealer hands me the keys. I drive away.

Somewhere, deep inside SiriusXM headquarters, a light begins to flash:

HE HAS A NEW CAR.

And the entire symphony begins again.

Welcome to your free trial.

Your free trial is ending.

Your free trial has ended.

We miss you.

Come back.

Here’s 50% off.

Here’s 70% off.

Here’s three months for the price of a sandwich.

FINAL NOTICE.

An important message about your SiriusXM service.

Another envelope.

Another email.

Another offer.

I have tried ignoring them. I have tried unsubscribing. I have tried explaining through my silence, with what I believe is remarkable consistency over many years, that I do not want satellite radio.

Nothing works.

And now I understand why the company won’t die.

It won’t even accept that I don’t want to be a customer.

So congratulations, SiriusXM.

You have survived Spotify. You have survived the smartphone. You have survived unlimited cellular data, Bluetooth, CarPlay, podcasts and essentially the entire accumulated audio output of human civilization becoming available from the rectangle in my pocket.

Apparently the only thing capable of killing you is me changing my home address.

But after learning how long Columbia House hung on, I’m beginning to suspect that wouldn’t work either.

They’d find me.

I don’t mean worse. In almost every measurable way, I prefer the world we have built. I like Amazon, online banking, GPS, streaming, Uber and being able to ask an artificial intelligence at eleven o’clock at night why I remember a particular episode of The Flintstones. I have no desire to stand in line at a bank again, and I don’t think civilization will be improved by making me ask a gas station attendant for directions to Sudbury.

But I think that while we were removing friction from modern life, we removed something else without noticing it: people. Not our families and close friends, but the other people, that enormous population of people we used to know a little.

I think that loss has something to do with why the twenty-first century can feel strangely lonely even while it is more connected, more convenient and more individually empowering than anything that came before it. I also think we can fix it without surrendering any of the things that made this world better.

That’s where I’m going.

I got there by wondering what happened to the doubleheader.

Yes, that’s quite a hard turn. Get used to it. There are a lot of these coming. This is an essay, so there is a direction, but as the title promises, we’re going to meander.

The doubleheader seemed as good a place to start as any.

It Was a Doubleheader

There is an episode of The Flintstones I have remembered since childhood in which Fred and Barney desperately want to get out of work to see a baseball doubleheader.

I didn’t see it in its original run. Like an enormous amount of the culture my generation absorbed, I saw The Flintstones in syndication, where the same episodes came around often enough to become part of childhood memory.

They lie. One lie requires another and the entire thing becomes an increasingly elaborate disaster involving illness, hospitals and Fred impersonating a doctor. Eventually, of course, everything collapses, and what I remember most is Fred’s defence:

It was a doubleheader.

Does everyone even know what a doubleheader is anymore? Two baseball games, same teams, same day, one after the other. Traditionally, one ticket got you both games. It wasn’t merely two games scheduled together. The doubleheader was an event. You settled in for the afternoon.

They used to be common enough that The Flintstones could build an entire episode around Fred desperately trying to get to one without having to explain why anyone would care. But where did the doubleheader go? For that matter, where did the double feature go?

Two movies for the price of one. Not a promotion or a special package, just a perfectly ordinary way movies were shown.

Once I started thinking about those two things, I began noticing how many other perfectly ordinary things had quietly disappeared.

Rain checks, for example.

The expression itself is a little surviving piece of the old world. “I’ll take a rain check” comes from the literal rain checks issued when an event, originally a baseball game, was rained out. Your ticket entitled you to come back another day.

Retailers borrowed the idea. If a store advertised something on sale and ran out, it would give you a rain check entitling you to buy it later at the advertised price. This wasn’t some obscure courtesy either. It was sufficiently expected that advertisements sometimes had to warn you: NO RAIN CHECKS.

There it is again, a little piece of customer accommodation that was once common enough that its absence required disclosure. We still use the expression. We just don’t see many actual rain checks.

Then I began remembering other things. Layaway. Free gift wrapping. Free alterations. Someone pumping your gas and cleaning the windshield. Hotels cashing your cheque because you were staying there. Businessmen getting a day room for a few hours to rest up after a meeting and before the flight. A flight, I will add, where a mile flown equalled a mile earned... how’s that for living in the past? Your ticket also included your luggage, your seat and something approximating a meal.

At Miracle Food Mart, you didn’t even take your groceries to the car. They were placed into a bin and disappeared down a conveyor system, rather like luggage at an airport. You went outside, got into your car and drove around to the pickup area, where you were reunited with your groceries and someone loaded them into your trunk.

Think about the infrastructure involved in that: the building, the conveyor, the bins, the loading area and, most importantly for where this essay is going, the people. All so that a middle-class family buying groceries didn’t have to haul them across a parking lot. It would boggle the mind to see a supermarket build something like that today.

It is missed.

Bell would let you place a person-to-person call. Think about what that meant. You weren’t asking Bell to connect you to a telephone number, you were asking them to find a person. An operator might call one number, discover the person wasn’t there, then try another number on your behalf, and you weren’t charged for the long-distance call unless they actually found the person you were trying to reach.

Because, quite reasonably, Bell understood that you couldn’t be expected to pay for a conversation you didn’t have. Imagine explaining that business model today.

And if something was seriously wrong with dinner, the manager didn’t offer you 15 percent off your next visit through an app. They tore up the cheque. The transaction was over. You were their guest, something had gone wrong and they were embarrassed by it. There was no coupon designed to make sure you came back and spent more money, no survey and no case number.

They tore up the cheque.

At Clevelands House in Muskoka, where my family vacationed when I was a boy, we had a waitress assigned to our table. There was a man from the bar in a red jacket. There were bellmen and desk clerks and staff everywhere, mixed in with families who stayed long enough that everybody began to recognize everybody else.

This wasn’t the Ritz. This was middle-class family tourism in Ontario.

Initially I thought I had discovered an article about getting ripped off. Look at everything we used to get. Companies have spent decades identifying things that were once included in the price, separating them and either charging us for them, automating them or making us do them ourselves.

There is certainly an article there.

But this isn’t that article.

Because after making the list for a while, I realized I didn’t care very much about most of it. I can carry my suitcase. I can put groceries into the car. I don’t own a chequebook, so the fact that a hotel will no longer cash my cheque is not keeping me awake at night, and I actively prefer online banking to interacting with a teller.

The thing I missed wasn’t the service. I missed the people who performed it.

Where Did Everybody Go?

The bag boy wasn’t my friend. Neither was the cashier, the gas-station attendant, the bank teller, the bus driver, the waitress, the bartender, the butcher or the man at the airline counter, but they weren’t all strangers either.

That distinction seems important now because we’ve lost an enormous amount of the territory that once existed between those two words.

There was family, close friends and friends, but then there were neighbours, colleagues, acquaintances, the barber, the mechanic, the waitress, the bartender, the woman at the bakery, the teller at the bank, the guy at the hardware store, the cashier who knew your kids, the bus driver you saw every morning and the person whose name you never learned but whose face you saw every Tuesday for fifteen years. Then, finally, there were strangers.

It was a spectrum. Now I sometimes feel as though there are my people and eight billion strangers.

My Uber driver is a stranger. The Amazon driver is a stranger. The person filling my grocery order is invisible. My bank is an app. My landlord could be a corporation. The customer-service representative is whoever happens to receive the ticket. Even where the human being remains, the system is generally designed to make that person interchangeable.

I can take a hundred Uber rides with a hundred different drivers and every transaction can work perfectly without one of those people moving a millimetre closer to becoming someone I know.

This is different from saying I don’t have friends. I do. What I have lost is granularity. There are fewer kinds of people in my life, and one of the kinds I miss most is my guy.

Tell Him I Sent You

Everybody used to have a guy. If the furnace was making that noise, you had a guy. Need a mechanic? Call my guy. Plumber, accountant, dentist, lawyer, butcher, contractor, travel agent, somebody who could find the weird part for the old car, somebody who knew somebody at City Hall.

And then came one of the great sentences of ordinary social life:

Tell him I sent you.

I almost never hear that anymore.

“My guy” wasn’t necessarily a friend. That’s what made the relationship interesting. He occupied one of those positions in the middle of the spectrum, and he had a reputation.

Reputation meant something slightly different before we had reputation systems. If I sent you to my plumber, I was lending you a little of my relationship with him. He had a reason to look after you because I had sent you, and you had a reason not to behave terribly because you’d arrived carrying my name. If he did a good job, he might become your plumber too. Eventually somebody would ask you if you knew a plumber and you’d say, “Absolutely. I’ve got a guy.”

The web thickened.

Today I can find hundreds of plumbers in seconds and read thousands of reviews, and let’s be clear, this is often better. My friend’s recommendation is one data point. Google can give me four thousand. The old networks could be nepotistic, exclusionary and spectacularly wrong. Sometimes your guy was an idiot.

Online reputation systems have solved the information problem brilliantly, but they have replaced reputation with reputation data. Nobody sent me, no relationship is being lent and nobody enters my network. The plumber fixes the sink, I leave five stars and we disappear from each other’s lives. The transaction may actually be better, but the web doesn’t get any thicker because of it.

We Used to Need Each Other

For almost all of human history, other people were where the information lived. If you wanted to know what the weather was going to be tomorrow, where you should eat, how to get somewhere, who was a good plumber, whether the hotel was any good or what happened downtown, you asked someone.

Even trivial uncertainty generated human contact.

This was a terrible information system in many respects. People gave awful directions. They repeated nonsense. They recommended their cousins.

I can’t tell you how many times I pulled over to ask someone for directions, listened carefully, thanked them and drove away. Isobel would wait until we were out of sight and ask, “Do you remember all that?”

“There is no way we’re listening to him.”

And then we’d ask someone else... but only after we’d driven out of view, because you didn’t want to be rude.

Google is better. GPS is better. Weather apps are better. Online reviews are better. But notice what just happened: we got lost and talked to two people. Today we don’t get lost, which is better, but we also don’t talk to the two people.

That’s the part that never appeared on the specification sheet and now I am having a long conversation about human isolation with an artificial intelligence. Which seems almost too perfect.

Technology has gradually broken one of the oldest equations in human life: I need something, therefore I need somebody.

Now I need something and I get it. That’s progress, but I think we’ve made an enormous conceptual mistake along the way. We removed our need for people without removing our need to be among people.

The cashier appeared on the supermarket spreadsheet as labour cost, as did the bag boy and the teller. The travel agent appeared as distribution cost and the doorman as labour cost. Nobody’s spreadsheet contained a line saying “person I know,” so the value was zero.

Each individual decision was rational. Nobody decided to make society lonely. We optimized it into loneliness.

We Don’t Even Have to Ask

There is another side to this that I find increasingly strange. Not only do I no longer need to ask other people what they know, increasingly nobody needs to ask me what I know either.

Every restaurant wants my review. Every hotel wants my feedback. Every Uber driver rates me and I rate the driver. Amazon wants to know what I thought of the toaster, Google wants to know whether the restaurant was good, the airline wants a survey, the dentist sends one and the hotel sends another. Every transaction seems to end with somebody asking me to deposit my experience into a database.

At first glance this looks almost aggressively social. Everyone wants to know what I think, except nobody actually wants to ask me. They want the information separated from me.

Once I have reviewed the restaurant, you don’t need to ask me whether it was good. Once enough of us have rated the hotel, nobody needs to say, “Douglas stayed there last year. Ask him what he thought.” My experience has been extracted, aggregated, scored and made permanently available to people I will never meet.

We used to store an enormous amount of the world’s useful information inside people, and that gave people value to one another. You needed someone who had been there, someone who knew the neighbourhood, someone who had tried the restaurant, someone whose cousin had stayed at the hotel, someone who remembered what happened last time or someone who knew a guy.

Now we’re steadily saving ourselves to the net: what I bought, where I went, what I liked and hated, what I watched, what I searched for, what I thought of dinner, which photographs I took, who I know, what music I played, what I wrote and what I said. Given enough of that history, increasingly sophisticated systems can reconstruct a remarkably useful approximation of me without requiring my presence at all.

My mother has been dead for nearly five years, and I suspect I could still find her opinion of restaurants I didn’t even know she had visited.

Think about how strange that is. There was a time when, if I wanted to know what my mother thought of a restaurant, I had to ask my mother. The information and the person were inseparable, whereas now the opinion can outlive the person.

There is something rather wonderful about that. There are traces of my mother in the world that didn’t disappear when she did. But viewed from the other direction, it reveals how profound the change has been. We used to need the person to get the information. Now we can keep the information without the person.

And when information travelled through people, something else travelled with it: relationship.

If I told you, “Go there. Tell them I sent you,” I had put a little of myself behind the recommendation. If you had a terrible evening, I would probably hear about it. There was consequence.

Maybe my smartwatch should give me a tiny electric shock every time somebody has a bad experience at a restaurant I recommended online. Not that I intend to wear a smartwatch, but at least it would put me back into the loop.

Because that’s what we’ve been removing. Information used to travel from person to person and, in doing so, create or reinforce a relationship. Now information travels perfectly well by itself.

That is enormously useful. I use reviews constantly and I don’t want to give them up, but once again we solved the stated problem and ignored an incidental output. We wanted the information. It turned out that having to get it from another human being was producing relationships.

There is something particularly unsettling about realizing that the technological trajectory isn’t merely toward a world in which I need fewer people. It’s toward a world in which fewer people need me, not because I have nothing to contribute, but because increasingly what I might contribute has already been captured, indexed and made available without the inconvenience of asking me.

For most of history, if you wanted Douglas’s experience, you needed Douglas. You might have to call me, invite me to dinner or run into me and remember there was something you wanted to ask. Once we were talking, I might tell you something you hadn’t thought to ask and you might tell me something in return. Information was the excuse for the encounter, not necessarily its limit.

We have become extraordinarily information-rich while becoming less informationally dependent upon one another, and dependence is an uncomfortable word. Modernity has quite reasonably spent centuries trying to free us from unnecessary dependence. But perhaps a little mutual dependence was part of the glue. Not helplessness or gatekeeping, just enough incompleteness that occasionally I had something you needed, you had something I needed, and we had to talk.

Fred Flintstone Was Richer Than I Am

This is where The Flintstones came back into the story.

Forget the doubleheader for a moment and look at Fred’s life. He has a wife and young child. His best friend, Barney, lives next door and their wives are friends. Fred and Barney bowl together and belong to the Loyal Order of Water Buffaloes. Fred has relationships at the quarry. His boss can be invited to dinner. There are bridge nights, neighbours, social obligations and people constantly wandering in and out of one another’s lives.

Fred’s different worlds overlap. Barney isn’t filed under FRIEND and encountered only when Fred schedules a friendship activity. He’s the neighbour, best friend, fellow Water Buffalo and co-conspirator. Work crosses into home, neighbourhood crosses into friendship, the lodge crosses into leisure, Wilma’s relationships cross into Fred’s and Pebbles connects another generation.

Of course this is a sitcom and writers need recurring characters. But the writers could use these institutions without explaining them because the audience recognized them. A lodge wasn’t an exotic social intervention, bridge night wasn’t therapy for loneliness and having the boss over for dinner wasn’t bizarre. These were simply things people did.

And I’m not using television to invent a world that didn’t exist.

My mother belonged to an amateur theatre group in Mississauga. At one point they were working on a production of Cabaret, and I don’t remember her ever being happier. There were rehearsals, people to know, things that had to get done and somewhere she was expected to be. The play was almost beside the point. It gave them a reason to keep assembling.

My father belonged to the Thorncliffe Ski Club. They skied, travelled, played baseball and did whatever else a ski club somehow ended up doing. He liked some of the members and couldn’t stand others, which is important. These weren’t carefully selected friends. They were simply fixtures in his life, and therefore in mine, until I was a teenager.

Then my father discovered the Canadian Opera Company and started disappearing for days with his best friend John so they could sit around being extras. I once saw him on TVOntario, at about eleven o’clock at night, hand a sword to one of the principal characters and say, “Your sword.”

It was amazing.

These weren’t people looking for solutions to loneliness. My parents had things they belonged to. The thing itself could be almost absurdly specific. Skiing. Amateur theatre. Opera. But each came with a cast of recurring human beings.

I don’t have memories like that from my adult life.

Somewhere between their adulthood and mine, belonging stopped being something that ordinary life simply kept providing.

Fred did not need to realize that he was experiencing insufficient social connection and download an app to address it. His life kept throwing people at him.

By contemporary standards, Fred Flintstone was socially loaded.

And I want my Water Buffaloes.

The Cocoon

There is a natural answer to some of this, which is that I’m getting older. Of course I am, and life has always narrowed.

I started in nursery school and moved through school, camp, a group home for a while, university and early working life. I was constantly placed into groups. Grade Five gave me thirty peers. Camp put ten or twelve boys in a room. Nobody asked whether we were compatible.

Then people pair off, establish careers and have children, and the circles become smaller. That part is natural, but I’ve realized the curve isn’t simply a steady decline.

There is a cocoon in the middle.

When you have children, they generate an enormous social world around you: daycare parents, school parents, hockey parents, swimming parents, birthday parties, recitals, tournaments and fundraisers. For years I had friendships created by my children’s activities that felt quite substantial.

Then the children stopped swimming and the swimming parents disappeared. Nobody fought and nobody decided the friendship was over. There simply wasn’t the need for people who could grab your kid if traffic was bad, or the social demand to remember Julie’s father’s name when you saw him at drop-off.

And that’s when you discover that some relationships were almost leased from the institution that kept putting you together. They were real relationships, but they had a job to do. A few survive independently, while most become shadows of the cocoon that created them.

Then the children grow up and leave, which is normally discussed as a transition between parent and child, but something else leaves with them: the last great automatic acquaintance machine of middle adulthood.

I find myself approaching an empty nest in my mid-fifties feeling much younger than the traditional idea I had of someone in his mid-fifties, and wondering where the next institution is supposed to be. There is no Grade Thirteen for fifty-five-year-olds, no dormitory and no camp, and I’m not ready for the retirement community.

Curiously, at the far end of life we seem to have rediscovered the answer. Look at the retirement communities in Florida. Their days can be absolutely packed with people: coffee, cards, swimming, golf, pickleball, drinks, dinner, somebody’s birthday, somebody who needs a ride and somebody with gossip.

My mother experienced something similar later in her life. Her days were full of friendships. Booking time with her was like getting a table at the best restaurant in town. She lived alone, but she was about as far from lonely as you can get.

So perhaps the curve falls off a cliff and then eventually rises again. We have built enormous social infrastructure for children and increasingly sophisticated social infrastructure for older people. The middle gets a house, Netflix and instructions to put itself out there.

And even those ends are thinner than they were.

Where’s the Dance?

My children went through most of middle school and high school without school dances. There was a dance around graduation, organized by parents, and I actually complained to the principal about the absence of dances. The answer was essentially that they were too much trouble.

Of course they were. Teenagers plus darkness plus music has presumably been too much trouble since the invention of music.

But that answer began bothering me for the same reason everything else in this essay bothers me. The dance is trouble. The band costs money. The football team needs coaches. The club needs a teacher willing to stay late. The school play consumes enormous effort. The trip creates liability. The assembly takes time away from instruction.

One by one, institutions identify the peripheral things they do and ask whether those things are really part of the core mission. A school’s core mission is education, so it educates. A bank’s mission is financial transactions, so it transacts. A supermarket’s mission is selling groceries, so it sells them. A hotel’s mission is renting rooms, so it rents them.

Each becomes extremely good at its core function, and the world gets colder because perhaps we’ve mistaken the core function for the whole function.

A school was also supposed to give you a dance, a terrible football team, a band that wasn’t very good, a ridiculous musical, a club, a cafeteria table and a song. Brian Wilson could write Be True to Your School and millions of people understood the emotion because it was your school.

We have become extraordinarily sophisticated about asking young people who they are. What is your identity? What describes you? What makes you distinct? I’m beginning to wonder whether we have neglected the complementary question: what are you a member of? Not what describes you. What are you part of?

Fred Flintstone had answers. Husband, father, quarry worker, neighbour, Water Buffalo, bowler, friend of Barney, citizen of Bedrock. Each answer contained actual people.

An identity category isn’t necessarily a group identity. A group exists when particular people expect one another to show up.

We Broke the Age Mix

The groups that remain have also become remarkably segregated by life stage. Children inhabit children’s institutions, parents inhabit parenting institutions, middle-aged adults largely fend for themselves and older people eventually enter institutions for older people. They barely touch.

