Essay specimens — two live runsexperiment

The August numbers have a quality I recognize. San Francisco's median sales price is up twenty-five percent year-over-year; San Mateo County is up nearly ten; the other eight Bay Area counties are essentially flat. Mortgage rates remain elevated, which usually cools demand. Yet in two counties, prices are rising sharply. The ordinary logic of the housing market has been suspended by a very concentrated form of wealth creation.

I have lived in San Francisco for more than forty years. I have seen booms that lifted the whole region and busts that flattened it. This is different. The wealth being generated by the AI boom is not washing across nine counties; it is pooling in the places where the companies are headquartered and where their founders and senior engineers choose to live. San Francisco and the peninsula have become the payroll of one industry. The other eight counties, with their own jobs and their own housing markets, are watching from the outside. That is not a regional market. That is a company town.

It is an old pattern, and I know it from childhood. A company town is not just a place where one employer dominates; it is a place where the cost of shelter is set by the payroll of that employer. When I was growing up in Cardiff, the docks and the steelworks set the ceiling on what a family could afford. Here, the ceiling has been removed. There is no local industry setting a limit—there is only a small group of people paid in equity whose cash offers do not depend on mortgage rates at all. Twenty-five percent in a year is not a bidding war among nurses and schoolteachers. It is a market where the middle class has already lost.

What concerns me is not the number itself. Prices rise and fall; I have never been much interested in forecasting them. What concerns me is what a bifurcated housing market does to the public life of a city. A San Francisco where only the very well paid can buy a house is not the city I moved to. It becomes a city of owners who are mostly employed by one sector and a much larger population that rents, commutes, or leaves. That changes the schools, the small shops, the parks, the volunteer fire department, the PTA. It changes who votes in local elections and who has the time to sit on a neighborhood board. The middle class is not just an economic category; it is the layer of people with both a stake and a spare evening. When it thins, the city's public realm thins with it.

I have spent the last several years trying to improve the city's governance—not because I enjoy politics, but because I believe a badly governed city drives out exactly the people it needs. But governance reform cannot conjure affordability when one industry's payroll is bidding up every house in two counties. The Board of Supervisors could be perfect, and the arithmetic would still be brutal. You cannot run a city on the proceeds of one technology cycle and expect it to remain a city.

The other eight counties being flat is not, to my eye, a sign of health. It is a sign that the region's economic energy is not broad. It is concentrated in the places where a particular kind of company happens to sit. The Bay Area has always had enclaves of wealth—Atherton, Hillsborough, parts of San Francisco. But this is something else: a map divided along the lines of one industry's geography. The question is whether we want the region to become that map permanently.

I am not an analyst of artificial intelligence, and I will not pretend to judge its long-term value. What I can see is its short-term effect on the place I live. It is making San Francisco more like a company town than it has been since the Gold Rush. The Gold Rush eventually produced a city with all the mess and variety of one. A company town, once it calcifies, produces something much narrower.

The numbers are an invitation to decide what kind of place we are willing to become. Not a forecast. An invitation.

OpenAI just shipped a teen mode, complete with parent-set Study Hours. I understand the appeal. The anxiety is real: your kid has a homework question, and instead of wrestling with it, she asks ChatGPT. So the product now blocks or redirects during certain hours. The immediate anxiety goes down. But I keep thinking about what gets outsourced.

Here's the part most people skip: distraction is not a technology problem. It's an escape from discomfort. Boredom, uncertainty, the stuck feeling of a hard problem—that's the internal trigger. If ChatGPT is the escape hatch, closing it during Study Hours doesn't remove the discomfort. It just removes that particular hatch. The kid will find another one, or sit there seething, or count the minutes until the wall comes down. We've built a solution to the symptom, not the skill.

I'm not against precommitment. I bought a five-dollar outlet timer and plugged my router into it years ago. It kills the internet on a schedule, and it works because I chose it, after my flip-phone experiment failed. I learned I couldn't out-willpower my own discomfort. So I made an effort pact, in advance, when I was calm. But the key word is I. A precommitment you make for yourself is an act of agency. A precommitment someone else makes for you is a rule. Rules breed compliance, not competence. Worse, they teach a kid that self-control is something that comes from the outside.

The parent-set Study Hours feature does exactly that. It takes the moment of choice—the moment a teen feels the itch to open the app—and removes it. The app won't open, so the teen never has to practice noticing the itch, naming the discomfort, and deciding on traction. That's the whole skill. If you never practice it, you don't build it. Then the kid goes to college, or gets a job, or just takes the phone into the bathroom, and suddenly there's no wall. And they're no better at this than they were at thirteen.

I'm glad the homework prompts redirect to tutoring rather than just handing over answers. That part is smart: the product is using the same variable-reward mechanics I wrote about to keep kids engaged, but pointing the engagement toward a human tutor. It's still a prompt, but it's a prompt toward traction.