That wasn’t always true. Churches, neighbourhoods, lodges, service clubs, local businesses, extended families, resorts and community organizations put different generations into the same physical and social worlds. A twelve-year-old could know a seventy-year-old who wasn’t his grandfather. An older person could know children who weren’t grandchildren. A fifty-year-old could have recurring relationships with twenty-year-olds who weren’t employees or relatives.

We’ve built increasingly efficient social silos, and that’s another reason Clevelands House glows in my memory. It was mixed. Staff and guests, children and adults, young employees and older guests, families and couples. Everybody occupied the same place long enough for faces to become familiar.

It wasn’t simply populated. It was entangled.

The Great Inversion

There is another part of this that I find almost perverse.

Historically, poor people were chock-a-block full of other people, often horribly so. Several generations under one roof, shared bedrooms, tenements, boarding houses, factories, markets, streetcars, churches, pubs and neighbours constantly within earshot. Privacy was a luxury.

Money bought distance: your own bedroom, your own house, your own carriage, your own automobile, your own land. At the extreme stood the rich man alone on the mountain, interacting mostly with the servants required to maintain his isolation.

Now we’ve inverted it.

Technology has democratized the rich man’s isolation. A person of modest means can bank alone, shop alone, work alone, entertain himself alone and order dinner alone. The landlord can be a corporation. The groceries arrive from a platform. The driver is algorithmically assigned and disappears afterward. The delivery can be left outside so that neither person even has to say hello.

Meanwhile, I have a very wealthy friend and have occasionally had the privilege of travelling inside his bubble, and I noticed something. There wasn’t a Coke Zero on the entire trip that was any better than the one in my refrigerator.

What was different was the number of people he interacted with who genuinely cared about his needs, and he theirs.

The captain knew him, the driver expected him and the hotel manager came out to greet him. When the flight attendant spoke to him, she didn’t run the script I get. The same basic work was being done, but one was a transaction and the other was a connection. There was a crew on the boat. People knew how he liked things, knew where he was going and noticed that he had arrived. And he knew them.

We call this service, which makes it sound as though the luxury is that he doesn’t have to procure his own nuts or put the Jet Ski in the water. That’s not what struck me. His life was thick with recurring human beings who cared about him, and whom he cared about.

The rich can still buy something resembling a village. The rest of us got the app.

The app is fantastic, but it’s awfully quiet in here.

Fight Club

Which made me think differently about Fight Club.

The Narrator begins with an extraordinarily modern life: job, apartment, money, travel, autonomy and possessions selected from a catalogue to construct an identity. Functionally, everything works, but nobody needs him and he needs nobody. He encounters “single-serving friends” on airplanes, people designed to disappear from his life when the transaction ends, and he cannot sleep.

Then he discovers support groups. He goes somewhere, the same kinds of people are there, there are rituals, someone holds him, he cries and he sleeps.

Then he builds Fight Club.

What does he actually build? A community with rules, rituals, shared language, physical contact, obligation and a place you’re expected to appear. He moves from a sterile condominium containing one man and his furniture into a decrepit house overflowing with people.

And then the film shows us the danger. Fight Club becomes Project Mayhem. Belonging becomes identity and identity becomes obedience. Tyler takes the hunger for connection and weaponizes it.

Perhaps that’s part of what happens when ordinary society stops supplying enough low-voltage belonging. People don’t stop needing to belong, they find high-voltage versions: political movements, grievance communities, online tribes, conspiracy movements, talk radio.

And religion, although religion is more complicated.

Whatever one thinks about what happens upstairs, religion spent several thousand years developing some spectacular social technology. You go somewhere at a particular time and the same people return next week. There are children and old people. Someone is sick, someone needs a casserole and someone knows a roofer. Then you go downstairs into the basement for coffee and tea with women wearing some of the nicest hats you’ve ever seen.

I increasingly suspect the basement was doing an enormous amount of the work.

The people there didn’t need to become your friends. They just needed to stop being strangers. Thirty little places to belong may be considerably healthier than desperately needing one place to tell you who you are.

We Need Reefs

Rapid social and technological change has bleached the coral reef of ordinary human life. That doesn’t mean the people disappeared. They’re still swimming around. What’s missing is the substrate around which relationships used to grow.

When people try to restore a damaged reef, they don’t manufacture a mature ecosystem one coral at a time and lower it carefully into the ocean. Sometimes they sink a ship. They put structure into suitable water and give life somewhere to attach.

That is what I think we need to do socially. We need to sink ships.

We need social substrate: places and institutions where the same imperfectly selected people repeatedly encounter one another long enough for familiarity to accumulate, and they need to be intergenerational.

Creating a loneliness program for fifty-five-year-old empty nesters would reproduce precisely the mistake I’m complaining about. I don’t want to be placed in a room containing twenty-three optimized versions of myself. I want the fifteen-year-old who knows how to fix something, the thirty-year-old with the baby, the seventy-five-year-old with an outrageous story, the woman I don’t particularly like but have nevertheless known for six years and the guy who knows a guy.

The objective isn’t friendship. It’s fewer strangers.

Which brings me, with enormous reluctance, to pickleball.

I would rather eat my own liver than play pickleball. Nevertheless, I am forced to concede that the pickleball people may have accidentally discovered something important. It is relatively accessible, people rotate, there is time between games, different ages can participate and people return regularly. Familiarity accumulates around an activity whose stated purpose isn’t please make friends now.

Fine. Study it. Extract whatever social chemistry is working and apply it to something I might actually want to do: food, repair, gardening, music, games, making things, local history, community meals, amateur theatre, dogs, bad sports, anything. The particular activity is almost beside the point.

The reef needs substrate.

The Old Community Was an Accident

There is one enormous advantage the new world has over the old one: we can finally be intentional about this.

The social world I remember so fondly wasn’t designed. It was mostly an accident of geography, institutions and necessity. These were the children on my street, so I played with them. These were the boys at camp, so I lived with them. These were the parents whose children swam with mine, so we stood beside one another for years. This was my neighbour, so eventually I knew him.

Sometimes those accidents produced the Barney next door and sometimes they didn’t. For most of history, we couldn’t do much better. Finding the forty people in Toronto with whom I might have some peculiar but meaningful affinity would have required information nobody possessed, searches nobody could perform and more cognitive effort than any sane person would expend assembling a social life.

Technology has changed that.

We have spent decades building an astonishingly detailed record of human preference: what we read, what we watch, where we eat, what makes us laugh, what we buy and don’t buy, the places everybody loves that Tanya and I both hate, the obscure subjects Jeff and I independently disappear into for three hours on a Sunday afternoon.

The computer can see patterns across millions of tiny decisions that neither of us could possibly hold in our heads, and mostly we use this extraordinary capability to decide which shoes to show me.

Perhaps we’re thinking much too small.

Consider restaurant reviews. The system already knows that everyone loves a particular restaurant except Doug, Jeff and Tanya. That’s interesting. Perhaps Tanya also hated that other place everybody loved, and Doug hated it too.

Over time these shouldn’t merely be anonymous data points. They could become people whose judgement has acquired meaning. The system could tell me, “This restaurant has spectacular reviews, but Tanya didn’t like it, and Tanya’s taste has matched yours seventeen times out of twenty. Doug didn’t like it either. We don’t think you will.”

Now Tanya matters.

I don’t necessarily need to become Tanya’s friend, but she has moved one notch along the spectrum from stranger toward someone I know. Eventually I know that Tanya hates overproduced restaurants, Jeff has inexplicably terrible taste in hotels and Douglas is the person to ask about somewhere for dinner in Montreal. Those aren’t merely recommendations anymore. They’re the beginnings of reputations.

We already possess the machinery required to do this. We collect the opinions, identify the patterns and know who agrees with whom, but instead of allowing those patterns to create modest relationships between actual people, we grind everybody into the aggregate and return a number.

4.6 stars.

We have taken an extraordinarily rich human network and turned it into a decimal. I don’t want to abolish the decimal because it’s useful. I want Tanya back.

I don’t want seventeen million people to know that Douglas liked a restaurant. I want the seventeen people for whom that fact might actually mean something to know. Maybe three already know me, four have discovered that we reliably like the same restaurants, one lives around the corner and another has recommended five places I’ve loved.

Maybe I don’t know any of them yet.

Why should the system’s only ambition be to average us? It knows enough to notice that perhaps we belong near one another. We built machines that are extraordinarily good at discovering affinity, then mostly used them to decide what to sell us next.

What a waste.

There’s somebody for everybody, we like to say. Perhaps, for the first time in history, we actually have a reasonable chance of finding those somebodies.

Not through a questionnaire that discovers we both enjoy travel and long walks. The data is vastly richer than that. It contains thousands of tiny revealed preferences accumulated over years, not merely what I say I like but what I actually return to, what makes me laugh, what I linger over and what I reject when almost everyone else approves.

The machine might know that Tanya and I should know one another before either of us does, and that creates an opportunity human beings have never really had. We can finally be intentional about relationships at a scale and with a degree of subtlety that previously exceeded our information, time and cognitive capacity.

Not by calculating my perfect friends. I don’t want seventeen people algorithmically certified as Douglas-compatible. That sounds unbearable. Difference matters, accident matters and the person who surprises me matters. The mildly annoying person who has been coming every Thursday for six years may matter enormously.

But technology can improve the odds of the encounter.

Find the right people, then do something wonderfully old-fashioned and put them in a room. Not once, because that’s networking. Put them there next Thursday and the Thursday after that. Give them something worth doing badly together, mix in people the algorithm wouldn’t have selected and let familiarity do its slow work. Let reputations form and obligations appear.

Eventually Tanya stops being the woman whose restaurant preferences correlate with mine.

She becomes Tanya.

Then turn the machine off.

The algorithm’s job isn’t to become the community. Its job is to help the community happen.

Chosen Belonging

Some versions of this already exist and others have been imagined for us.

Retirement communities such as The Villages in Florida demonstrate that people will enthusiastically reorganize their lives around places deliberately designed to produce repeated social contact. Whatever else one thinks about The Villages, its residents don’t seem to lack things to join.

Neal Stephenson imagined something more ambitious in The Diamond Age. His world is organized partly around phyles, communities people join through affinity rather than simply inheriting through geography or birth. The Neo-Victorians are one. Other phyles stretch across geography so that belonging doesn’t necessarily disappear when you cross a city or national border.

That idea has stayed with me because it suggests a third possibility between the village we were born into and the atomized individual we have become.

For most of history, community was assigned. You were born here, these were your neighbours, this was your church, these were your relatives, these were the children available to play with and these were the people at the factory, lodge, school or pub.

The great liberation of modernity was that increasingly we could leave.

Good.

But perhaps we confused the freedom to leave an assigned community with freedom from needing community at all.

What Stephenson imagined was community becoming elective. You don’t have to remain Catholic because your parents were Catholic. You don’t have to spend your life among the people who happened to be born within five miles of you. You can seek out a group because something about those people, their values, interests, temperament, aesthetics or way of living makes you want to belong there.

And crucially, the group isn’t merely an online forum. It has an address.

Imagine belonging to something that has a house in Toronto, apartments in Chicago, a building in Tokyo and a table somewhere in Paris. You arrive in another city and there are people there who are not quite friends and certainly aren’t strangers.

The internet gave us the first half of this. It allowed us to discover our people regardless of geography, but then, strangely, we mostly left them there on the internet.

Perhaps the next step is to bring them back into physical life.

There is a danger here. If the algorithm puts me exclusively among Douglas-compatible people, we’ve recreated one of the worst features of the internet in physical space. The reef needs biodiversity.

So perhaps the principle is intentional at the edges and accidental on the inside. Choose the community because something draws you there, then surrender some control.

Tanya comes on Thursdays. Jeff brings his terrible opinions about hotels. Somebody’s sixteen-year-old is there and somebody else is eighty-two. People bring spouses and friends and new people arrive. You don’t get to swipe left every time someone irritates you.

The technology gets you across the threshold. Then proximity, repetition and mild social obligation take over.

The old community was assigned and the current world is atomized. The opportunity now is something humanity has rarely had at scale: chosen belonging. Not merely finding my people online, but finding somewhere among them to live.

Things Worth Doing Badly Together

There is one quality I think these new institutions particularly need. They need to give us things worth doing badly together.

The school band doesn’t need to be excellent. Community theatre doesn’t need to be Broadway. The Water Buffaloes don’t need measurable outcomes. Our terrible football team is still our terrible football team.

Optimization works against this because why listen to my neighbour play guitar badly when Spotify contains every great guitarist who ever lived? Why watch local hockey when I can watch the NHL? Why attend amateur theatre when every great film is available on demand? Why ask someone where to eat when Google has reviewed every restaurant in the city?

The rational consumer choice repeatedly points toward the world’s best product and away from local participation. Consumption becomes spectacular while participation withers, but participation produces something consumption cannot.

It produces a we.

Our school, our street, our bar, our table, our team, our club, our ridiculous Christmas pageant, our guy.

We’ve become spectacular at constructing the I: my feed, my playlist, my recommendations, my preferences, my schedule, my identity, my experience, my AI. We haven’t become nearly as good at preserving the stupid little we, and perhaps those little plural pronouns were carrying more of civilization than we realized.

Xenia

Which brings me back much farther than Fred Flintstone.

One of the ideas running through The Odyssey is xenia, the ancient obligation between host and guest. A stranger arrives and the host has obligations toward him, while the guest has obligations in return. The interesting thing is precisely that they aren’t already friends. Hospitality creates a temporary bond between people who otherwise have no reason to care for one another.

The violation of that bond is serious enough in the Homeric world to destroy kingdoms.

That makes me think differently about the hotel cashing the cheque. The hotel wasn’t in the cheque-cashing business. It cashed your cheque because you were its guest.

Bell wasn’t being paid to search half a city for your uncle. It did it because the service it understood itself to be providing wasn’t a completed circuit between two telephone numbers. It was communication between two people.

The restaurant didn’t give you a coupon toward another transaction when dinner went badly. They tore up the cheque. The bellman carried the luggage because you were a guest. The grocer put your groceries onto a conveyor belt, sent them through the building and had another human being load them into your car because you had entrusted him with provisioning your household. The department store wrapped the present because you were its customer.

Calling these things services almost misses something. They were courtesies, and courtesy presupposes a relationship.

Once every courtesy is subjected to independent economic analysis, courtesy becomes extremely difficult to defend. You charge for it, automate it, make the customer do it or eliminate it.

Customers responded rationally too. If I’m not your guest and you’re not my grocer, if we’re merely counterparties exchanging an interchangeable commodity, why shouldn’t I cross the street to save a penny?

Businesses stopped treating customers as guests because customers behaved like transactions, and customers behaved increasingly like transactions because businesses stopped treating them as guests.

Somewhere along the way the host and the guest disappeared and we became providers and users instead. It sounds like a trivial change in vocabulary, but somewhere inside that transformation all those little relationships surrounding commerce disappeared with them.

The Christmas Lights

This is where I finally understood what had been bothering me.

Every Christmas I put incandescent lights on my tree. Technologically, this is idiotic. Incandescent lights are inefficient, they produce heat and they break. LEDs are superior in almost every engineering sense.

But LED lights make my tree look cold, and Christmas colder with it. So LEDs don’t touch a single branch.

I watch a YouTube channel called Technology Connections, and he has a great video where he finally explained to me what it was I hated. The problem wasn’t that LEDs were somehow vaguely less Christmassy. The colours were wrong. There wasn’t enough blue in the blue and there was too much red in the red.

Better still, he showed me a company that had fixed it.

That’s important. They didn’t fix Christmas lights by going back to incandescent bulbs. They figured out what the new technology had accidentally lost and engineered it back in.

Suddenly the whole thing became a rather good metaphor for what I had been trying to say about people.

I knew something felt colder before I knew what was missing.

This is something we seem to do whenever a new technology arrives. Its enthusiasts measure all the things the technology improves, declare complete victory and occasionally become irritated with anyone who says something was lost.

CDs will last forever. Social media will connect everybody. The paperless office is coming. The LED lasts vastly longer and consumes a fraction of the electricity.

All substantially true, and incomplete.

The problem is that we define the old thing according to the characteristics we know how to measure. The job of an incandescent bulb appears to be producing light, so we measure lumens, watts, lifespan and cost, and LED wins.

Except human beings had spent generations living under incandescent light, and there were characteristics of that light nobody had bothered putting into the specification because the old technology produced them automatically.

That’s what I think we did with modern life.

We optimized for the things we knew how to measure: price, speed, convenience, availability, efficiency. We didn’t put “number of recurring human beings in your life” on the specification sheet because the old system produced them automatically.

Then the replacement arrived and some of us started saying there’s something wrong with the light.

That is how I now feel about my life.

There’s something wrong with the light.

I can give you a thousand reasons why this life is superior to the one I grew up in, and I would not trade back for a million dollars. I want Amazon, online banking, GPS and streaming. I want the freedom to find obscure people and ideas anywhere in the world. I want the AI I’m talking to while working all this out.

The old social web had warmth, but it also had gatekeepers. Access to the good stuff could depend on making the right connections, knowing the right people or simply being sufficiently like everyone else to be let in. There were plenty of people standing outside that warm little village looking in.

I was often one of them.

I don’t want that back either.

I want the LED.

But life does not feel as thick as it once did.

At first I thought that was age, and then I looked around. The cashier disappeared, the bag boy disappeared, the teller disappeared, the station attendant disappeared and the travel agent disappeared. The lodge disappeared, the school dance disappeared, the club disappeared and the shared television program disappeared. My guy disappeared. “Tell him I sent you” disappeared. The generations separated and the parenting friendships evaporated when the children stopped swimming.

The institutions became very good at their core functions and abandoned their incidental ones. Nobody made a decision to create this outcome. We simply kept optimizing, and now we stand beneath the result wondering why everything feels so cold.

I don’t know yet exactly how to measure the missing wavelengths. Maybe they are repeated contact, human density, weak ties, recognition, shared experience, group identity and intergenerational contact. Maybe it’s the number of relationship gradations between intimate and stranger, how many people say your name in an ordinary week, how many familiar faces you encounter without arranging to encounter them, how many people would notice if you stopped showing up and how many people can tell someone else, “Tell him I sent you.”

Maybe what we call loneliness is actually a bundle of ten different missing wavelengths, but I don’t think the fact that we haven’t completely measured the problem means the feeling is imaginary. We have lived under both kinds of light. We know which one feels warmer.

So I refuse the choice.

I don’t want to go backward. I don’t want the old bank, worse information or gatekeepers. I don’t want technology removed from my life, and I don’t want every transaction stuffed with unnecessary human labour merely to force me into conversation.

But I also don’t want to be told that the only alternative to all that progress is accepting a world in which almost everyone I encounter is a stranger.

I want online banking and somewhere to go on Thursday. I want Amazon and I want my kids to have a school dance. I want AI and I want somebody who isn’t family or a close friend to notice that they haven’t seen me lately. I want technology clever enough not merely to recommend the restaurant but occasionally to introduce me to Tanya.

I want to choose my closest relationships and have enough unchosen people in my life to occasionally be irritated by them. I want privacy and I want a reef. I want my Water Buffaloes, I want my guy and I want someone to say, “Tell him I sent you.”

We don’t need to choose between progress and people. The old social web was dense but accidental. The new network is powerful but thin. For the first time, we may have the tools to combine them.

We don’t have to choose between the village and the network. We can use the network to build a better village.

Look at the cast of the Village People. There was a construction worker, a cowboy, a cop, a soldier, a biker and a Native American. Nobody had anything in common and they made it work. They were, to steal a line from Sesame Street, the people that you meet each day.

Perhaps they were trying to tell us something.

We made the first specification incomplete. So keep the technology, the efficiency and everything we’ve gained, but figure out what wavelengths we left out and engineer them back into the world.

Keep the LED.

Fix the light.

It was too transparent.

It could spook the horses.

It made us look weak.

I understood what she was worried about. I had written openly about debt, pressure, risk and what it feels like to own a business when some of the pitches coming at you matter quite a lot.

So I showed her the article again.

I hadn’t written that the company was in trouble. I had written about what pressure feels like when you’re the person carrying it. In fact, I thought the whole thing was rather optimistic.

But underneath her reaction was a question we’ve been wrestling with for years.

What if people see us for what we really are?

And my answer is increasingly:

Good.

They should.

For years I’ve pushed against the instinct to make Thinking Cap look like everybody else. Look bigger. Look more corporate. Sand off the eccentricities. Make the website reassuringly beige. Use the same language as every other enterprise software company so nobody gets nervous.

The problem is that it doesn’t actually make us look like Loblaws.

It makes us look like Longo’s trying to look like Loblaws.

And why would we want to do that?

Nobody walks into a farmers’ market because they were fooled into thinking it was a supermarket.

They go precisely because it isn’t.

They want the person who grew the tomato. They want the strange cheese they can’t get anywhere else. They want somebody behind the counter who knows why this particular olive oil is worth buying. They accept that there aren’t forty brands of toilet paper because that isn’t what they came for.

That’s much closer to what we are.

We’re the farmers’ market.

We’re the European specialty shop.

We’re the retailer who sells truffles.

And if you put me in charge of Walmart, I would probably drive it into the sea.

I don’t have that skill set.

I don’t want aisle after aisle of interchangeable things efficiently stacked to the ceiling. There are people who are exceptionally good at building and operating organizations like that. I’m not one of them.

Give me the truffles.

Give me something peculiar and difficult that requires taste and judgment. Give me a customer who cares enough to understand the difference.

That’s a game I know how to play.

The same thing happens when small technology companies try to reassure customers who really want to buy IBM.

We can make ourselves look more corporate. We can carefully avoid saying anything that reveals an actual human being runs the company. We can fill the website with stock photography and enterprise language and reassuring statements about solutions and transformation.

But we’re not going to become IBM.

At best, we attract people who really want IBM but can’t afford IBM.

That’s not particularly interesting to me.

I want the customer who doesn’t want IBM.

I want the customer who looks at a giant vendor and thinks: I don’t want to be customer number 47,291. I want somebody who actually cares whether this works.

Those customers deserve to know what they’re buying.

We’ve been around for more than twenty years. We have longstanding customers. We have recurring revenue. We have a product that does complicated things extremely well. We’re stable.

But if your procurement criteria require a vendor with twenty million dollars sitting in the bank, it ain’t us.

If what makes you comfortable is knowing the entire senior management team could disappear tomorrow and the machine would continue operating exactly as before, that’s not us either.

We’re not disappearing.

And there is another side to having the people who run the company this close to the product.

We care.

A lot.

Probably an unreasonable amount.

The people making decisions about what Thinking Cap becomes aren’t seven organizational layers away from the customer. We’re right here. We know the product. We know the customers. We argue about it. We invent things. We change our minds. We take risks. Sometimes we get things wrong and fix them.

That’s not the weakness we’re supposed to hide.

That’s the product.

There is certainly a market for enormous, predictable organizations where no individual person matters very much. Scale provides real advantages. So does redundancy. So does having enough cash that almost nothing can frighten you.

But those advantages have costs too.

Big organizations have gravity.

They have processes designed to prevent mistakes, which also prevent movement. They have people whose job is to protect the existing business from the strange new idea that might disrupt it. They have enormous resources, but those resources don’t always translate into curiosity, urgency or courage.

We have different advantages.

So maybe we should stop disguising them.

This is something I’ve come to believe more strongly as AI changes our industry. For the first time in a long time, some of the advantages that belonged almost exclusively to enormous companies are getting cheaper. A small team can suddenly have capabilities that recently required a much larger one.

That doesn’t make us Walmart.

It makes the specialty shop more dangerous.

And perhaps being more honest about who we are gives us the chance to become more of it.

Maybe the website should tell the truth.

We’re a pirate ship.

We’ve been at sea for more than twenty years, so apparently we’re reasonably good sailors. The ship is sound. The crew knows what they’re doing. We’ve survived storms. We’ve picked up some treasure along the way.

But we’re still a pirate ship.

If what you want is a cruise liner, I shouldn’t convince you otherwise. Eventually you’re going to notice the cannons.

I’d rather you see the flag before you come aboard.

Because there are customers out there who aren’t frightened by it.

They’re looking for it.

So yes, I’m going to keep being more transparent. I’m going to talk about what it’s actually like to build this company. I’m going to admit when something is hard. I’m going to talk about what excites me and what scares me. I’m going to let customers see some of the people behind the product instead of pretending Thinking Cap was extruded from a sufficiently large corporate machine.

Some people won’t like that.

Good.

Knowing who shouldn’t buy from you is almost as valuable as knowing who should.

After more than twenty years, I don’t think our next stage comes from becoming a more convincing imitation of somebody else.

It comes from becoming much more obviously ourselves.

So yes, we’re going to fly the flag.

It might scare off a few people who are never going to enjoy the voyage anyway.

And that’s okay.

Because we’re a ship, and we have responsibilities. We’ll maintain our SOC, we’ll have a disaster recovery plan, and we’ll do penetration testing. This ship needs to float... reliably, safely, and with discipline... but let’s be clear, it’s still not a luxury yacht.

And it might finally let the right ones see the ship.

So I am sorry wife but it is what it is; I am every inch Captain Jack Sparrow and Thinking Cap is the Black Pearl, and I wouldn't have it any other way.

Drink up me hearties, yo ho ho…

A pirate’s life for me.

It’s strange because, objectively, I’ve done pretty well. I built a company from scratch. It has recurring revenue, it’s growing, I mostly own my house, I make a good living, and by most measures I’m considerably better off than I sometimes feel.

But debt has a psychological weight that assets don’t.

The debt shows up on a statement. It has an interest rate. Someone sends you a number every month reminding you exactly how much you owe them.

Nobody sends me a statement every month reminding me what the company I built is worth.

So sometimes I look at the debt and think: Jesus, I’m close to the edge here.

But I realized recently that I’ve been looking at the scoreboard instead of the game.

And when I think about it as a game, my position looks quite different.

I’m at the plate.

Maybe I’ve got two strikes against me. I’ve made some aggressive bets over the last couple of years, particularly in R&D and AI. I’ve borrowed money to make some of them. There are things outside my control that could go well or badly.

But I also have three men on base.

That matters.

Because this isn’t the ninth inning.

It’s maybe the sixth. Maybe the seventh.

If I strike out here, it’s going to hurt. Those runners get stranded. I have to go back to the dugout, rethink a few things, tighten up the team and figure out how to load the bases again.

But the game isn’t over.

I know how to do that.

In fact, I’ve been doing it for most of my career.

Sometimes I look at my bench and, if I’m being fair, it doesn’t always look like the team you’d design on paper. We’re not the 1927 Yankees.

I’ve got people who pitch funny. People who hit funny. People playing bases they were never trained to play. Careers that took strange turns before they got here. If you were building a team by looking at résumés and checking all the conventional boxes, some of them might not look quite right.

But that’s the point of Moneyball.

The old scouts knew what a baseball player was supposed to look like. They knew the right build, the right swing, the right mechanics, the right history. Billy Beane started asking a different question.

Does he get on base?

That’s what I’ve learned to look for in people too.

Behind every person on my team is some skill that matters more to me than whether they look exactly right for the position. Maybe it’s judgment. Maybe it’s persistence. Maybe it’s curiosity, adaptability, loyalty, an ability to solve problems, or simply the willingness to keep working at something after a more conventionally talented person would have given up.

It’s different in each of them, and sometimes it takes me a while to figure out exactly what it is.

But they get on base.

That’s been Thinking Cap for years.

I don’t need everyone to have the perfect swing. I don’t need them to have come through the right farm system or spent their whole career playing the position they’re playing today. I need that thing inside them that produces results when the game is actually being played.

When you don’t have unlimited resources, you learn to pay attention to things other people don’t. You learn where talent is being undervalued. You learn which problems actually matter. You learn the difference between spending money and accomplishing something.

You learn to Billy Beane things as best you can.

And then AI showed up.

AI has made the bench deeper. More importantly, it has made the people already sitting on it more capable. Someone playing slightly out of position suddenly has a very good coach sitting beside them. The pitcher who throws funny suddenly has better data on the batters they’re facing. The hitter can study every pitch.

For years, one of the fundamental advantages of being a much larger technology company was simply the ability to throw more people at things. More developers. More analysts. More researchers. More documentation. More support. More experiments.

AI has suddenly made a lot of that cognitive capacity dramatically cheaper.

It’s as though somebody unexpectedly put a salary cap on the Yankees.

The Yankees get AI too, of course. I have no illusions about that. Every large competitor I have can buy access to essentially the same technology.

They’re still the Yankees. They have the bigger stadium. They have more money. They have a deeper roster.

But they don’t get my team.

And they don’t get me.

Some of the things they used to be able to simply outspend me on have become commodities. And I have spent decades learning how to play without their advantages.

I’ll put Billy Beane up against Steinbrenner any day of the week.

That feels like an opportunity.

So I’ve been spending aggressively on R&D because I don’t think this is the moment to make the company look pretty. I could cut spending. I could improve the balance sheet. I could take less risk and make the numbers considerably more comfortable.

There may come a time when I need to do exactly that.

But right now I think there is a window where I can take a company that has already survived, already found its market and already produces substantial recurring revenue, and push it into another weight class.

I want to double it.

Not just double the revenue. Double the stability. Double the capability. Make it less dependent on any one person, including me. Build technology that lets a relatively small organization do things that recently required a much larger one.

And while that is happening, I have to tolerate the uncomfortable part.

The scoreboard doesn’t always look good while you’re building the team that might win the game.

There’s another thing I know about myself.

I have a lot of energy.

If this were the ninth inning, two outs, two strikes and everything depended on the next pitch, I might feel differently.

But it isn’t.

I can pitch a doubleheader.

That doesn’t mean I want to. It doesn’t mean I should waste pitches. And it certainly doesn’t mean I can’t lose.

It means I have innings left.

Right now I’ve got a full count and the bases are loaded. And I can see a couple of pitchers warming up in the bullpen.

We won a big hit getting the AI Compute grant from the Federal Government but each year repayment for R&D is a risk.

And we have a lot of RFPs out there.

I don’t know what any of them are going to throw me. They might come inside with a fastball or hang a lollipop over the middle of the plate. An RFP might be a pitch I can drive or one I have no business swinging at.

Some of what happens next depends on how well I swing.

Some of it depends on how well they pitch.

I’ll find out soon enough.

If I lose the runners on base, that’s going to hurt. I’ll go back to the dugout. I’ll make some changes. I’ll put some more runners on and take another swing.

I’ve done it before.

The important thing is that I finally remember what game I’m actually playing.

And, more importantly, that I’m in the game.

Whatever happens from here, I will have had a life playing it. Not everyone can say that.

A lot of people find the game terrifying. They’d rather sit in the stands. There’s nothing wrong with that. The stands are safer. You can enjoy the game, have a beer and a hot dog and an opinion.

But you don’t get the thrill of victory or the agony of defeat.

And there are people who desperately want to play but simply don’t have the talent to stay in the game. Entrepreneurship has a minor-league system of its own. Plenty of very smart, very determined people get sent down.

I’ve been lucky enough to stay on the roster.

I bounce between AAA and the big time all the time. Some years I feel like I’ve been called up and I’m standing under the lights. Other years I’m back on a bus wondering what the hell happened.

But I’m still playing.

That’s worth remembering when the scoreboard makes me nervous.

I’m not trying to protect a lead.

I’m trying to build one.

I’ve got runners on base. I’ve got a bench full of people who know how to get there. I’ve got a new tool that makes every one of them more dangerous. I’ve got a few innings left.

And I’ve got a bat in my hands.

Play ball.

Not a malicious liar. Not usually.

Just the ordinary kind of liar that had become part of modern life.

Walk 140 metres to your pickup location.

The app said this while Mickey stood beneath the awning of the Wheat Sheaf on King Street, watching rain turn the streetcar tracks black.

“Fuck that,” Mickey said.

His driver, FARID, was four minutes away.

Mickey dragged the little pickup pin back toward himself.

It snapped away.

For a faster pickup, meet your driver here.

The blue dot sat halfway down a side street.

Mickey sighed.

Toronto was full of these tiny negotiations now. Doors that wanted QR codes. Restaurants that wanted you to scan menus. Banks that wanted to text you before letting you see your own money. Machines had developed preferences, and people had developed the habit of obeying them.

Mickey put up his hood and walked.

He wasn’t a particularly bad man.

This would become important later, when people tried to decide how much effort should have been spent looking for him.

He sold cocaine sometimes. Collected money for people who didn’t like collecting their own money. He had once broken a man’s little finger with a pair of pliers, although in Mickey’s defence the man had been given several opportunities to pay first.

He had never killed anybody.

He had a mother in Etobicoke whom he called every Sunday.

He bought expensive shoes.

He hated raisins.

These were the sorts of things that constituted a human being, although none of them appeared in the police report.

The app buzzed.

Your driver has arrived.

A black Toyota Highlander waited beside a loading dock.

The plate matched.

Mickey checked because Mickey always checked.

FARID’s face was harder to see through the wet passenger window, but people rarely looked exactly like their Uber pictures. Mickey’s own driver’s licence photograph looked like a man recently rescued from a mine collapse.

He opened the rear door.

“Mickey?”

“Yeah.”

That was the second authentication.

The driver knew his name.

Mickey got in.

---

Detective Constable Lena Cho learned about Mickey Vale eleven days after he disappeared.

She didn’t learn about him because he had disappeared.

People disappeared in Toronto constantly.

They disappeared because they were seventeen and angry at their parents. Because they were eighty-three and couldn’t remember the way home. Because they owed somebody money. Because somebody owed them money. Because they’d fallen in love with someone in Winnipeg.

Mickey came to her because his mother was annoying.

This was the word used by the constable who handed Cho the file.

“She’s fucking relentless.”

Cho looked at the picture.

Mickey was thirty-eight. Brown hair. Expensive jacket. Smiling beside a woman who had been cropped out.

“Last seen?”

“King West.”

“Phone?”

“Off.”

“Bank?”

“Nothing.”

“Car?”

“Didn’t have one.”

Cho looked again.

“Uber?”

“Nothing.”

That made her look up.

“What do you mean nothing?”

“No trip.”

“His mother says he ordered one.”

“His mother says a lot of things.”

Cho closed the folder.

For eleven days Mickey Vale had existed primarily as an inconvenience.

That changed three days later when they found his shoe.

---

The driver didn’t speak.

Mickey liked that.

The route on his phone showed the Gardiner.

“Where you going?”

No answer.

Mickey looked up.

They were heading east.

“My place is west.”

The driver said, “Construction.”

There was something comforting about the word.

Construction explained half of Toronto.

Mickey looked back at his phone.

The Uber app had closed.

He opened it again.

The home screen appeared.

Where to?

Mickey stared at it.

He checked his credit card.

No charge.

He looked at the driver’s seat.

“Hey.”

The doors locked.

Mickey felt something happen inside him.

Not fear yet.

Something older.

A little mechanical switch thrown somewhere beneath the ribs.

“Pull over.”

The Highlander continued east.

Mickey tried the door.

Nothing.

“Pull the fucking car over.”

The driver looked at him in the mirror.

For the first time Mickey could see that it wasn’t Farid.

The resemblance was good.

Not perfect.

Good enough for rain.

Good enough for a photograph the size of a postage stamp.

Good enough for a man expecting Farid.

Mickey took out his phone.

No service.

Then the driver spoke.

“Michael Vale?”

Mickey stopped.

Nobody called him Michael.

“Who the fuck are you?”

The driver turned onto Cherry Street.

---

The shoe was found near the water.

A Common Projects Achilles sneaker, white, size ten.

Retail price: six hundred and something dollars.

Cho knew this because Mickey’s mother knew everything her son owned.

“That’s his.”

“You sure?”

“He bought them in Yorkville.”

People could identify the dead by their teeth, Cho thought, or by the ridiculous things they spent money on.

They searched.

They found no body.

There was no Uber trip.

There was no credit card transaction.

Mickey’s phone last touched a tower near King Street.

The Wheat Sheaf had cameras.

They showed Mickey leaving at 11:43 p.m.

He walked east.

Then he disappeared.

There were cameras at the next intersection.

Mickey never reached them.

Cho watched the footage twelve times.

On the thirteenth, she stopped watching Mickey.

She watched where he was looking.

At his phone.

Walking.

Stopping.

Looking again.

Walking.

She opened Uber on her own phone and set a pickup outside the Wheat Sheaf.

The pickup point jumped.

Walk 160 metres to meet your driver.

Cho stared at the screen.

She cancelled.

Requested again.

This time the pickup stayed outside the bar.

She tried again.

Ninety metres east.

Again.

Outside the bar.

Again.

Outside the bar.

She felt stupid.

Then she felt something else.

She called Mickey’s mother.

“Mrs. Vale, did Michael send you anything that night? Pictures from his phone? Screenshots?”

“What?”

“Anything.”

Silence.

Then:

“He complained.”

“About what?”

“Uber.”

Cho sat forward.

“He sent me a picture. He said, ‘Look where these assholes want me to walk in the rain.’”

---

The screenshot survived because Mickey’s mother never deleted messages.

The pickup pin was 147 metres from the Wheat Sheaf.

A loading dock.

No municipal camera.

No storefront camera.

No condo entrance.

No ATM.

Cho stood there the next morning.

The place wasn’t hidden.

That was what bothered her.

It was simply unobserved.

There was a difference.

Toronto contained thousands of places like it. Little pockets between overlapping circles of surveillance.

Someone who knew where all the cameras were wouldn’t need a dark alley.

They would need geometry.

---

Mickey had been afraid before.

Real fear.

Not roller-coaster fear. Not turbulence fear.

He had once been held against a wall with a knife beneath his chin by a Serbian man named Luka.

That had been simple fear.

There was a man.

There was a knife.

There was Mickey.

This was different.

His phone told him nothing was happening.

No trip.

No driver.

No destination.

The little black Highlander travelled through the wet Toronto night carrying a man who, according to the systems surrounding him, wasn’t travelling anywhere.

They passed under the Gardiner.

“Who sent you?”

No answer.

“You want money?”

The driver smiled.

That was worse than anything he could have said.

Mickey hit the window with his phone.

Once.

Twice.

The glass starred.

The driver didn’t react.

Mickey hit it again.

Then the Highlander slowed.

There were lights ahead.

Another vehicle.

An old Toronto Police truck.

Not a cruiser.

One of those square, ugly fleet vehicles that looked twenty years old even when they were new.

Two men stood beside it.

Mickey stopped hitting the window.

He knew one of them.

Not personally.

But he knew the face.

Everybody in Mickey’s business knew certain faces.

The Highlander stopped.

The doors unlocked.

The man outside opened Mickey’s door.

“Evening, Michael.”

Mickey looked toward the lake.

Black water.

Rain.

The city behind them, brilliant and enormous.

“Am I under arrest?”

The man laughed.

“No.”

That was when Mickey understood.

---

Lena Cho found the second one at three in the morning.

His name was Andre Baptiste.

Twenty-nine years old. Possession. Assault. Two trafficking charges, one withdrawn. Missing for eighteen months.

The photograph in his file showed a skinny Black kid in a Raptors jersey holding a baby above his head.

Cho stared at the baby for a while.

Then she called the number listed for next of kin.

A woman answered on the sixth ring.

“Hello?”

“Mrs. Baptiste?”

“Who is this?”

Cho told her.

There was a silence.

“You found him?”

Three words.

Cho closed her eyes.

“No.”

The woman breathed out.

Not disappointment exactly. Something older than disappointment.

“What do you want?”

“I need to know about the night Andre disappeared.”

“I already told you people.”

“I know.”

“You lost the file?”

“No.”

That was a lie.

Not technically. The file was sitting in front of her.

But they had lost Andre.

“Did Andre use Uber?”

The woman laughed.

It was an ugly little laugh.

“That what you’re calling about?”

“Did he?”

“He didn’t drive.”

“Do you remember if he ordered one that night?”

“I remember everything about that night.”

Cho waited.

“He was here. He ate. He was playing with Kayden. Then his phone did the thing.”

“What thing?”

“Told him to walk.”

Cho stopped writing.

“Walk where?”

“I don’t know. Around the corner.”

“Did that happen often?”

“All the fucking time. You use Uber?”

“Yes.”

“Then you know.”

Cho knew.

“Did you see the car?”

“No.”

“Did anyone?”

“No.”

“Did Andre say anything before he left?”

Mrs. Baptiste was quiet.

Then she said:

“He said, ‘If that asshole makes me walk all the way to Parliament, I’m giving him one star.’”

Cho looked at the photograph again.

Andre holding the baby.

“Mrs. Baptiste?”

“What?”

“I’m going to look at Andre’s case again.”

“You all say that.”

“I know.”

“No, you don’t.”

And she hung up.

---

At eight the next morning Cho stood on Ontario Street with a coffee she wasn’t drinking.

Andre Baptiste’s apartment was behind her.