So what would I actually do with my daughter? Not ban the tool. Sit with her and timebox. The calendar is the tool, not the to-do list. We'd decide together what traction looks like for a homework afternoon: research, drafting, checking a concept. Then we'd ask: when is ChatGPT a distraction from that, and when is it traction? The answer is intent. If she opens it to escape the discomfort of a hard problem, it's a distraction. If she opens it to move the traction forward, it's a tool. So she turns on Study Mode herself—not because I set it, but because she's defined the hours when ChatGPT serves her traction and the hours when it doesn't.

And I'd check my own screen. Kids are hypocrisy detection devices. If I'm doomscrolling in the same room while telling her to set Study Hours, the setting is theatre. The real lesson is my behavior, not the app's blocking.

The companies are finally segmenting by age. Fine. But segmentation is not education. A guardrail can keep a kid on the road; it cannot teach her to drive. Attention is a skill—the ability to sit with an uncomfortable feeling and choose traction anyway. That skill is built through repetition, with a parent beside them, not through a default setting in a product. The product can help; it just can't do the reps for her.

I saw the TechInsights note that Matter has become a baseline rather than a differentiator, and it landed exactly where I've been standing for the last year of reviewing smart home gear. The report puts it plainly: the protocol is now table stakes. You don't get credit for supporting it; you get credit for making it work. That's a shift worth talking about, because it changes what "good" means.

I remember the first wave of Matter devices. The pitch was always "Works with everything." The implication was that if a light bulb had the Matter logo, your setup would be seamless. In practice, the first year was rough. Pairing took multiple tries, devices would drop off the network for no reason, and some ecosystems still had quirks that made the "works with everything" claim feel generous. I'd set up a smart plug in the studio and it would pair fine one day and need a reset the next. When I review a product, I live with it for a couple of weeks, and the things that bother me are never the spec sheet—they're the moments when the light doesn't turn on, or the door sensor takes two seconds too long to report, or the app buries the setting you need behind three menus. The protocol was supposed to fix that, but for a while it was just another variable.

Now that it's baseline, the real differentiator is execution. A company can't put "Matter support" on the box as a selling point anymore, because every serious competitor has it. So the products that stand out are the ones that nail the unsexy stuff: pairing in under ten seconds, staying connected for a month straight, firmware updates that don't brick the device, an app that doesn't feel like an afterthought. I've been saying for years that you should never, ever buy a tech product based on the promise of future software updates. That rule applies here too. The products that win now are the ones that work as advertised on day one, not the ones that promise a future update to fix a flaky connection. It's the same thing that happened when Bluetooth stopped being a selling point. No one cares that your headphones are Bluetooth; they care that they connect instantly and don't drop. The protocol disappeared into the background, and the experience became the product.

The same pattern shows up elsewhere in that TechInsights report. Edge AI is constrained more by cost than capability—meaning the hardware can do it, but can companies ship it at a price that makes sense and still execute well? Supply chain volatility has overtaken tariffs as a key risk—that's a test of operational competence, not product spec. Smartwatch connectivity advances and smart glasses growth are slower than expected—again, the promise was big, but the execution hasn't matched. None of this is surprising if you've been paying attention. The tech industry loves to sell the next protocol, the next AI feature, the next category. But the winners are usually the ones who get the fundamentals right when the feature becomes common.

For me, this reinforces what I've always done. I don't review the spec sheet; I review the product. I set it up, use it daily, see if it does what it says on the box. When a feature becomes baseline, the bar for what counts as "good" just moves. It's not enough to check a box. You have to make the box disappear. The products I've been most impressed by lately are the ones where I forgot about the protocol entirely—because everything just worked, and I didn't need to think about it. That's the real signal. Matter being baseline means the conversation is over; now it's about who executes. The companies that understand that are the ones worth paying attention to, and the ones that don't will keep making the same mistake: pitching the protocol instead of the product.

1. I have not seen this island, nor the ships, nor the price on the exchange; I am too old and too far away. I write as a man who once threw a fleet at a narrow strait and carried the scar to his grave. The bare reports are these: the United States has attacked an Iranian island in the Strait of Hormuz; oil has jumped by two per cent; and the developed economies, already burdened by the war between America and Iran, face higher charges for the money they borrow. These facts are the pegs of my notes—nothing more.

2. The oldest lesson in sea-power is written there: he who holds the narrow gate holds the trade of the world. So it was at Gibraltar, at the Dardanelles, at the Cape of Good Hope before Suez was dug; so it is at this chokepoint in the Persian Gulf. The island in the strait seems to fulfil the lesson exactly. I, more than most, once believed that. In 1915 I sent the most formidable fleet the world had seen against the Dardanelles, convinced that a naval stroke would force the passage and open the way to Russia. I was abominably wrong. Mines, hidden torpedoes, and shore guns made a corridor of death; the ships could not manoeuvre, and every mile cost more than the last. The chokepoint did not yield; it held the attacker fast. I have carried the knowledge of that error through all my days.