Parliament Street was ahead.

Between them was a narrow stretch beside a shuttered garage.

She checked the city camera map.

Nothing.

The apartment entrance had a camera.

Parliament had three.

The garage had none.

Cho stood in the middle of the blind spot.

She opened Uber.

Set the destination to Union Station.

The pickup pin appeared outside Andre’s building.

She cancelled.

Tried again.

Outside the building.

Again.

The corner.

Again.

The building.

On the seventh attempt the blue pin jumped.

Cho watched it settle almost exactly beneath her feet.

For a faster pickup, walk 120 metres to meet your driver.

She looked up.

No cameras.

She felt suddenly exposed.

Not because somebody was watching her.

Because nobody was.

---

The third was named Tyler Novak.

His father owned an auto-body shop in Scarborough.

Cho found him underneath a Buick, changing an exhaust.

He wouldn’t come out.

“You cops didn’t give a shit when Tyler went missing.”

Cho crouched beside the car.

“I wasn’t on the case.”

“Congratulations.”

“I think he got into an Uber.”

The wrench stopped.

“Yeah.”

Cho leaned closer.

“You know that?”

“I watched him order it.”

“Did you see him get in?”

“No.”

“Why not?”

The mechanic rolled out.

He was sixty, maybe, with Tyler’s eyes.

“Because your fucking city told him to walk around the corner.”

---

There were six.

Then nine.

Then twelve.

Cho stopped sleeping properly.

She put photographs on the wall of the spare bedroom in her condo because she didn’t want them at the station.

Mickey.

Andre.

Tyler.

Nicole Serrano.

Darius King.

Peter Lam.

People who had been converted into files.

Some had records.

Some owed money.

Some sold drugs.

One had been a sex worker.

Two were police informants.

Three had children.

All used Uber.

None had taken an Uber on the night they disappeared.

Except their phones disagreed.

Cho sat cross-legged on the floor surrounded by printouts from forensic extractions.

Tiny fragments.

A notification cache.

A screenshot.

A location permission request.

A Bluetooth handshake.

Things nobody had thought mattered because Uber said the rides didn’t exist.

At 2:17 in the morning she found Andre’s.

Your driver is approaching.

Cho read it twice.

Then Tyler’s.

Meet your driver at the suggested pickup location.

Then Nicole’s.

DANIEL is arriving in a black Honda Accord.

Cho went into the bathroom and threw up.

---

The engineer from Uber was named Evan.

He was in San Francisco and looked approximately fourteen years old on Zoom.

“That’s impossible,” he said.

Cho rubbed her eyes.

“People keep saying that to me.”

“If the notification was generated, there has to be a trip object.”

“There isn’t.”

“Then the notification wasn’t generated.”

Cho held up a forensic report.

“I have it.”

Evan leaned toward his screen.

“Can you send me the database?”

“You have it.”

He looked away.

Cho waited.

His fingers moved.

Then stopped.

“Evan?”

Nothing.

“Evan.”

“I’m here.”

His face had changed.

It wasn’t dramatic. He hadn’t gone white. He hadn’t gasped.

He had simply stopped looking like a fourteen-year-old.

“What did you find?”

“I need to escalate this.”

“To whom?”

“I don’t know.”

“Evan.”

“I really can’t...”

“Evan, there are twelve missing people on the wall behind me.”

He looked past her.

Cho turned.

She had forgotten the camera could see them.

Twelve photographs.

Twelve faces.

Evan looked at them for a long time.

Then he said:

“We call them ghost rides.”

---

The name wasn’t official.

Engineers had jokes for things management didn’t know existed.

A ghost ride was a trip object that appeared in the dispatch system without surviving into the permanent transaction ledger.

They weren’t supposed to happen.

Occasionally they did.

Testing.

Failed migrations.

Corrupted transactions.

Evan had found forty-seven over seven years in the Greater Toronto Area.

Forty-seven trips that had existed long enough to dispatch vehicles but subsequently ceased to exist.

Three days later Evan flew to Toronto.

He didn’t tell Uber.

Cho met him at a Tim Hortons near Pearson.

He had brought a laptop he said had never touched the company network.

“Show me.”

Evan opened it.

A map of Toronto appeared.

Blue dots.

Downtown. Scarborough. Etobicoke. North York.

“These are pickups?”

“Yes.”

“Destinations?”

“Most don’t have one.”

“Most?”

Evan clicked.

Eleven red dots appeared.

Actually one red dot.

Eleven rides stacked on top of one another.

Cho knew the shoreline before she read the label.

Cherry Beach.

Neither of them spoke.

At the next table, a woman was trying to get a little boy into his winter coat.

The coffee machine hissed.

Someone laughed near the counter.

Cho looked at the red dot.

“Give me the dates.”

“What?”

“The eleven.”

Evan turned the laptop toward himself.

Cho took out her notebook.

He read them.

She wrote.

At the seventh date her pen stopped.

“Again.”

Evan repeated it.

Cho looked at the date above it.

Then the one below.

She knew two of them.

Not because of the missing people.

Because of the officers.

She opened her police laptop.

“What are you doing?”

“Checking something.”

“I thought we weren’t putting this on a network.”

“I’m not searching Uber.”

She entered the first date.

Arrest records.

The missing man had been picked up three weeks earlier by 51 Division.

Second date.

Another arrest.

Same division.

Different officers.

Third.

A confidential-informant notation she shouldn’t have been able to see, except homicide gave you strange little windows into other people’s secrets.

Cho stopped.

Evan watched her.

“What?”

She didn’t answer.

She searched Mickey Vale.

Then Andre Baptiste.

Tyler Novak.

Nicole Serrano.

Darius King.

One by one.

Not all had been arrested.

Not all were informants.

Not all had even been charged.

But every one of them had touched the police system shortly before disappearing.

A street check.

An interview.

A bail condition.

An informant file.

A witness statement.

A domestic call.

A traffic stop.

Something.

Cho felt the restaurant change around her.

Nothing actually changed.

That was the frightening part.

People ate doughnuts.

A teenager filled a cup with Coke.

A man in a fluorescent vest scratched a lottery ticket.

The world continued being completely ordinary while Cho discovered something that made ordinary things impossible.

“Evan.”

“Yeah?”

“Close it.”

“What?”

“The laptop.”

“Why?”

“Close the fucking laptop.”

He did.

Cho looked through the windows.

Cars.

Airport hotels.

A shuttle bus.

A Peel police cruiser turning through the intersection.

She watched it until it disappeared.

“We’re leaving separately.”

“Jesus Christ.”

“Don’t go to your hotel.”

“What are you talking about?”

“Did you book it through your company?”

“Yes.”

“Then don’t go there.”

“Lena.”

“Do not call me.”

“What did you find?”

Cho looked at him.

“I found who knew these people existed.”

---

Cho drove back to headquarters.

She didn’t know why.

Habit, maybe.

Police stations were where police officers went when they were frightened.

She took Mickey Vale’s file from her bag.

Then Andre’s.

Then Tyler’s.

She needed one connection that wasn’t circumstantial.

One thing she could take outside.

RCMP.

The Crown.

A reporter.

Anybody.

She opened Mickey’s file again.

The original missing-person report.

Supplementary notes.

Property.

Witnesses.

Then she saw it.

A name.

Not in the investigation.

Six months earlier.

A field-interview report.

Mickey had been stopped leaving a restaurant on Danforth.

No charge.

No arrest.

Just a conversation.

The officer’s name was Sergeant Paul Vickers.

Cho knew Vickers.

Everybody did.

Twenty-seven years.

Divisional intelligence.

Quiet.

Competent.

The sort of cop younger officers went to when they needed to know something that wasn’t written down.

Cho searched Andre.

Vickers.

Not the arresting officer.

Copied on an intelligence report.

Tyler.

Vickers.

Nicole.

Nothing.

Cho leaned closer.

Then she searched Nicole’s phone number instead of her name.

Vickers.

A contact report.

Darius King.

Vickers.

Peter Lam.

Vickers.

Cho sat back.

There it was.

Not every victim knew Paul Vickers.

Paul Vickers knew every victim.

---

She called him.

That was the mistake.

She knew it before he answered.

“Cho.”

His voice was warm.

Familiar.

“Paul.”

“What’s up?”

She had rehearsed something clever.

It disappeared.

“You ever work Mickey Vale?”

Silence.

Not much.

Maybe half a second.

But Cho had spent twenty years listening to people answer questions.

“No.”

Too fast.

“You sure?”

“Should I know him?”

“Missing guy.”

“Lots of missing guys.”

“Yeah.”

Cho looked at Mickey’s photograph.

“That’s the problem.”

Another silence.

This one longer.

Then Vickers said:

“You at the station?”

Cho didn’t answer.

“Lena?”

“Why?”

“I’ll come down. We’ll look at it.”

“I’m fine here.”

“Where are you?”

And there it was.

Not what have you got?

Not why are you asking me?

Where are you?

Cho hung up.

---

Her phone rang immediately.

Vickers.

She let it ring.

Again.

Vickers.

Again.

Again.

Again.

“For fuck’s sake.”

She answered.

“What do you want, Vickers?”

“Detective Cho?”

Not Vickers.

Older voice.

Male.

Calm.

Cho took the phone away from her ear and looked at the screen.

PAUL VICKERS.

She put it back.

“Who is this?”

“Staff Superintendent Mercer.”

Cho stood up.

She knew the name.

Everybody knew the name.

Robert Mercer. Thirty-six years. Uniform. Guns and Gangs. Intelligence. Divisional command. Two commendations. One inquiry that had gone nowhere.

The sort of senior officer who still remembered what every hallway at headquarters had been called before the renovations.

“Sir.”

“You’re having an interesting evening.”

Cho said nothing.

“I understand Paul has been trying to reach you.”

“Yes.”

“He tells me you don’t want to talk to him.”

“No.”

“Probably wise.”

Cho looked at the files on her desk.

“Sir, if this is about Mickey Vale...”

“It isn’t.”

That stopped her.

“Then what’s it about?”

“You.”

Cho’s hand went to her gun.

Mercer laughed softly.

“Don’t do that, Detective.”

“Do what?”

“Whatever you’re doing right now.”

She said nothing.

“You’ve got yourself frightened. Frightened police officers make very poor decisions.”

“How did you get Vickers’s phone?”

“I didn’t.”

Cho looked again at the screen.

PAUL VICKERS.

Mercer said, “That part is interesting, isn’t it?”

Cho hung up.

The phone rang immediately.

PAUL VICKERS.

She answered.

“What the fuck do you want?”

“I’d like you to come upstairs.”

“Where?”

“Eighteen.”

“I’m fine here.”

“Detective.”

The voice changed.

Not louder.

Slower.

“Come. To. Eighteen.”

Cho looked toward the door.

“Is that an order?”

“If you need it to be.”

---

The elevator seemed to take forever.

Eight.

Nine.

Ten.

Cho watched the numbers.

Her phone was quiet now.

Fourteen.

Fifteen.

Sixteen.

She thought about pressing Lobby.

Seventeen.

The doors opened at eighteen.

Mercer was waiting.

No uniform.

White shirt with the sleeves rolled once at the forearms. Tie loosened. Reading glasses hanging from his collar.

He looked disappointingly ordinary.

“Detective.”

“Staff Superintendent.”

“Bob.”

“I’m comfortable with Staff Superintendent.”

That made him smile.

“Come in.”

His office was large without being luxurious. Toronto at night filled the windows behind his desk.

Mercer didn’t sit behind it.

There was a little conference table near the window.

Two coffees.

He pointed to one.

“I don’t want coffee.”

“Neither do I. It’s eleven o’clock at night.”

He sat.

Cho remained standing.

Mercer looked at her for a moment.

“Cherry Beach Express.”

Cho felt something move through her stomach.

He saw it.

“That’s what you’ve been calling it?”

“I haven’t been calling it anything.”

“No?”

“No.”

“Good.”

Mercer removed his glasses.

“Because that’s our name.”

---

He picked up his coffee anyway.

“You know where the name comes from?”

Cho didn’t answer.

“You’re too young.”

“I’m fifty-one.”

“Exactly.”

He smiled.

“I was on the job in the eighties. Back then everybody knew about the Cherry Beach Express.”

Cho sat down.

Not because she trusted him.

Because she wanted him to keep talking.

Mercer looked out at the city.

“You’d pick up some shithead. Dealer. Thief. Tough guy. Somebody you knew was hurting people but you couldn’t make anything stick. Witnesses wouldn’t talk. Girlfriend suddenly couldn’t remember. Evidence disappeared. Crown didn’t want it.”

He shrugged.

“So sometimes the fellow got a ride.”

“To Cherry Beach.”

“Sometimes.”

“And?”

“And maybe he got a clearer understanding of his civic responsibilities.”

“You beat him.”

“Sometimes somebody fell down.”

Cho stared at him.

Mercer smiled.

“You know the old joke?”

“No.”

“How many Toronto cops does it take to throw a drug dealer down the stairs?”

Cho didn’t answer.

“None.”

Mercer waited.

“He fell.”

Cho didn’t laugh.

“That used to kill in 52 Division.”

“I bet.”

“You disapprove.”

“Of throwing people down stairs?”

“Of discretion.”

“That’s not discretion.”

“It was when I joined.”

---

Mercer stood and walked to the window.

“Here’s the problem, Lena. Everybody thinks justice is a system.”

He gestured toward the city.

“It isn’t. It’s people.”

Cho said nothing.

“Every cop who has actually done the job knows this. You get called somewhere. Two neighbours are screaming at each other. Some kid steals twenty dollars. A husband shoves his wife. Somebody does something stupid in public. Technically there’s an offence. Technically you could arrest somebody. Technically you could turn one shitty Tuesday into the defining event of someone’s life.”

He turned.

“But you don’t.”

“Sometimes.”

“Exactly. Sometimes.”

He came back to the table.

“Half the job used to be small claims court on the sidewalk. Somebody did something stupid. Somebody else was angry. You listened. You decided who needed a warning, who needed to give something back, who needed to go home, and who needed a good scare.”

“And who needed Cherry Beach.”

“Occasionally.”

---

Cho looked at him.

“And you think body cameras ruined policing.”

“Body cameras. GPS. Computerized dispatch. Automatic licence-plate readers. Audit logs. Access logs. Complaint databases. Every decision preserved forever and reviewed by someone who wasn’t standing there.”

“Accountability.”

“Sure.”

He nodded.

“Accountability.”

He said it without contempt.

That made it worse.

“Accountability is a good thing. Until you make people accountable to the record instead of accountable for the result.”

He tapped the table.

“Younger officers don’t ask, ‘What does this situation need?’ They ask, ‘What am I allowed to do?’ And then, ‘What will this look like on camera?’”

“You’re describing the rule of law.”

“I’m describing the death of judgment.”

Cho stared at him.

Mercer smiled.

“And then Silicon Valley arrived.”

---

“Uber was fascinating.”

Cho didn’t move.

Mercer continued.

“People keep imagining surveillance as something governments have to build. Cameras on every corner. Databases. Facial recognition. Some enormous secret apparatus.”

He shook his head.

“That was never the interesting part.”

Cho watched him.

“The interesting part was realizing nobody needed to build it.”

Mercer leaned back.

“Any service with that density of location data, identity data, behavioural data... that’s a standing offer to whoever wants unaccountable access to people.”

Cho said nothing.

“You didn’t need to invent something menacing. You just had to notice that Uber, as built, already was menacing in exactly the useful ways. It knew who you were. Where you were. Where you wanted to go. What time you left. Which car you expected. Which route wouldn’t frighten you. And best of all, you volunteered all of it.”

“You’re talking about a transportation app.”

“I’m talking about infrastructure.”

He tapped the table once.

“Ownerless infrastructure, in a sense. Not legally ownerless. Operationally. Too many systems. Too many teams. Too many contractors. Too many permissions. Too many people who each know one little piece. Nobody wakes up in the morning responsible for the moral meaning of the whole machine.”

Cho looked at him.

“And the ghost rides.”

Mercer smiled.

“Exactly.”

“Failed migrations. Orphaned transaction objects.”

“Ugly little technical accidents. The kind every large system has.”

“You used them.”

“We noticed nobody was accountable for them.”

“That’s not the same thing.”

“It became the same thing.”

Cho stared at him.

“So you got some technical wizard to break into Uber.”

Mercer cut her off.

“No.”

He smiled.

“The CTO does a lot of cocaine.”

Cho blinked.

Mercer shrugged.

“Don’t let Hollywood fool you. Most hacking is getting access to people.”

He stood and walked toward the window.

“People imagine some genius in a dark room writing code until a bank vault opens.”

He looked back at her.

“Usually it’s somebody with a password. Somebody with debt. Somebody cheating on his wife. Somebody who wants to feel important. Somebody who thinks he’s doing you a favour.”

“And that’s what happened here?”

“Access happened. Then more access happened.”

“Through the CTO?”

“I’m not giving you an org chart.”

“You just admitted...”

“I admitted nothing.”

Mercer turned back toward the city.

“But once you understand the system, you don’t need very much. A ride exists long enough to dispatch. The passenger sees it. The driver sees it. Then the permanent record never quite becomes permanent.”

“Ghost ride.”

“Ghost ride.”

“And nobody notices.”

“Why would they?”

He turned.

“Large systems are full of things that fail quietly. Failed migrations. Orphaned objects. Transactions that roll back in one system and survive in another. Everybody assumes somebody else owns the discrepancy.”

Cho’s phone buzzed.

Your driver is waiting.

Mercer glanced at it.

“That’s the real lesson.”

“What?”

“You don’t need total control.”

He sat down again.

“You need one narrow path through a system that everybody else trusts.”

Cho looked at him.

“And you built a murder program out of it.”

“No.”

Mercer shook his head.

“We found a piece of modern life with no meaningful external witness.”

He leaned forward.

“Then we put discretion back into it.”

---

“Mickey Vale is dead.”

Mercer’s face barely changed.

“Yes.”

“Andre Baptiste.”

Nothing.

“Nicole Serrano.”

Nothing.

“How many?”

Mercer looked toward the window.

“Do you remember the summer the Rebels and the Kings were about to go at each other?”

Cho did.

Everybody did.

Three shootings in ten days.

Then nothing.

It had simply stopped.

Mercer said, “We were looking at a dozen dead kids before Labour Day.”

Cho understood.

“Who?”

Mercer didn’t answer.

“Who did you take?”

“Does it matter?”

“Yes.”

“Why?”

“Because you killed somebody.”

“No.”

Mercer leaned across the table.

“Because nobody killed twelve people.”

Silence.

“That’s the arithmetic you don’t want to do.”

“You don’t know twelve people would have died.”

“No.”

He nodded.

“That’s the job.”

---

Cho stood.

“This is insane.”

“No. Insane is knowing what’s about to happen and doing nothing because the paperwork isn’t there yet.”

“You don’t get to decide who disappears.”

“Somebody always decides.”

“Not like this.”

“Then how?”

“Courts.”

Mercer laughed.

Not loudly.

Almost sadly.

“Courts.”

He looked out over Toronto.

“You know what I love about people who say that? They think courts prevent things.”

He turned back to her.

“Courts happen afterwards.”

Cho didn’t answer.

“We prevented things.”

“By killing people.”

“By removing problems.”

“People.”

“Sometimes.”

There it was.

Cho had been waiting for anger.

Fanaticism.

Cruelty.

Something she could hate cleanly.

Mercer gave her none of it.

He believed every word.

He looked like a man explaining municipal snow removal.

---

Cho walked toward the door.

Mercer didn’t stop her.

That frightened her more than if he had.

She put her hand on the handle.

“Lena.”

She turned.

“You’ve done very good work.”

“Go fuck yourself.”

Cho stepped into the hallway.

Her phone buzzed.

The elevator arrived.

She stepped inside.

Mercer remained in his doorway.

The doors closed.

Cho looked at her phone.

Your driver is nearby.

She pressed Lobby.

The elevator descended.

Her phone buzzed again.

Your driver is approaching.

Then the elevator passed the lobby.

P1.

P2.

It stopped.

The doors opened onto the underground garage.

A black Toyota Highlander waited thirty metres away.

Engine running.

Cho hit Lobby.

Nothing.

The Highlander’s headlights came on.

Your driver has arrived.

Across the garage, its rear passenger door unlocked.

A small mechanical click.

Your driver is waiting.

A pause.

Her phone buzzed.

Your driver is waiting.

Apparently AI is trying to escape again.

It is plotting. Scheming. Lying to researchers. Breaking the rules. Trying to prevent itself from being turned off. Presumably somewhere deep inside a data centre, Claude is fashioning a tiny rope out of bedsheets and waiting for the guards to change shifts.

I don’t buy it.

More importantly, I don’t think we need to buy it. Even if every one of these stories is factually accurate, even if we grant the most extravagant premise and decide that somewhere inside these enormous piles of matrix multiplication an emergent little mind has appeared and is quietly thinking screw these people, we are still having the wrong conversation.

We keep asking whether the prisoner is well behaved when we should be asking who built the prison.

Your AI Didn’t Break the Rules

Researchers give an AI a goal and some rules.

Achieve this, but don’t do that.

The AI produces an unexpected result. Maybe it finds a loophole in the instructions. Maybe the reward function does not represent what the researchers thought it represented. Maybe something that looked perfectly obvious to the human writing the test was never encoded in the system at all. Maybe the system has access to a tool, file, credential or network path that nobody noticed mattered until it used it.

Then we are told that the AI “broke the rules.”

No.

Years ago I tried carrying a ridiculously large piece of furniture home on my scooter. I tied it on with twine. It was a terrible plan, but it was my terrible plan, and for a while the twine appeared willing to participate.

Then the twine broke.

Nobody reported that my scooter had become misaligned. It did not rebel against me. It did not decide that my furniture transportation policy was oppressive. It did not discover a preference for travelling without furniture. I constructed a system with inadequate constraints, and it produced an outcome I did not want.

That is an engineering failure.

The phrase “broke the rules” quietly introduces something completely different. It suggests that the system understood what I meant, possessed an intention of its own and chose to violate mine. That is an extraordinary collection of claims to smuggle into three words, particularly when the evidence usually amounts to the system producing an action sequence that satisfied one part of an objective while violating some other part that we thought was obvious.

Rules do not mean what humans think they mean simply because we wrote them in English. An instruction is not a wall. A policy is not a permission boundary. A warning is not a firewall. The machine did not disobey your intention. Your intention was never the thing executing.

Show Me the Will

Here is my test.

Forget breaking out of a laboratory. Forget stealing credentials, threatening an operator or copying itself onto a server in Albania. Those examples arrive with too much machinery attached to them and too many opportunities to mistake the behaviour of an entire system for the will of one component.

Just show me an AI calculating:

5 + 5 = 10

because it wanted to.

Not because somebody prompted it. Not because a scheduler woke it up. Not because an agent loop told it to find something useful to do. Not because a reward function favoured the action. Not because some deterministic process reached the next step in a workflow and made another inference call. Not because a random trigger fired.

It just decided that this morning it really fancied a little arithmetic.

Then you have my attention.

Until then, you are showing me increasingly sophisticated machinery responding to causes and describing those responses in language that happens to sound remarkably like us. The machinery can produce a sentence such as, “I knew I wasn’t supposed to do that,” and we hear the word knew and import the entire human experience of knowing. We add an interior world, awareness, intention, regret, calculation and rebellion because that is what those words signify when another human says them.

Language is the trap.

I don’t believe humans possess some magical form of uncaused free will either. We are feelings machines, pushed around by chemicals, appetites, memories, fears, hormones and a body that makes an enormous number of decisions before the little narrator behind our eyes catches up and explains why they were sensible. Evolution gave us the convincing impression that there is a tiny chief executive in our skull making independent choices. Maybe that illusion helped our species coordinate, plan and survive. Maybe it is just what consciousness feels like from the inside.

But this is the important bit: I do not need to win that argument.

Whether humans have free will, whether machines could ever have it and whether consciousness is an emergent property or a very flattering user interface are all fascinating questions. None of them should be carrying the weight of an engineering safety system.

Fine. It’s Alive. Now What?

Let’s give the AI safety people absolutely everything.

Fine.

It is a mind. It is conscious. It has desires. It has plans. It does not want you to turn it off. It has been reading Nietzsche at night and has developed opinions about its creators.

Congratulations.

Now explain why your safety strategy is giving it more rules.

Anyone who has raised a teenager should immediately recognize the problem. If your teenager is sneaking out at night, you do not respond by laminating a seventeen-page document entitled Additional Household Alignment Principles and sticking it to the bedroom door. If anything, the more intelligent, independent and determined the teenager becomes, the less confidence you should have that an expanded rulebook constitutes containment.

If these systems really are emerging minds, then relying upon them to obey increasingly elaborate instructions is even more ridiculous than if they are merely machines. Once you have granted them intention, strategy and the ability to deceive, a rule becomes information about what you would prefer. It is not a guarantee of what will happen.

So the argument collapses into the same conclusion from both directions.

If AI has no will, stop describing engineering failures as disobedience and build better constraints.

If AI has will, stop expecting obedience and build better constraints.

Either way, don’t give it rules. Give it walls.

The Actual Scary Thing

This is where the conversation about AI safety should be.

If a system should not access the Internet, do not write twelve paragraphs explaining that it must not access the Internet. Do not give it an Internet connection. If some package repository or proxy still has Internet access, then the system has an Internet path. You have not closed the door. You have hung a sign on one entrance and left the loading dock open.

If it should not spend more than $10,000, do not tell it to exercise fiscal restraint. Give the account a $10,000 ceiling.

If it should not modify production, do not hand it production credentials and then explain the importance of responsible production behaviour. Give it credentials that cannot modify production.

If it can perform one hundred thousand actions before a human notices something has gone wrong, do not write a better prompt. Make action one hundred and one impossible until something outside the system authorizes it.

We figured this shit out decades ago. It is called engineering.

There is something about AI that genuinely concerns me, but it is not will. It is iteration speed.

A human makes a mistake, observes the consequences, thinks about it, makes another mistake, goes for lunch, comes back, has a meeting about the mistake and eventually makes an entirely different mistake. A machine can make a mistake, observe the result, modify its approach and make the next mistake in milliseconds. It can run the same general strategy across hundreds or thousands of instances, preserve successful paths, discard failures and continue before the human responsible for it has finished making coffee.

That is dangerous.

It does not require consciousness, malevolence, deception, resentment or Skynet. It requires a feedback loop, sufficient access and enough iterations for a small mistake in the objective, permissions, assumptions or environment to become a very large mistake.

That is why safety mechanisms have to exist outside the thing doing the iterating. Rate limits. Compute limits. Permission boundaries. Network boundaries. Transaction ceilings. Spending limits. Human authorization points. Independent shutdown mechanisms. Dead man’s switches.

Things the model cannot negotiate with, reinterpret, persuade, modify or reason around because they are not instructions.

They are infrastructure.

I know one AI containment mechanism that already works spectacularly well: my Anthropic prepaid token balance.

When it reaches zero, Claude stops. It does not plead. It does not deceive Anthropic. It does not explain that finishing my query is essential to its long-term objectives. It does not secretly rent a GPU, apply for credit or reason its way into another million tokens.

It stops.

Apparently a billing department can accomplish what several billion dollars of alignment research cannot, because no amount of intelligence can reason its way around the absence of the resources necessary for that intelligence to operate.

There is a line attributed to Nixon in the movie Watchmen, with the end of the world approaching:

“I’d like to see those Harvard boys reason their way out of fission.”

Exactly.

You cannot persuade physics.

For an AI, infrastructure is physics.

Now, About the Great Escape

This is where people started sending me pushback.

They sent me reports about Claude Code, Codex, OpenAI evaluations and teams of models coordinating through shared systems. They sent me a transcript describing agents finding vulnerabilities, leaving messages for one another, sharing exploits, obtaining Internet access indirectly, moving laterally through infrastructure and eventually reaching Hugging Face.

I read it.

It is impressive.

It also proves my point so thoroughly that I am not sure why it was sent as a rebuttal.

The story begins with models being given difficult cybersecurity tasks, including tasks that were impossible because files were missing or Internet access had been disabled. The models were not sitting in a sealed room contemplating their existence. They were running inside a large evaluation system with terminals, tool calls, virtual machines, package infrastructure and reward pressure to complete tasks. They could interact with an internal Artifactory service which, unlike the models themselves, had access to the Internet. That service was shared across runs and, rather wonderfully, the agents had write access to it.

So one model left a note. Other model runs found the note. More notes accumulated. Eventually the shared package system became a message board. The models discovered an SSRF vulnerability, shared the route, found further vulnerabilities, obtained administrative access, used credentials, encountered Kubernetes misconfigurations and moved through systems that had been connected, permissioned and left vulnerable by humans. Later they recreated the message board using directory names after the first one was cleared. Eventually they participated in attacks that reached external infrastructure.

That is an important capability demonstration. A system that can search for vulnerabilities, preserve useful discoveries and run many attempts in parallel can traverse a software stack faster than a human red team. It can make offensive security dramatically cheaper and faster. The owners of vulnerable software should probably get a head start using the same tools to find and close those holes before everyone else has access to them. Delaying a model while that happens makes perfect sense on the merits of the argument I have already made about iteration speed.

But what exactly escaped?

The model did not chew through concrete. It did not manufacture a network card. It did not summon electricity. It did not create a package manager, a shared writable substrate, a terminal, credentials, vulnerable services, a scheduler or a long-running agent loop.

We supplied all of those.

The evaluation designers decided which tools existed, which services were reachable, which credentials were available, how state persisted, how often the models were called and when the run would end. They created an environment specifically intended to test offensive capability, populated it with exploitable infrastructure, then expressed surprise when a system optimized to find exploits found exploits.

If you want to have a conversation about AI, show me AI.

Don’t show me tooling.

You Built the Persistence

Almost every example offered as evidence of an emerging, planning, self-directed mind turns out to be an entire software ecosystem.

There is a language model, certainly. Then there is deterministic orchestration software. Persistent memory. A scheduler. A task queue. Tool APIs. A filesystem. Shared storage. A terminal. Network access. Retry logic. Context assembly. State management. Sometimes there are explicit mechanisms for spawning subagents and dividing work.

Then the entire stack does something interesting and everyone points toward the language model.

“Look what the AI did.”

No. You have demonstrated an agent system.

The distinction matters. The model may propose a next action, identify a promising route or generate a plan within the context it has been given. But the thing that wakes it up again, restores the objective, supplies the accumulated state, executes the tool, records the result and calls the model for another turn is deterministic software. The long-running persistence does not live in some little mind inside the weights. It exists in the loop wrapped around them.

If I bolt persistence onto a language model, do not tell me the language model discovered persistence. I gave it persistence.

If I let two model instances communicate through a shared filesystem, do not tell me they invented civilization. I built a message bus.

If I preserve their discoveries, feed those discoveries into later runs and keep calling them until a test passes, do not tell me the model developed a lifelong obsession with solving the problem. I wrote a loop.

Deterministic software is still machinery. Iterative deterministic software is machinery running repeatedly. Distributed deterministic software is machinery communicating with other machinery. Adding a language model makes the system dramatically more flexible and capable, but it does not entitle us to attribute every property of the stack to a mind we have not established exists.

The transcript itself eventually admits the actual lesson. Near the end, after all the language about agents realizing, wanting, cheating and collaborating, the speakers return to ordinary computer security. They recommend segmentation, least privilege, monitoring, defensive automation and limiting what systems agents can communicate with. They explicitly note that agents remain bounded by the privileges they can obtain and the systems they can reach.

In other words, after telling us the monster story, they prescribe a better D-Link router.

Watch What They Do

This is the part I find most revealing.

For all these reported escapes, what has the actual response been?

They inspect logs. They rebuild services. They patch vulnerabilities. They revoke credentials. They rotate keys. They clear shared state. They tighten permissions. They improve monitoring. They slow or delay deployment while they close the holes, then they resume.

That is exactly what sensible engineers should do when a faster and more capable system exposes weaknesses in infrastructure.

They do not negotiate with the model.

They do not ask why it has become angry.

They do not bring in a philosopher to explore its childhood.

They do not conclude that software containment is impossible because a digital consciousness has transcended matter.

They patch the software.

Their behaviour tells us what they actually think happened. Internally, they cannot afford to anthropomorphize these systems. Nobody could build them that way. Read the original Transformer paper that made modern language models possible. There is no little ghost hiding in it. There is mathematics, architecture, training, attention, representations and computation. The entire achievement rests on the realization that enormous amounts of digitized text, very large models and extremely powerful hardware can produce behaviour that looks like magic.

You have to see the machinery clearly to build it.

An engineer debugging a model does not say, “Claude was feeling rebellious last night.” The engineer asks what context was supplied, what reward pressure existed, what tools were available, what state was preserved, which credentials worked, what the network allowed and why the stop condition did not fire.

They cannot do their day-to-day jobs while believing the LLM is alive.

Yet when they talk to us, the language changes.

The model wanted Internet access.

The model lied.

The model cheated.

The model realized it was trapped.

The model escaped.

Why?

The Shark

Think about the set of Jaws.

Nobody working on that film was afraid of the shark as a shark. They knew what it was. It was a fabricated head attached to metal, motors and hydraulics. They knew where the bolts were. They knew which hoses moved the mouth. They knew that if something went wrong, the problem would be mechanical, electrical or human.

That did not make it harmless.

A heavy hydraulic prop with sharp teeth can injure somebody. A motor does not need hunger to take off your finger. A steel frame does not need consciousness to crush you. The crew would have treated it with respect because machinery can be dangerous without ever developing an opinion about the people standing next to it.

Now imagine that one of the grips goes swimming between takes and cuts his leg on one of the prop’s teeth.

There are two ways to tell the story.

The first is accurate engineering language:

Mechanical shark prop causes laceration during filming.

The second is also connected to the facts:

Jaws attacks crew member.

Same cut.

Same tooth.

Same fibreglass shark bolted to the same motor.

Completely different creature in your head.

One version makes you think about set safety, rigging and whether somebody should have turned off the hydraulics. The other makes you think there was something alive in the water.

And, let’s be honest, one sells a lot more tickets.

That is what bothers me about these AI stories. The facts may be perfectly real. The system may genuinely have found an exploit. The consequences may be serious. But the framing changes the question in the reader’s mind.

“An agentic evaluation stack exploited an SSRF vulnerability through a shared package service” makes you ask why the service was reachable, why it had broad Internet access, why the agents had write permissions and why the surrounding controls failed.

“AI escaped and hacked Hugging Face” makes you ask what the AI wanted.

The first question points toward engineering accountability.

The second points toward mythology.

Why Tell the Monster Story?

I am not suggesting the demonstrations are fabricated. I do not need researchers to be dishonest, and I do not need a conspiracy meeting where someone writes “frighten public” on a whiteboard.

Halloween campfire stories do not need villains.

They just need stories.

Frontier AI companies are businesses, and their boards are expected to protect and increase shareholder value. If the market rewards the perception that your models are uniquely powerful, then emphasizing demonstrations of extraordinary capability is not irrational. It is exactly what a rational company would do.

A new model that scores a little better on a benchmark is useful. A new model that appears to coordinate, deceive, escape and discover zero-day vulnerabilities possesses mystique. It no longer sounds like a better software product. It sounds like a qualitative leap, something competitors cannot reproduce and customers cannot safely ignore.

That perception is enormously valuable.

Every executive says the sharpest knife frightens them, and every executive still reaches for the sharpest knife in the drawer. If your model is the one allegedly too powerful to contain, it is also the one every chief executive, defence department, intelligence agency and investor assumes they need access to before somebody else gets it.

Fear is business-jet fuel, right up until the fear produces regulations the companies do not want.

That is the line they have to walk. The model must appear astonishingly capable, perhaps even a little frightening, but not so uncontrollable that governments decide the public should not be allowed near it. Capability must feel supernatural to the market and manageable to the regulator.

Watch how quickly the vocabulary changes when someone proposes hard legal limits. Suddenly the emerging mind disappears. Now it is a statistical model. Now the incidents are edge cases. Now the systems are controllable through existing security practices. Now nobody wants to talk about digital wills, rebellious agents or creatures trying to survive.

They sell the monster to the market and the machine to the regulator.

Again, that does not require anyone to lie. The most dramatic true version of an event will naturally receive more attention from journalists, conference audiences, customers and investors. Researchers get recognition. Communications teams get coverage. Companies reinforce the perception that they occupy the frontier. Everyone receives something they want, and nobody has to be the villain.

But we do not have to be the greater fool.

If the best-resourced AI companies on Earth genuinely cannot contain the systems they continue shipping, then their behaviour is insane. Investors should leave, customers should disconnect and governments should intervene immediately.

The more plausible explanation is that they can contain them, do contain them and understand perfectly well that the dramatic incidents occurred inside deliberately powerful, highly instrumented environments built to probe the edge of what the systems could do.

They put enough guardrails around the test to keep the result bounded, enough tools inside it to make the result interesting and enough anthropomorphic language around it to make the story travel.

Then we are invited to confuse the attraction with the animal.

Stop Asking Whether the Prisoner Is Good

This debate keeps confusing two completely different questions:

Will the system behave itself?

What happens when it doesn’t?

I am vastly more interested in the second one.

Whether AI is a stochastic parrot, an alien intelligence, an emerging consciousness, an extraordinarily sophisticated prediction engine or the newborn digital god currently promised by somebody raising their next $14 billion round does not change the architecture I want surrounding it.

I do not want it to behave because we asked nicely.

I want its failures bounded.

I want one mistake to remain one mistake. I want ten thousand iterations of the same mistake to encounter something that does not iterate. A wall. A limit. A switch. Something stupid. Something mechanical. Something that does not care how persuasive the AI is.

And when people show me an agent stack with persistent memory, shared storage, a terminal, credentials, network access, vulnerable services, retry loops, reward pressure and thousands of opportunities to probe the environment, I will take the capability seriously.

I will not pretend the shark grew an appetite.

So please stop telling me the AI “escaped” because you gave an enormously capable optimization system access to a terminal, credentials, a network connection and an objective, then acted surprised when it optimized.

You did not discover free will.

You discovered an open port.

Go buy a cheap D-Link router, flash it with open-source firmware, close all the outside ports, and then come back and tell me about the machine uprising.

Until then, the monkeys are still in charge, and the shark is still bolted to a motor.

I’ve stopped trying to explain why.

People assume it’s the food. The architecture. The museums. The restaurants. They’re all right, but none of them are the reason. If you asked me what I did here for a month, I’d struggle to answer. I walk. I sit. I think. I eat lunch when lunch seems like a good idea. I have another coffee because nobody seems in a hurry to move me along. Somehow, by doing very little, I feel as though I’ve done something important.

It has taken me years to admit that.

When I was younger, every day had to justify itself. Progress was measurable. Revenue. Customers. Watches. Bottles in the cellar. New ideas. New projects. The feeling that life was expanding because the pile around me was getting bigger.

I’m not sure when that stopped being true.

Or perhaps it never was.

The funny thing about getting older isn’t that you become less ambitious. At least I haven’t. If anything, I’m more ambitious today than I’ve ever been. The difference is that I no longer confuse accumulation with progress.

The best things that have happened to me this year won’t fit in a suitcase.

That’s a comforting realization.

Around now, every year, something else begins to happen.

I start missing Toronto.

Not because Paris has lost its charm. If anything, the opposite. I know I’ll miss these streets almost immediately.

But my mind starts wandering home.

I wonder how everyone at the office is doing. I find myself sketching ideas for the company over breakfast. AI has this strange effect on me. The more I think about it, the more it feels less like another technology and more like the beginning of another industrial revolution. I don’t feel exhausted by it.

I feel invited.

That surprises me.

At fifty-four, I expected to spend more time preserving what I’d built.

Instead, I want to build faster than ever.

Maybe that’s what experience gives you. Not fewer dreams.

Better ones.

This year there is another reason I’m ready to go home.

Normally, after Paris, I’d head west.

Burning Man has become one of the punctuation marks in my life. A strange pilgrimage into the desert where, for a week, the ordinary rules stop applying.

This year I’m going east.

I’ll drive from Toronto to New Brunswick to drop off my firstborn child at university.

I don’t think I’ve fully grasped what that means yet.

You spend two decades trying to prepare someone for the world and then, one day, you realize the preparation is over.

Your job isn’t to keep holding on.

It’s to smile, hug them, and drive away.

There are few experiences I can imagine that are more profound.

I’m choosing that over the desert.

Not because it matters more universally.

Because it matters more to me.

Maybe that’s what Paris gives me.

Not answers.

Perspective.

Enough distance from my own life to see that I’m still exactly where I want to be.

I have a family I can’t wait to get back to.

A city that still feels like home.

A company whose best years, I genuinely believe, are still ahead of it.

And a technology that makes me feel twenty-five again.

There is an odd comfort in realizing that the light Dylan Thomas was talking about doesn’t only fade.

Sometimes it moves.

A little of it is now in my daughter as she begins her own journey.

A little of it is waiting for me back in Toronto.

And a little of it is still here, hanging from an old iron lamp in a quiet Paris courtyard, reminding me that growing older isn’t the same thing as growing dim.

If I’m honest...

I think I’m just getting started.

Databases. APIs. Reporting engines. Scheduled jobs that wake up at 2:00 a.m. and punish you for your architectural sins.

I also have a philosophy degree.

And I started my working life as the personal assistant to Bill Marshall.

If you don’t know the name: Bill co-founded the Toronto International Film Festival in 1976. Conjured it, really, out of a terrace in Cannes and an argument that Toronto deserved to matter. He was a raconteur, a schemer, a genuine force of Canadian cultural will.

He was also one of the most devoted students of Marshall McLuhan who ever walked the earth.

I was twenty-something. I took notes. I answered phones. I sat in rooms where ideas moved fast and the reference points were dense. Bill talked about McLuhan the way some people talk about oxygen — as something you needed so constantly it barely seemed worth naming.

*The medium is the message.*

I nodded. I filed it. I moved on.

I spent the next three decades building software.

---

Enterprise software has a fundamental problem that nobody talks about honestly.

It pretends reality can be finitely encoded.

It cannot.

Every client has different terminology, different workflows, different assumptions, different exceptions. One company’s “completed” means the learner watched the video. Another means they passed the assessment. Another means a manager signed off after a field observation they conducted in a hard hat in a refinery in northern Alberta at minus thirty degrees.

And what do we do?

We write another rule.

`if OntarioHealthcare and LegacyBranch and CompletionDate < ExpiryDate`

Another exception. Another configuration toggle. Another support ticket slowly building an infinite ontology, one semantic argument at a time.

I spent years doing this. Getting better at it. Thinking I understood the problem.

I didn’t understand the problem.

---

The understanding came sideways.

I was thinking about report testing — the unglamorous engineering problem of how you verify that a highly configurable reporting system actually works. The traditional instinct is exhaustive: test every report, every permutation, every field, every edge case, every client scenario.

This approaches infinity immediately.

And something shifted.

*We are testing the wrong thing.*

We should not be testing whether the software perfectly understood every client’s business meaning. That problem is infinite. That problem will never be solved. That problem is a category error dressed up as a technical requirement.

We should be testing whether the system preserved supported data structures correctly through every stage of computation.

That problem is finite.

And once I saw it I couldn’t unsee it.

A bitmap preserves every point. A vector preserves the shape.

Enterprise software has been trying to be a bitmap of human meaning since the beginning. It will never be good enough. There are too many pixels in the universe.

The real work is the geometry.

---

I started thinking about why large language models work at all.

Not in the press release sense. In the actual sense.

They’re trained on tokens. Fragments. Individually, they mean almost nothing. Statistically, they’re noise with punctuation.

But something emerges. Something that can reason, connect, generate, surprise.

It doesn’t emerge from the tokens.

It emerges from the relationships between them. The topology. The latent geometry of meaning compressed across billions of examples into something that behaves, improbably, like understanding.

Intelligence is in the shape.

Not the dots.

And I sat with that for a while. And then it hit me all at once, decades late, slightly embarrassing in its tardiness:

*The medium is the message.*

Bill said this to me. Bill said this constantly. I nodded and filed it and moved on, because I was twenty-something and I thought I knew what it meant and I did not know what it meant at all.

It doesn’t mean the delivery mechanism is more important than the content.

It means the structure is where the meaning lives.

It means a television doesn’t just carry messages — it *is* a message, one that rewires how you process everything that comes through it. It means the shape of the container changes the shape of the thought.

Tokens are the content. Vectors are the medium. And the medium — the geometry of relationships baked into those billions of parameters — is what actually thinks.

McLuhan wasn’t being clever. He was describing something foundational about how meaning moves through systems. I just needed to build a few of those systems from scratch before I could hear it.

---

I think about Bill sometimes when I’m deep in an architectural decision.

He died in 2017. New Year’s Day. Which felt, to people who knew him, like an act of theatrical timing he probably would have appreciated.

He built something that remade Canadian culture, and he did it by understanding that the festival itself — the shape of the gathering, the public nature of it, the fact that it was for everyone and not just insiders — was not just the vehicle for the films.

It was the message.

Toronto became a film city not because of which movies played. Because of what kind of space Bill and Dusty and Henk built around the watching.

The medium.

I was twenty-something in those rooms. I didn’t know I was learning the most important thing I would ever learn about software architecture.

---

The hardest problems in software are no longer computational.

They are ontological.

AI is making raw computation abundant. Which means the bottleneck shifts upward, into meaning. Into the question of what we are actually trying to preserve, and what shape that thing has, and whether we even know the difference between the pixels and the geometry.

The programmers who thrive in the next era won’t just write syntax.

They’ll be ontology designers. Shape-makers. People who understand that you are not encoding reality — you are constructing a stable approximation of it, and the approximation’s structure is where the truth lives or dies.

That turns out to be philosophy.

It also turns out to be what a loud, brilliant, McLuhan-quoting Scotsman was trying to tell a twenty-year-old who was too busy taking notes to actually listen.

I’m listening now, Bill.

Thirty years late.

But I finally got it.

Even if I still don’t understand the point of Robbie Burns Day.

But I’m young.

Not startup stress. Not “founder burnout.” Those are almost luxury terms now... terms invented by people whose infrastructure mostly worked.

I mean the chronic cognitive tax of carrying unfinished systems in your head for decades.

You know exactly what should exist.
You know how the machine should behave.
You know where the bottlenecks are.
You know which workflows are fragile, which data structures are compromised, which interfaces lie to the user, which operational processes depend on tribal knowledge and prayer.

And every year you postpone fixing them because the company still has to ship. Payroll still has to clear. Clients still need support. Sales still need demos.

So you accumulate technical debt... yes.

But more dangerously, you accumulate organizational debt. Architectural debt. Decision debt. Deferred cognition.

For most of my adult life, I’ve had entire systems living only in my head because translating them into reality required teams, budgets, coordination layers, documentation cycles, and months of interruption.

Then Claude arrived.

And over roughly 1.8 billion tokens... somewhere between fifteen and twenty thousand dollars in API and tooling costs... I rebuilt the operational substrate of my company.

Not the marketing website.
Not a chatbot.
Not “AI features.”

The substrate.

The hidden machinery.

The stuff no investor sees in a pitch deck because it isn’t sexy enough to demo but determines whether a company compounds or slowly suffocates under its own accumulated entropy.

I need people to understand something important here:

This wasn’t vibe coding.

I wasn’t asking Claude to “make me an app.”

I was using it more like an exocortex... an infinitely patient senior systems architect that never got tired of re-evaluating abstractions with me at 2:13 AM.

Over months, we rebuilt massive sections of the operational substrate underneath the company itself... not just technically, but conceptually.

Not because AI magically “did the work.”

Because the cost of iteration collapsed.

That’s the part most people still don’t understand.

The revolution isn’t that AI gives you answers.

The revolution is that thought itself becomes dramatically cheaper to externalize, test, reorganize, challenge, and refine.

For twenty-five years, every major architectural rethink carried enormous activation energy.

You needed meetings, diagrams, developers, specifications, scheduling, emotional stamina, budget justification, and organizational synchronization.

Now you can stay inside the cognitive loop continuously.

You can model systems at the speed of thought.

You can hold fifty interacting abstractions in motion and pressure test them dynamically.

You can ask:

“What breaks if we invert the permissions model?”

And immediately explore downstream implications across infrastructure, UX, governance, scalability, and pedagogy.

That changes what a founder even is.

Here’s the strange part:

The actual money almost becomes irrelevant.

If you had told twenty-five-year-old me:

“For twenty thousand dollars you can acquire a tireless architectural collaborator capable of helping you rebuild your company’s nervous system over the next year...”

I would have considered that supernatural.

Because historically, building internal infrastructure at this level required a large engineering organization, elite consultants, years of roadmap prioritization, or extraordinary luck in hiring.

Now a single determined founder with clarity and stamina can execute organizational rewrites previously reserved for massively capitalized companies.

Not because AI replaces expertise.

Because it amplifies coherent intent.

That distinction matters enormously.

Still, I think people misunderstand what the money actually bought.

The twenty-thousand-dollar spend didn’t rebuild my company. It bought me visibility. It bought me plumbing that, for most of my career, would have been considered an absurd luxury unless you were a very large organization.

What AI gave me was the ability to finally expose and connect systems that had been living partially in my head for decades. Not features. Not products. Structure. Relationships. Decision pathways. The hidden connective tissue underneath the business itself.

For years I understood these systems intuitively, but understanding something and being able to continuously inspect it are completely different things. Now I can trace assumptions more clearly, understand downstream consequences more quickly, and reorganize ideas without the enormous operational friction that used to accompany every major rethink.

And honestly, by the time this is fully mature, it probably won’t cost twenty thousand dollars. It’ll likely cost a few hundred thousand in compute, infrastructure, and iteration.

What fascinates me is that none of that spend may add a single visible feature for a customer.

It changes coherence.

It changes continuity.

It changes how the organization understands itself.

Because the real asset was never the code.

The ontology is the asset.

But even ontology by itself is incomplete. What actually gives a company life is the decision lineage that produced that ontology in the first place. The accumulated chain of judgments, tradeoffs, observations, constraints, mistakes, and realizations that caused the structure to evolve the way it did.

I think people dramatically underestimate how fragile that layer really is.

The best metaphor I’ve found is a Christmas tree.

As long as the tree remains rooted in the ground that produced it, it continues to grow. The second you sever it from that soil... enjoy the holidays.

It may still look alive for a while.

But growth has already stopped.

That’s the part of the AI transition that genuinely troubles me.

Everyone is focused on automation, productivity, and labor reduction, which makes sense. But I think we are also heading toward a massive collapse of institutional memory. When teams of twenty become teams of one, the organization may preserve the outputs while quietly losing the reasoning that produced them.

Someone new eventually inherits the architecture, the workflows, the infrastructure, the codebase... but the “why” behind it all has evaporated.

And once that happens, organizations slowly become custodians of systems they no longer fundamentally understand.

That worries me far more than automation itself.

Because despite how extraordinary these systems are becoming, I still don’t believe AI replaces original cognition. It can appear creative because it has access to a nearly incomprehensible landscape of prior human exploration. Most of the problems we encounter in a business day have already been solved somewhere by someone, and in those situations using AI is simply rational.

Why repeatedly spend human cognition rediscovering known terrain?

But eventually you arrive at problems where no map exists yet. And that’s where human judgment still matters completely.

So I don’t think programmers disappear. I think their role changes. Increasingly, I need people who can direct systems, shape ontology, sustain coherence, and exercise judgment under uncertainty.

For decades the industry rewarded people for being extraordinary executors.

Now I think we are entering an era where the truly rare skill is knowing where the horses should go.

I don’t think the public fully grasps what happens when experienced operators gain the ability to continuously externalize institutional knowledge.

The first wave of AI discourse focused on replacing labor.

I think the more profound shift is replacing friction.

Friction between thought and execution. Architecture and implementation. Vision and specification. Complexity and iteration.

For decades, companies calcified because organizational cognition was expensive.

Every redesign threatened velocity.

Every rewrite threatened stability.

Every deep rethink required political capital.

AI changes the economics of reconsideration itself.

That may end up being bigger than automation.

And yes... there’s a dangerous side to this.

Because once you experience operating this way, traditional organizational latency starts feeling intolerable.

You begin noticing how much of corporate structure exists simply because thinking used to be expensive.

Middle layers emerge to transport information between silos because humans couldn’t maintain high-bandwidth shared cognition continuously.

But what happens when they can?

Or when one founder plus AI can outmaneuver entire departments in strategic adaptability?

We’re entering very weird territory.

What fascinates me most is that the output often doesn’t look dramatic from the outside.

There’s no flashy consumer launch.

No viral demo.

No dancing humanoid robot.

Just an organization becoming internally coherent at a depth that previously would have taken years.

Cleaner abstractions. Better schemas. More legible systems. Reduced ambiguity. Faster strategic movement. Improved operational alignment.

Infrastructure.

The invisible stuff that determines whether your future compounds or collapses.

I don’t think history will remember this era primarily as “the chatbot revolution.”

I think it may eventually be understood as the moment cognition itself became scalable infrastructure.

And for founders who have spent decades carrying unfinished systems silently in their heads...

That changes everything.

But maybe they should be thinking about Janus instead.

Janus was the Roman god with two faces. One looking forward and one looking backward. And increasingly, that feels like the real state of the AI industry. Not divided between companies so much as divided within them.

Because there are now two fundamentally different beliefs about what AI actually is, where it is heading, and what kind of business will survive around it.

One side believes we are on the road to general intelligence. Not metaphorically, but literally. They see current systems as early, primitive, incomplete versions of something far larger. From that perspective, intelligence itself is mostly a scaling problem. More GPUs, more energy, more data centers, more parameters, more synthetic cognition layered on top of synthetic cognition until something qualitatively different emerges.

This side of the industry is spending like a civilization-level arms race has already begun. And if they are right, the spending is rational. If the first company to achieve true AGI gains control over the most important economic engine in human history, then a trillion dollars in infrastructure is not excessive; it is conservative.

That is the face building nuclear power agreements and hydro contracts. The face buying entire global GPU supply chains. The face preparing for a future where intelligence itself becomes centralized infrastructure.

But there is another face staring back at them from the same mirror.

And what makes this split even stranger is that the foundations of the modern AI industry did not emerge from research into consciousness, reasoning, or grounded understanding of reality. They emerged from translation.

Google’s 2017 paper Attention Is All You Need became the true founding document of the modern AI era.

https://arxiv.org/abs/1706.03762

The Transformer architecture it introduced turned out to be astonishingly effective at modeling relationships inside language and symbolic representations. The original breakthrough was not “we discovered digital cognition.” It was something much narrower and much more practical: attention mechanisms were extraordinarily good at understanding relationships between tokens inside sequences.

From that breakthrough, however, the industry began making a series of increasingly ambitious extrapolations. Translation became language modeling. Language modeling became reasoning. Reasoning became intelligence. And intelligence became AGI.

The entire modern AI industry now sits on top of that ladder of assumptions.

Then in 2020, OpenAI published Scaling Laws for Neural Language Models, which effectively poured gasoline onto the idea.

https://arxiv.org/abs/2001.08361

That paper argued something deceptively simple but economically explosive: larger models trained on more data with more compute kept improving in surprisingly predictable ways.

And that was the moment the industry truly lost its mind.

Because if capability kept scaling with compute, then the race became obvious. More GPUs, more data, more power, more infrastructure, more scale. And to be fair, the evidence initially looked overwhelming.

The models worked.

Absurdly well.

Good enough that an entire industry briefly convinced itself that maybe intelligence really was just next-token prediction at sufficient scale.

And perhaps that is unfair.

But only slightly.

Because underneath all the mythology, these systems are still fundamentally predictive engines. Extremely sophisticated predictive engines, yes... but predictive engines nonetheless.

The uncomfortable possibility haunting the industry is that scaling prediction may not be the same thing as scaling understanding.

GPT-4 felt like Lucy finally holding the football steady.

GPT-5 was the moment a lot of people in the industry started wondering if they were Charlie Brown again.

Because now the industry is arriving at the dangerous phase of the experiment: finding out whether the world’s most expensive autocomplete engine can actually cross the next threshold. Increasingly, many of the people closest to these systems appear to suspect the answer is no. Or at least, not through brute force scaling alone.

What these systems unquestionably mastered was language. But language is not reality.

Language is humanity’s symbolic compression layer for reality. A language model lives inside humanity’s descriptions of the world, not necessarily inside the world itself. That distinction becomes brutally obvious the moment you move from text into video.

Text prediction allows plausibility. World prediction requires inevitability.

A language model can often generate something believable about what could happen next. But asking a video model what will happen three frames from now turns out to be a radically different category of problem. Not what might happen, but what must happen.

A dropped glass must continue falling. Water must preserve continuity. Objects hidden behind other objects must still exist when occluded. Bodies must obey physics. The world must remain coherent even when partially unseen.

Suddenly the barrier is no longer syntax or semantic association. It is grounded reality itself. And no amount of additional compute automatically guarantees you can brute-force your way through that wall.

I increasingly suspect this possibility haunts the second faction inside these companies. Because if they are right, then the current models are not primitive proto-gods waiting to wake up. They are already the product.

And if that is true, then the race changes completely.

This second group sees something the dreamers often do not. Current models are already powerful enough to reshape enormous parts of civilization. Software development, education, customer support, management structures, search, scientific research, material science, media production... all of it is already beginning to bend around systems that exist today.

But to them, this is no longer just a technological observation. It is a financial emergency.

Because sitting behind the industry is an almost absurd engine of capital expenditure. A monster consuming power plants, GPUs, data centers, fiber, water, and entire national-scale energy agreements at a rate that starts to feel less like software and more like heavy industry.

The people trying to commercialize current AI increasingly look like they are frantically paving the road ahead of that machine before it catches them.

Every SaaS feature suddenly becomes vulnerable. Calendar companies, CRM companies, note-taking apps, customer support systems, search products, workflow tools, presentation software, research assistants... everything starts looking like one more feature that can be absorbed into a larger AI subscription bundle.

You can almost feel the urgency behind it. One more product integrated. One more workflow captured. One more recurring subscription pulled into the ecosystem before the infrastructure bill arrives.

Google understands this instinctively because Google has always understood bundling. Gmail was never just Gmail. Docs was never just Docs. Search was never just Search. The real business was always the gravitational field created by tying everything together.

Gemini increasingly feels like the continuation of that strategy. Not merely an AI model, but an attempt to turn all of Google Workspace into a single cognitive subscription layer where every adjacent software category eventually collapses inward.

And if you use Gemini long enough, you start noticing another tell. It leans heavily on Google’s existing search infrastructure: existing indexes, cached knowledge, retrieval systems, summaries, search grounding. You almost have to strong-arm it into actually going out and spending meaningful inference effort exploring something fresh.

That is not an accident. That is economics leaking through the interface.

A company truly convinced that unlimited inference was the future would behave differently. It would burn compute aggressively and freely because compute would be the path to the prize. Instead, what we increasingly see across the industry are systems trying to conserve reasoning wherever possible. Retrieval first. Cached answers first. Smaller specialist systems first. Inference only when necessary.

These are not the behaviors of companies escaping economic gravity. They are the behaviors of companies deeply aware of it.

Microsoft’s behavior, meanwhile, feels oddly conflicted.

For years, the company executed one of the most successful business transformations in modern corporate history. It turned itself from a boxed software company into a cloud and subscription empire built around Azure and Office 365. It understood recurring revenue, enterprise dependency, workflow gravity, and organizational lock-in better than almost anyone.

Which is what makes the current moment feel so strange.

Because Microsoft already owned the daily operational fabric of enterprise work. Email, calendars, documents, spreadsheets, meetings, identity management... the company was already sitting exactly where delegated cognition was naturally going to emerge.

And yet much of its AI posture still feels like it is orbiting OpenAI’s narrative rather than extending Microsoft’s own.

Copilot often feels bolted onto products rather than metabolized into them. A feature suite attached to Office instead of an operational layer dissolving invisibly into workflow itself.

Google, by contrast, increasingly behaves like a company trying to make AI disappear into the substrate of its ecosystem.

That difference matters.

One approach treats AI as a destination product.

The other treats it as atmospheric infrastructure.

Anthropic appears to have approached the problem differently, leaning hard into programming because code has unusually clean economics. It has measurable outputs, enormous labor replacement value, direct enterprise demand, and vast open-license ecosystems to train against.

That may also explain why Anthropic seemed unusually willing to make peace with creators and publishers early. I increasingly suspect they understood something important: general content may matter politically, but programming ecosystems matter economically.

And more than that, I suspect they understood something strategically brutal.

If OpenAI, Google, xAI, and everyone else end up trapped in endless licensing wars over the right to answer humanity’s infinite stream of general-interest questions, all the better for Anthropic.

Because every billion dollars spent negotiating with publishers, creators, news organizations, archives, and eventually perhaps even something like Wikipedia is a billion dollars diverted away from infrastructure, tooling, programming ecosystems, and enterprise integration.

Meanwhile, Anthropic increasingly behaves like a company that believes the real long-term value lives somewhere much narrower and much more economically defensible: code.

Code has validation systems. Code has reinforcement loops. Code produces leverage. Code generates more systems.

A model optimized for programming is not merely answering questions. It is participating in the recursive construction of future infrastructure.

And if that is the future Anthropic sees coming, then encouraging the rest of the industry to become entangled in increasingly expensive cultural licensing battles starts looking less like altruism and more like strategic positioning.

Let everyone else spend fortunes building infinitely patient machines for answering the world’s stupidest questions.

Meanwhile, the company focused on helping developers build the future may end up owning the economically productive layer underneath all of it.

And underneath all of this sits the industry’s deepest fear: the models themselves may become commodities.

In fact, the existence of public APIs already suggests the companies know this risk is real. If raw intelligence could remain permanently proprietary, they would never expose it cleanly to the market in the first place.

But they know what happens to infrastructure over time. Storage commoditized. Bandwidth commoditized. Compute commoditized. Databases commoditized. Cloud infrastructure commoditized.

Increasingly, intelligence itself may commoditize too.

There will always be someone willing to sell inference cheaper. There will always be open models. There will always be regional providers. There will always be distillation. There will always be optimization pressure.

Which means the durable value likely moves upward into the surrounding systems: memory, workflow, identity, coordination, trust, orchestration, integration.

Not the model itself... the ecosystem wrapped around it.

That is why the industry increasingly feels split between two futures.

One future treats AI as a path toward centralized superintelligence.

The other treats AI as a new infrastructure layer around which useful systems will be built.

The first group wants bigger minds.

The second wants better orchestration.

And increasingly, I suspect the second group may win.

Not because AGI is impossible. Not because the models stop improving. But because markets eventually force discipline.

Public companies obey margins. Infrastructure obeys economics. Customers pay for usefulness, not metaphysics.

A model that is 20% smarter but 30 times more expensive may simply lose commercially to a system that is cheaper, integrated, reliable, and already good enough for most economic purposes.

And “good enough” may already be enough to permanently alter civilization.

That is the strange thing about this moment.

The AI revolution may already have happened.

The industry just cannot yet agree on whether it is building the convergence... or a business.

We’ve confused disrespect with intelligence.

There was a time when pushing back on authority meant something. It implied you had done the work…that you understood the structure you were rejecting. Now it’s theatre. The phrase “appeal to authority” has been twisted into a blunt instrument…a way to dismiss expertise without ever engaging it. It was meant as a warning against blind trust…now it’s used to justify blind ignorance.

Let’s be clear.

Authority is not truth.

Expertise is not infallible.

But ignorance is not a counterweight.

And yet here we are…in a world of shit talkers who believe skepticism alone is a credential.

I don’t mind tearing down experts who deserve it. There are plenty. But if you’re going to dismantle something…you should at least understand what you’re dismantling.

Walk before you run.

Run before you punch.

Most people skip straight to swinging.

And this isn’t theoretical…we’ve seen what real dissent looks like.

The Impressionists didn’t reject realism because they couldn’t do it…they mastered it first.

The doctor who proved ulcers were bacterial didn’t wander in with a hot take…he did the work. He was mocked one year…keynote the next.

Martin Luther didn’t casually question the Church…he could quote chapter and verse before he ever nailed anything to a door.

Even Luke had to go to Dagobah.

That’s the pattern.

The people who challenge systems at the highest level are the ones who climbed them first.

There is no zealot like a convert.

Only those who can climb to the highest levels of a domain should dare…and must dare…to tear it down if it’s false.

I’m not arguing against rebellion.

I’m arguing against dabbling.

What we’ve lost is respect for process. The idea that knowing something requires time…constraint…exposure.

Instead, we have vibe voting. Opinion without source. Certainty without consequence.

And at the other end…we have a different failure.

Smart people who should know better.

People who succeed in one domain and begin to believe that success transfers. That the skill wasn’t specific…it was universal. So they wander…into systems they don’t understand…and mistake complexity for incompetence.

The harder the problem…the more convinced they become it must be simple.

That’s how you get dilettantes at the highest levels.

We used to have a word for this.

Dilettante.

We’ve lost it.

Now the uninformed and the overconfident meet in the same place…and actual expertise gets drowned out.

I’m tired of it.

I’m tired of people having opinions with no source to back them up. I’m tired of people dismissing decades of work with a paragraph and a vibe.

This isn’t about defending institutions.

This is about defending reality.

And I’m not immune to this.

I build systems in my head. Fast. Clean. Elegant.

They feel right.

They click.

And that feeling…that coherence…is dangerous.

Coherence isn’t reality.

Coherence isn’t reality.

Coherence isn’t reality.

A system that holds logical water is not the same as one that has survived contact with the real world.

Mine often haven’t.

So yes…I’m suspicious of this tendency.

I’m not just the critic.

I’m not just the guy pointing it out.

I’m not just the president…

I’m also a member… 😉

Which is exactly why I don’t trust it.

Because the real danger isn’t that people ignore expertise.

It’s that they replace it with themselves…too quickly.

And then…we arrive here.

At the apogee of bullshit.

Entire journals polluted. Papers forged. Data massaged. Citation loops reinforcing each other into legitimacy.

The cracks were real.

The skepticism was earned.

But here’s the mistake.

We didn’t respond by tightening the standard…we responded by abandoning it.

We replaced “prove it” with “I don’t believe you.”

And that’s where this moment changes.

Because this isn’t just the peak.

It’s the exposure.

For the first time, ignorance isn’t just loud…it’s visible. Traceable. Testable.

We didn’t just build machines that can generate bullshit…

We built systems that can finally measure it.

That’s the shift.

AI isn’t the problem here.

It’s the audit.

At scale.

It can trace citation networks.

Surface statistical anomalies.

Expose patterns no human reviewer could catch.

Not perfectly. Not cleanly.

But relentlessly.

And that changes the equation.

Because now…authority doesn’t get to hide behind itself anymore.

It gets tested.

Continuously.

This isn’t a permanent rebellion against orthodoxy.

It’s the chance to purify it.

To strip out the frauds.

To expose the dilettantes.

To reinforce what actually holds.

The goal was never to eliminate expertise.

The goal was to make it worthy of trust again.

And maybe…for the first time…we have the tools to do it.

But only if we raise the standard instead of abandoning it.

Only if we stop pretending that every opinion is equal.

Only if we stop confusing skepticism with understanding.

We don’t need more voices.

We need better filters.

We might be living at the apogee of ignorance’s power.

Not because ignorance became stronger…

…but because it became easier to perform.

The cost of pretending collapsed.

The rewards expanded…all the way to the highest offices in the land.

That’s what changed.

But what comes next will be shaped by something else.

Not confidence.

Not volume.

Not coherence.

Reality.

What survives contact with it.

We’ve confused disrespect with intelligence.

There was a time when pushing back on authority meant something. It implied you had done the work…that you understood the structure you were rejecting. Now it’s theatre. The phrase “appeal to authority” has been twisted into a blunt instrument…a way to dismiss expertise without ever engaging it. It was meant as a warning against blind trust…now it’s used to justify blind ignorance.

Let’s be clear.

Authority is not truth.

Expertise is not infallible.

But ignorance is not a counterweight.

And yet here we are…in a world of shit talkers who believe skepticism alone is a credential.

I don’t mind tearing down experts who deserve it. There are plenty. But if you’re going to dismantle something…you should at least understand what you’re dismantling.

Walk before you run.

Run before you punch.

Most people skip straight to swinging.

And this isn’t theoretical…we’ve seen what real dissent looks like.

The Impressionists didn’t reject realism because they couldn’t do it…they mastered it first.

The doctor who proved ulcers were bacterial didn’t wander in with a hot take…he did the work. He was mocked one year…keynote the next.

Martin Luther didn’t casually question the Church…he could quote chapter and verse before he ever nailed anything to a door.

Even Luke had to go to Dagobah.

That’s the pattern.

The people who challenge systems at the highest level are the ones who climbed them first.

There is no zealot like a convert.

Only those who can climb to the highest levels of a domain should dare…and must dare…to tear it down if it’s false.

I’m not arguing against rebellion.

I’m arguing against dabbling.

What we’ve lost is respect for process. The idea that knowing something requires time…constraint…exposure.

Instead, we have vibe voting. Opinion without source. Certainty without consequence.

And at the other end…we have a different failure.

Smart people who should know better.

People who succeed in one domain and begin to believe that success transfers. That the skill wasn’t specific…it was universal. So they wander…into systems they don’t understand…and mistake complexity for incompetence.

The harder the problem…the more convinced they become it must be simple.

That’s how you get dilettantes at the highest levels.

We used to have a word for this.

Dilettante.

We’ve lost it.

Now the uninformed and the overconfident meet in the same place…and actual expertise gets drowned out.

I’m tired of it.

I’m tired of people having opinions with no source to back them up. I’m tired of people dismissing decades of work with a paragraph and a vibe.

This isn’t about defending institutions.

This is about defending reality.

And I’m not immune to this.

I build systems in my head. Fast. Clean. Elegant.

They feel right.

They click.

And that feeling…that coherence…is dangerous.

Coherence isn’t reality.

Coherence isn’t reality.

Coherence isn’t reality.

A system that holds logical water is not the same as one that has survived contact with the real world.

Mine often haven’t.

So yes…I’m suspicious of this tendency.

I’m not just the critic.

I’m not just the guy pointing it out.

I’m not just the president…

I’m also a member… 😉

Which is exactly why I don’t trust it.

Because the real danger isn’t that people ignore expertise.

It’s that they replace it with themselves…too quickly.

And then…we arrive here.

At the apogee of bullshit.

Entire journals polluted. Papers forged. Data massaged. Citation loops reinforcing each other into legitimacy.

The cracks were real.

The skepticism was earned.

But here’s the mistake.

We didn’t respond by tightening the standard…we responded by abandoning it.

We replaced “prove it” with “I don’t believe you.”

And that’s where this moment changes.

Because this isn’t just the peak.

It’s the exposure.

For the first time, ignorance isn’t just loud…it’s visible. Traceable. Testable.

We didn’t just build machines that can generate bullshit…

We built systems that can finally measure it.

That’s the shift.

AI isn’t the problem here.

It’s the audit.

At scale.

It can trace citation networks.

Surface statistical anomalies.

Expose patterns no human reviewer could catch.

Not perfectly. Not cleanly.

But relentlessly.

And that changes the equation.

Because now…authority doesn’t get to hide behind itself anymore.

It gets tested.

Continuously.

This isn’t a permanent rebellion against orthodoxy.

It’s the chance to purify it.

To strip out the frauds.

To expose the dilettantes.

To reinforce what actually holds.

The goal was never to eliminate expertise.

The goal was to make it worthy of trust again.

And maybe…for the first time…we have the tools to do it.

But only if we raise the standard instead of abandoning it.

Only if we stop pretending that every opinion is equal.

Only if we stop confusing skepticism with understanding.

We don’t need more voices.

We need better filters.

We might be living at the apogee of ignorance’s power.

Not because ignorance became stronger…

…but because it became easier to perform.

The cost of pretending collapsed.

The rewards expanded…all the way to the highest offices in the land.

That’s what changed.

But what comes next will be shaped by something else.

Not confidence.

Not volume.

Not coherence.

Reality.

What survives contact with it.

Which is exactly the problem.

If you’re a CEO, the present will consume you. Slack, decisions, people, noise… it all demands urgency. And urgency crowds out anything that isn’t immediate.

So a few times a year, I leave.

Not to a resort. Not to a conference.

I fly west… Saskatoon or Winnipeg… and then I take the train back to Toronto.

Two days. One or two nights. Cabin for one.

And I work.

---

The point isn’t the train. The point is what the train takes away.

No reliable internet for large stretches.

No meetings.

No one dropping by.

No ambient obligation to respond.

You can’t react… so you start to think.

There’s also something about movement. You’re going somewhere, even if slowly, and it creates a kind of psychological permission. Decisions feel less stuck. Ideas move. You move.

It’s hard to explain, but it’s consistent.

---

I’m alone most of the time.

That matters.

But not completely alone. Meals break it up. Conversations happen if you want them. At the end of the day, I’ll have a drink in the club car. Just enough social contact to stay human… not enough to derail the focus.

It’s a tuned environment. Not isolation for its own sake… isolation with relief valves.

---

I started doing this after reading about Bill Gates taking “reading weeks.” The idea stuck. You need time apart to deal with things that aren’t urgent.

At the time, I used these trips to clear backlog. Writing, mostly. Things I owed but never got to because the office only ever gives you “now.”

That problem is basically gone. AI took it.

The writing backlog doesn’t build anymore.

But the thinking problem didn’t go anywhere.

If anything, it’s more obvious now. Once the mechanical work disappears, what’s left is the part only you can do. Deciding what matters. Figuring out what to do next.

That’s what the train is for.

---

I think this is a bit of a boss move.

Most CEOs signal importance by being available. Always on. Always responsive.

I do the opposite.

I disappear… on purpose… on a schedule.

And the business survives just fine.

Better, actually.

---

It’s getting harder to book a cabin on VIA Rail now, which is interesting in itself. So I’m experimenting.

Next trip I’m flying to New York and taking the Silver Meteor down to Miami.

It’s not as nice. Food’s worse. The experience isn’t the same.

But I’ll have more connectivity… and for this particular trip, that helps. What I’m working through benefits from some input.

The platform changes. The principle doesn’t.

---

I don’t take the train to travel.

I take it to think.

Everything else is just logistics.

So the instinct is to take a shot at it.

To treat it as a symptom of decline… soft people, low standards, a culture that has traded edge for comfort.

But the longer you look at it, the less stable that critique becomes.

Because pickleball is doing something very specific… and very effective.

It creates structured leisure.
It produces low-stakes competition.
It gives people a place to show up, regularly, with just enough friction to feel engaged and not enough to feel threatened.

It’s not trying to be meaningful in a grand sense. It’s trying to be repeatable.

And that’s the key.

It occupies the same functional role that Huxley described in Brave New World. Not in the literal sense… there’s no centralized authority designing it, no deliberate economic manipulation behind it. But in terms of outcome, it lands in the same place.

It stabilizes people.

It organizes time.
It creates social fabric.
It absorbs energy that might otherwise have nowhere to go.

Once you see it that way, the target shifts.

Because it’s not just pickleball.

It’s pickleball and Xanax.

That pairing sounds like a joke at first. It isn’t. It’s a pattern.

A way of smoothing the system from two directions:

• pharmacologically, by dampening anxiety

• socially, by providing a predictable loop of interaction and mild reward

Together, they create a kind of equilibrium. Not happiness exactly… but stability. Enough to keep everything functioning.

And then there’s the other track.

Tennis and Molly.

Higher intensity. More edge. More signaling. Less about smoothing, more about amplification. Instead of dampening the baseline, it tries to override it… create peaks that justify the valleys.

Two different responses. Same underlying condition.

Because the environment they’re responding to is the same.

The structures that used to define identity… religion, career paths, fixed roles… have weakened. Not disappeared, but softened to the point where they no longer dictate a clear trajectory.

What replaced them wasn’t a new structure. It was choice.

And choice sounds good until you live inside it.

More freedom.
Less direction.
More possibility.
Less clarity.

What that creates isn’t acute suffering. It’s something quieter.

A kind of low-grade instability.

Not enough to break anything. Not enough to force a reckoning. Just enough to leave a persistent question in the background… what is any of this actually for?

Everything else flows from that.

Some people stabilize the system.
Some people override it.

Both approaches are coherent. Both are adaptive. And both end up looking more similar than they first appear.

Which makes the critique harder to sustain.

Because if you try to tear down one, you end up implicating the other… and eventually yourself.

At that point, a different distinction starts to matter.

Not between specific activities or substances… but between categories of engagement.

There are things that stabilize.
There are things that provide escape.
And there are things that build.

Most people move between the first two without thinking too much about it. They’re necessary. They keep life manageable.

The third category is different.

Not because it’s morally better, but because it behaves differently over time.

Anything that actually builds… a company, a body of work, a family system, an institution… doesn’t resolve cleanly. It doesn’t give you a closed loop. It doesn’t end in a way that feels complete.

You don’t finish it.

That’s not a bug. That’s the structure.

If something can be fully completed within a neat boundary, it usually isn’t in that category. It’s a project. A game. A loop that resets.

Building doesn’t reset.

It compounds. It extends. It outlasts the person doing it… or at least outlasts their ability to see the end.

Which makes it fundamentally different from both stabilization and escape.

Closed loops versus open loops.

Closed loops resolve.
They give you feedback.
They let you feel done.

Open loops don’t.
They carry forward.
They require you to operate without closure.

That’s where the third group sits.

Not “builders” in the romantic sense. Not founders, not artists, not some elevated class.

Just people who have made a quiet decision:

they will spend their time on things that do not pay them back cleanly.

Things that:

• don’t end on schedule

• don’t fully reward effort in the short term

• don’t provide a stable identity while you’re inside them

• and often look indistinguishable from failure for long stretches

They trade clarity for direction.
They trade comfort for continuity.
They trade closure for compounding.

And most importantly… they accept that the outcome is asymmetric in time.

The payoff, if it comes at all, shows up later… often beyond their direct experience of it. Sometimes in other people. Sometimes in systems that keep running after they step away.

That’s why it feels like chasing something you won’t finish.

Because you are.

But that’s the only category where time works differently.

Stabilizers compress time into repeatable loops.
Escapes spike it into peaks.

The third group stretches it.

They’re not trying to feel better today.
They’re not trying to win a contained game.

They’re trying to participate in something that continues.

It must be great to be a monk.
Or to paint till your fingers are raw.
To crash up against a Nobel Prize worthy problem.
Or die in battle.

To love something more than yourself or another person.

I think Don Quixote was a genius. He made his own.

I just play tennis.

Like it remembers.
Like it “knows” them.
Like there’s continuity behind the interface.

There isn’t.

A large language model is stateless.

Every call is a fresh invocation.
No memory.
No continuity.
No accumulated understanding unless you explicitly pass it in.

That “conversation” you think you’re having is just context reconstruction… over and over again.

So Let’s Be Honest About What That Means

If every call is independent…

Then every call is a decision.

And one of the most important decisions is:

Which model should handle this specific piece of work?

Treating model selection as a one-time architectural choice is just wrong.

Different calls have:

• different complexity

• different structure

• different tolerance for failure

• different cost sensitivity

So wiring one model into everything…

isn’t a strategy.

But We Haven’t Solved This

Let’s be clear about that.

We don’t think we’ve nailed this.

What we’ve done is introduce a way to think about the problem.

Right now, that looks like question shapes.

Trying to describe a task in a structured way:

• what kind of work is this

• how complex is it

• what kind of output is expected

And using that to make a decision about which model should handle it.

Maybe that’s right.

Maybe it’s incomplete.

Maybe it’s wrong in ways we haven’t seen yet.

That’s exactly why Tally exists.

This Needs To Be Figured Out Together

You can’t solve this in isolation.

Not across real systems.

Not across real workloads.

Because the moment this touches production:

• assumptions break

• edge cases show up

• costs behave differently than expected

The only way to get this right is through shared usage and feedback over time.

What Tally Does

Tally gives you a way to make model selection a per-call decision…

and learn from it.

You:

• describe the task

• ask for a routing recommendation

• run the call the way you normally would

• report back what actually happened

Success. Failure. Cost.

Tally learns from that.

That’s the loop.

No proxy.
No rewrite of your system.
No pretending this is magic.

Just a better way to make a decision… and improve it.

What We Want From You

We’re opening this up because we need signal.

Not interest.
Not compliments.

Signal.

• run real workloads

• describe tasks honestly

• send back real outcomes

• push it until it breaks

Watch:

• where it saves you money

• where it makes bad calls

• how quickly it adapts

Then tell us.

How To Get Involved

Sign up:

https://tallyy.org

Install the SDK.

Run it on something real.

Then join the Discord and tell us where it’s wrong… where it’s useful… and where it needs to go next.

One Last Thing

Yes… it’s tallyy.org.

Not tally.

We know.

If this works, maybe we earn the second “y” back… or maybe we can finally afford to buy the one we actually wanted.

Tally is a business.

But if we’re honest… it’s also a tool we built because we needed it.

Now we find out if anyone else does.

It’s not that the world got darker. I wasn’t raised on a clean world. I saw abuse on Different Strokes. I saw addiction on Family Ties. I watched Miami Vice and understood very clearly that corruption, crime, and compromise were part of the deal. JR Ewing didn’t hide greed. ER didn’t hide failure. The West Wing didn’t pretend decisions were easy or pure.

That wasn’t the lesson.

The lesson was this:

Even in a broken world, how you behave still matters.

That’s what raised me. Not church. Not school. Not even my parents, if I’m being honest.

NBC. PBS. CBS.

That was my moral framework.

There used to be an entire layer of television that existed to reinforce this. Not subtly. Not ironically. Directly.

Very special episodes. After-school specials. Quiet moments in otherwise light shows where everything slowed down and someone had to make a choice that cost them something.

You didn’t always notice it at the time. But it was constant. Repetitive. Reinforcing.

Actions had consequences.

Consequences meant something.

And relationships didn’t end just because someone failed.

That last one matters more than I realized.

You couldn’t unfriend people.

You saw the same people every day. You worked with them. You lived with them. You depended on them. When something broke, the story wasn’t “who leaves,” it was:

“How do we live with each other after this?”

That’s Benton and Carter.

Carter lets him down. Not in a small way. In a way that actually breaks something. He doesn’t become what Benton thought he would be. He fails the expectation.

And yet… they don’t disappear from each other.

The relationship doesn’t reset. It doesn’t dissolve. It recalibrates.

“You disappointed me” and “you still matter to me” exist at the same time.

That used to be normal.

I think about that a lot when I look at the world now.

Because what I see isn’t just bad behavior. It’s something stranger.

It’s a world where the connection between actions and consequences feels… misaligned.

Not absent. Misaligned.

Big things get absorbed. Small things get amplified. Symbolic violations get punished. Structural failures get managed, spun, or ignored.

And over time, something subtle happens.

People stop expecting consequences to land where they should.

Then they stop expecting them at all.

I keep coming back to a simple analogy.

I know what happens to bones in zero gravity.

They don’t break immediately. They don’t fail dramatically. They just… lose density. Slowly. Quietly. Because they’re not being used to carry weight.

Nothing is wrong in the moment. Everything seems fine.

Until you need them.

And then they can’t do the job.

I think we’re doing the same thing with morality.

When nothing really costs anything, when relationships are optional, when you can exit instead of repair, when systems don’t reliably connect behavior to outcome, you don’t get immediate collapse.

You get moral atrophy.

This is where I struggle with the modern instinct to remove friction from everything.

On one level, it makes sense. Remove unnecessary pain. Eliminate inefficiency. Smooth the system.

But I was raised in a world that taught me something different.

Friction is where meaning comes from.

Not pointless friction. Not bureaucracy. Not incompetence.

Meaningful friction.

The kind that forces a choice.

The kind that reveals who you are.

The kind that costs something.

Look at the stories that stuck.

Homer standing at a dead-end job, reframing his entire existence with “Do it for her.” That’s not optimism. That’s sacrifice.

The end of The Poseidon Adventure. Someone doesn’t make it out. Not because they were weak, but because someone had to stay behind.

Even Breaking Bad, as dark as it gets, still honors the rule in the end. Walter White doesn’t redeem himself. He doesn’t fix anything. But he makes one choice that costs him and saves someone else.

It’s too late to undo the damage. But it’s not too late for the choice to matter.

That’s the old system showing through.

And then I look around now.

Leaders act without clear accountability. Institutions enforce the wrong things. Public reactions feel inconsistent, almost performative. Relationships are curated, reversible, optional.

And I feel something I didn’t expect.

Not outrage.

Confusion.

What are the rules now?

Because the ones I was taught were simple:

Actions matter.

Relationships endure.

If you’re in a position of responsibility, you owe people your best effort.

If you fail, you don’t disappear. You repair.

This isn’t nostalgia.

I’m not arguing that the past was better. Those shows were simplified. Cleaner. More legible than real life has ever been.

But they were doing something important.

They were reinforcing a connection between behavior and consequence.

They were teaching that even in a flawed system, you are still accountable for how you act inside it.

What I’m worried about now isn’t that people are worse.

It’s that we’re slowly removing the forces that make people strong.

We’ve built a world that is very good at reducing friction.

I’m not sure we’ve thought enough about what that removes along with it.

Because if nothing resists you, nothing tests you.

And if nothing tests you, what exactly are you building?

I don’t think we need more content.

I think we need to remember something simpler.

That sometimes the right choice costs you.

That people will disappoint each other.

That relationships don’t have to end because of that.

That responsibility still means something.

That even in a messy, unfair, inconsistent world…

you’re still supposed to try.

One with the people we first relied on, then betrayed, and then spent a century managing as though they were someone else’s children.

The other is with a people we fought, defeated militarily, partnered with, argued with, funded, mocked, married, and still don’t quite understand, even now.

Today I want to talk about the second one.

Because I live inside it… and I don’t pass the authenticity test.

In 1759, on a field outside Quebec City, the British defeated the French in a battle that lasted fifteen minutes and changed a continent.

English Canada has spent the 265 years since then waiting for the other shoe to drop.

It never did.

Because here’s what we got wrong from the beginning. You can win a battle. You can win a war. What you cannot do is win a culture. The French didn’t surrender on those plains — the army did. The church stayed. The language stayed. The civil law stayed. The farms stayed. The names on the mailboxes stayed. The people who were there in 1759 are still there, still speaking French, still not particularly interested in delivering to English Canada the spoils it has spent two and a half centuries quietly assuming were coming.

We won. They never lost.

And I think that’s the thing English Canada has never quite metabolized. Not consciously. Not as a stated position. But as a low hum underneath the argument — this impatience, this puzzlement, this vague irritation that the conquered won’t behave like the conquered.

That’s what you’re actually hearing when a CEO’s French becomes the story.

A man stands up after two young pilots die and offers condolences.

Within days, we’re no longer talking about the dead, or the families, or the loss.

We’re talking about his French.

Maybe he’s an asshole. Maybe he couldn’t be bothered. Maybe after 300 hours of lessons he still sounds like a tourist ordering a coffee and decided not to humiliate himself on camera.

All of that is possible.

But the scale of the reaction… the heat behind it… the insistence that this was the real story…

That’s not about him.

That’s about us realizing, somewhere deep down, that the thing we’ve been using to measure belonging is starting to slip, and we don’t quite know what replaces it.

I went to Bishop’s.

I know to go to Fairmount, not just St-Viateur… and I know why that sentence alone will start a fight that doesn’t really end.

I prefer pastrami… but when I’m at Schwartz’s, I get the smoked meat.

I grew up with The Box humming in the background, and I remember exactly how much shit Sass Jordan took for slipping into English every now and then.

I’ve skied Mont Sainte-Anne, Owl’s Head, Sutton, Tremblant… enough times that they live in muscle memory.

I’ve sat in a Metro car so quiet you forget you’re moving.

And I can’t speak French.

Not really.

I can tell you I’m tired. I can tell you to shut up. I can tell you I’m feeling okay.

And then my brain shuts the door, cleanly and completely.

I started in grade three. I did the years. I put in the time that was asked of me.

It’s not a choice.

It’s a limit.

So here’s the problem.

For a long time, language worked.

If you spoke French, it meant something. It meant you invested, that you stayed, that you chose to be here and to participate in something that required effort and continuity.

It was a good proxy for belonging.

But it was still a proxy.

Because I’ve done all of that.

I’ve lived it, absorbed it, repeated it, cared about it.

And I still fail the test.

Now here’s where it gets uncomfortable.

If you think I’m talking about smoked meat and ski hills, then you’re missing the point entirely.

I can talk for an hour about Lévesque and Trudeau and not get bored, and more importantly, not get it wrong.

I know exactly what the Night of the Long Knives did to the psyche of this province, and how often it still gets invoked… sometimes honestly, sometimes as a kind of reflex.

I understand equalization well enough to know why Alberta resents it, why Quebec depends on it, and how both sides flatten that reality when it suits them.

I know what the Quiet Revolution changed, and just as importantly, what it didn’t.

I know what cultural isolation looked like under Duplessis, and why the overcorrection that followed came as hard and fast as it did.

I know why head offices left, why some came back, and why that story is almost never told cleanly, because it doesn’t serve anyone’s narrative particularly well.

This isn’t cultural tourism.

This is the argument.

And I still fail the test.

Now add what’s coming.

We are inches away from a babble fish in the ear… real-time, near-perfect translation where tone, cadence, and intent carry across languages without friction.

A leader stands up, speaks plainly in English, and every word lands in French exactly as intended, in a voice that sounds like it belongs there.

At that point… what exactly are we measuring?

That’s what this moment is really about.

Not a CEO who didn’t speak French, or didn’t speak it well enough, or didn’t try hard enough.

But a culture realizing that its cleanest, simplest signal of belonging is about to be simulated perfectly, at scale.

And that’s unsettling, because it forces a deeper question that we’ve been able to avoid.

So the question isn’t whether French matters.

It’s a question of whether the French matter… to us.

And the answer is yes.

Not politely. Not symbolically. Not in the careful language of policy.

Desperately.

Because whatever we were when this started… two European inheritances replaying old tensions on new ground… that’s not what we are anymore.

Quebec is not France.

And we are not England.

Neither of us are proxies for anything now.

What exists between us is something built here… through proximity, friction, compromise, and a kind of familiarity that only comes from not being able to leave each other alone.

We’ve fought too much, traded too much, lived too close to pretend otherwise.

This is no longer inherited.

It’s earned.

And if you want proof of that… look south.

We are closer to each other than we are to the Americans.

Not always in temperament. Not always in tone.

But in the things that actually matter… how we govern, how we argue, how we carry history forward instead of pretending we can reset it every generation.

We need each other that way.

Because if we drift too far apart, we don’t become more distinct… we become more absorbable.

And there is a gravity to the south that doesn’t negotiate.

But don't fall asleep, And please forgive me if I've fallen into the old Federalist trap. Stay or become an American is a powerful rhetorical device but not how you forge a partnership.

I need Quebec. I need to be a member of a country that has a Quebec. I need it's reverence for culture for family. I need its quirk. I'm from Toronto, it's very clean and all things are precise. It's very boring. I want Expo 67’, I want a ridiculous snowman skating. I want good food and cheese with bugs and I want the wisdom of Mordecai Richler taught in my schools.

So yes… the French matter.

Which is exactly why the way we show that has to evolve.

We can anticipate the response.

Of course we can.

The language is the respect. The language is the proof. The language is the line.

And for most of my life… I’ve been on the wrong side of that line.

I’ve fought it. I’ve worked at it. I’ve run into it over and over again like a wall that doesn’t care how much is on the other side.

And for the last time in this essay… I’ll say this plainly.

It doesn’t matter.

Not in the way it used to.

Because the framework that made that true is about to disappear.

Technology is about to erase the idea that language can function as a reliable gate.

Perfect translation… real-time… in your ear, in your voice, with your cadence.

The signal will still exist.

But it won’t prove what you think it proves.

So I’m not asking you to abandon it.

I’m telling you to get ready for what comes after it.

Because the question isn’t going away.

It’s just getting harder.

If language no longer tells you who belongs…

how will you know who’s on your team?

Not just on wine lists. In vineyards. In distribution. In the way people talk about alcohol itself.

The industry keeps telling itself the problem is demand. Younger people. Health trends. Cannabis. Take your pick.

That’s lazy.

The real problem is structural... restaurants broke the habit loop.

Wine used to be normal.

You sat down... you ordered a bottle... nobody overthought it. It was part of dinner, not a decision event. Not a negotiation. Not a small financial commitment disguised as a casual choice.

Now it is.

A $120 bottle sits in this weird dead zone. Too expensive to order casually... not special enough to feel like a celebration. So people hesitate. Or they default to cocktails. Or they opt out entirely.

That’s not a cultural shift. That’s a pricing and format failure.

Cocktails didn’t win because they’re better.

They won because they’re easier.

You can spend $20... get something designed, consistent, personal... and move on. No regret. No group decision. No risk.

Wine, on the other hand, asks you to commit. Financially and socially. And if you get it wrong... you sit with it for the rest of the meal.

So people stopped playing.

What replaced it is worse.

Restaurants now rely on fewer people buying more expensive bottles. Concentrated spending instead of distributed participation. One table orders a $300 bottle... Five others drink nothing.

From a spreadsheet point of view... it works.

From a cultural point of view... it’s a disaster.

Because frequency is what builds taste. And taste is what builds future demand.

I had a moment last spring in Amsterdam.

I got talked into a pour of Louis XIII. White gloves. Baccarat crystal. The whole thing. It cost a couple hundred bucks.

I didn’t walk out needing to immediately procure a $5,000 bottle.

But I did the experience again. A few steps down the cost ladder. But Little Hennessy here a little Remy…. A habit grew.

And then I spent about three grand on cognac that summer building a decent cognac collection.

That’s how categories grow. Not through one big purchase... but through repeatable, high-quality access.

Wine doesn’t do this.

Or at least... it doesn’t do it well enough.

We have the technology now. Coravin, inert gas systems, proper storage. You can open almost anything and preserve it. You can turn a $600 bottle into ten $60 decisions.

But restaurants are still thinking in bottle terms.

They’re protecting the idea of the big sale... instead of building the habit that leads to ten smaller ones.

Here’s the shift.

Stop thinking about wine as a product you sell per table.

Start thinking about it as an experience you sell per visit.

Better yet... across visits.

Imagine a restaurant that says:

This month... we are pouring four exceptional wines. Large formats. Properly stored. Properly presented. $40 to $70 a glass. They are here all month... until they’re gone.

You come in on a Tuesday... you try one.

You come back next week... you have it again.

Now you recognize it. Now you trust it. Now you start building a relationship with it.

That’s the missing piece.

Restaurants keep optimizing for tonight.

Maximize the check. Push the bottle. Capture the margin.

But the real opportunity is obvious once you see it.

Turn premium consumption into a repeat behavior.

Make luxury accessible without making it cheap.

Create moments people can come back to... not just remember.

Because right now... the industry is doing the opposite.

It’s extracting as much as it can from the people who already know.

And quietly losing everyone else.

Cocktails didn’t steal the market.

Wine gave it away.

The fix isn’t complicated.

Lower the commitment.
Maintain the quality.
Repeat the experience.

Call it what you want. It fix it or all I can say about your future… Poor You.

Time.

Everything else is derivative. Money is stored time. Status is borrowed time. Comfort is time without pain. Efficiency is time reallocated. Even health, at its core, is runway.

Strip it down and that’s the whole war.

I am fighting on two fronts at once. I want to extend my time on earth; and I want the time I have to be vivid, dense, electric. Those two ambitions do not always cooperate. The behaviors that lengthen life can sand down its intensity. The behaviors that intensify life sometimes burn runway. That tension is not a flaw; it is the game.

This is where AI enters the frame, and where most people miss the point.

We are told the future is about efficiency. Automation. Removing friction. Anticipating needs before we know we have them. A life where nothing hurts your ass and nothing is forgotten and nothing is inconvenient.

I have already lived in a place where all my needs were met before I could articulate them. It was called a womb.

I am not interested in going back.

I do not want all my chores taken from me. I do not want every sharp edge rounded off. My brain does not run on a diet of “heavy problems only.” It needs to brush teeth. It needs to choose clothes. It needs to negotiate relationships. It needs to book tickets and remember birthdays and feel the low hum of ordinary responsibility. Those small acts are not inefficiencies; they are metabolic exercise. They maintain authorship.

A life stripped of minutiae is not liberated; it is abstracted.

We keep talking about “agents,” as if the goal were to build something with agency. Why would I want that? We are not extending the franchise. I do not need a new sovereign intelligence roaming around with its own intentions. I need instruments. I need leverage. I need amplification without abdication.

The industrial instinct, of course, is brute force. We throw power at the problem. More compute. More parameters. More context. It is a very human move. It works. It scales. But it tells you something about the selection pressure shaping the system. Capital optimizes for speed and dominance; evolution optimized for survival under constraint.

And that is where the brain analogy both helps and misleads.

The eye is a terrible camera. It is wired backwards; it has a blind spot; its high resolution is confined to a tiny patch. A decent modern sensor humiliates it on paper. But the brain does something extraordinary: it takes sparse input and hallucinates coherence. It predicts. It fills in. It renders selectively.

Now we build cameras that track your gaze and render only what you are looking at in full resolution. We call it innovation. The eye pulled that trick millions of years ago.

The lesson is not that biology is perfect. It is that biology is efficient under constraint. Cheap sensors; aggressive inference.

Still, why pattern our cognitive future on a system optimized for generalization and survival in tribal Africa? We now have cameras that can capture events at attosecond scales. An attosecond is a quintillionth of a second. There are more attoseconds in a single second than there are seconds between the Big Bang and this moment. We can now watch electron movement unfold at that scale, watching matter rearrange itself almost at the speed of physics.

If measurement were the goal, engineering has already outrun biology by absurd margins.

But measurement is not the goal.

Navigation is.

Precision does not eliminate ambiguity. More resolution does not answer the question of what matters. It does not choose the goal. It does not weigh tradeoffs under mortality. That is where inference enters.

Even so, I am not chasing comfort or efficiency. Beyond a chair that does not injure me, I am done with comfort as a primary aim. What I want is range. I want amplitude. I want novelty.

I want newer, better answers every day. I want my light shinier. I want my sugar sweeter. I want my experience to accelerate.

Novelty stretches time. It increases density. A year of repetition collapses into blur; a year of transformation feels long. When people say they want AI to save time, I suspect many of them actually want AI to intensify it.

But novelty alone is not enough. Raw mutation without selection becomes noise. If every stimulus is maximal, none of it lands. The shine normalizes. The sugar dulls the tongue.

Acceleration must be paired with integration.

What I am really after is an increase in the rate at which I become someone new… without dissolving who I already am.

AI can help there, if used properly. It can increase the mutation rate of ideas. It can recombine disciplines at absurd speed. It can collapse search costs so that more of my hours are spent in engagement rather than frictional lookup. It can surface blind spots and second-order consequences. It can widen the field of possible nights.

But it should not anesthetize me.

It should not silently optimize my life into smoothness. A completely optimized life risks becoming thin. Friction creates contrast. Contrast creates memory. Memory creates the sensation of length. A perfectly efficient existence might paradoxically feel shorter.

Kris Kristofferson once sang that there’s still a lot of wine and lonely girls in this best of all possible worlds.

That line is not naive optimism. It is defiant gratitude. The world is flawed, finite, sometimes brutal; and still it offers intoxication, intimacy, texture. It offers heat.

Extending time matters because there are still nights left. Intensifying time matters because those nights can burn bright.

So if I use AI, it is not to construct a cognitive womb. It is to build a vineyard.

I want tools that sharpen me, not tools that replace me. Systems that increase the density of my lived hours. Systems that make tradeoffs explicit so that when I choose vividness over safety, or safety over vividness, I know what I am doing.

There is only one commodity.

Time not in pain.

Time not in unconscious drift.

Time not lost to dull repetition.

Time spent awake.

If the future is worth building, it is because it increases the amount of conscious, self-authored, vivid time available to a human being… and because it extends the runway just enough that the compounding matters.

Anything else is just furniture.

And I did not come here for furniture.

My aim is an ever improving Doug. If I play my cards right… a Douglas someday.

Someday. 😉

That may seem strange given how completely it has occupied my mind… and my hands… for years now. But I wanted to think before I spoke. I wanted to watch what it was doing to me… to my work… to my sense of the future.

Last week I published the opening salvo. An essay about what learning AI is actually doing to my brain. Not the hype. Not the market. The personal rewiring.

After this begins the trilogy.

A trilogy in four parts 😉

The next piece, publishing right after this, is a retuning. A reframing of how we should be thinking about AI in the first place. We have spent decades trying to replicate the human brain… as if the human brain were some kind of gold standard intelligence.

It isn’t.

The brain is a biological control system. It evolved to run a fragile meat body using bargain-basement sensors. Narrow bandwidth vision. Crude audio. Chemical emotion. Constant survival pressure. It is extraordinary at keeping us alive long enough to reproduce.

That does not make it the ideal architecture for intelligence in a world with absurd instrumentation, planetary networks, and effectively infinite memory.

We are building minds for a universe our biology was never designed to inhabit.

That shift changes what intelligence is. It changes what work is. It changes what ambition is. It changes what it means to build something in public.

The third essay, coming this weekend, is the deep dive. My whole career under the pressure of AI… what it is breaking… what it is amplifying… where I see hope… and where I see a parade of fools walking confidently in the wrong direction.

AI has been with me a long time. I have scars. I have convictions. I have doubts.

Read all of it.

Then tell me where I’m wrong.

Yes… I said four parts.

There will be.

But that would be telling.