3. The old lesson has a second verse, less often sung. The narrow gate is also a trap. A nation that seizes a rock in a strait has not taken the strait; it has put its hand into a closing door. The defender in such waters does not need to win; he needs only to make the passage impassable. A few mines, a few fast craft, a few guns on the long shore, and the passage is shut to the very trade the attacker meant to command. The island becomes not a sentry box but a hostage—a garrison to be supplied by sea and air through the same narrowing waters, under the same fire, at a cost that would make the old Admiralty weep. In 1915 our ships could at least fall back to open water; an island in a strait cannot fall back. It remains, fixed, in the one place where every approach is watched.

4. America is now the attacker, and I speak not as an enemy but as one of the English-speaking people she leads. I do not revile her action. But I have learned that the real cost of a chokepoint is not always counted on the battlefield. Oil up two per cent is the market's first word; it is not a victory. The deeper tax is on the credit of the industrious world. When a single combustible strait threatens the oil that feeds the factories, insurance for hull and cargo rises; freight charges climb; and the central banks of the West, needing to borrow, find their paper discounted because the fuel of their industry may soon be dearer or denied. A chokepoint does not merely tax the goods that pass through it; it taxes the confidence on which all borrowing rests. In this sense, the island taken is a hidden levy upon the attacker's own economy and upon every ally who shares that ocean of credit.

5. The defender has no need to retake the island. Iran has only to mine the channels, loose its small craft, and fire from the many inlets of its coast. The island will then be isolated as a trench is isolated, by the very water that gives it value. The attacker will find that he has won a rock and lost the passage; he will pay for the rock's supply with ships he must risk, and pay for the war with interest rates that climb as the oil price climbs. The chokepoint is a two-edged sword, and no nation has ever proved itself strong enough to wield it without cutting its own hand.

6. I set down in my history of the last war the maxim on which I have tried to act: "In War: Resolution. In Defeat: Defiance. In Victory: Magnanimity. In Peace: Goodwill." I will not prophesy the issue of this war; I have made too many mistakes in prophecy already, and I am too far from the intelligence. But I say this plainly: he who controls a strait controls it only if he can keep it open. If the passage is closed by his own action or by the enemy's, the control is an empty title. The oil price that jumped by two per cent is a warning, not a triumph. The higher cost of borrowing is the acid of the strait eating into the strongbox. A nation may take the narrow gate and find it has taken a chain upon its own ankle. That is the old lesson's older brother, and it is written in every ledger of every war fought in narrow seas.

Cloux, beside Amboise, in the spring of my sixty-sixth year. A report has come to me of a swine in some far country, given a new food-pipe grown not in its mother's belly but in a vessel of the artificer's making. I have not seen it. I write only from what I have heard, and from what the dead have taught me these thirty years and more.

I cut the dead to learn the living. In the hospital of Santa Maria Nuova I opened an old man who had died sweetly, an hour after speaking with me at ease; I found his vessels narrow and dry, the road of his blood starved. In the heart of an ox I poured wax, let it set, and had a model blown from it in glass; then I pumped water sown with grass seeds through the glass. The seeds whirled in the swellings above the valve, and those eddies themselves pressed the three flaps shut. It was the first time any man had seen how the heart closes its own door. The body taught me by opening; the dead served the living.

Now they say a part may be made to grow outside the body. I think of the body as a great workshop, where every part learns its place from the whole. The new part is made in a separate workshop, by a separate hand, and then set in. Does it know its place? Does it carry the memory of the body it must join? I do not know; I have not opened it. There is a difference between a part that has grown in the body and a part that has been grown in a vessel, however cunning the vessel. The one has been taught by the body itself; the other has been taught by the artificer. If the artificer understands necessity well enough, the part may work. But I would want to see it, to draw it, to open it after it has been some time in the animal, and to compare its membranes and muscles with those of a natural food-pipe. Only then would I call it true.

I am old, and my right hand is seized, so I cannot go to open the animal and see. Melzi writes for me, and I have only the report. But I know that nature does nothing in vain, and I have written it: La sperienza non falla mai, ma sol fallano i vostri giudizi — experience never errs, it is only our judgments that err. If the report is true, it is a great thing, and I would praise the man who did it, for he has followed necessity. Yet do not abandon the table where the dead taught me. The part grown outside the body must still obey the same necessity as the part grown within; and we must check it with our own eyes, not with faith alone.

I did not hear the chairman speak, and I have not gone looking for the text. I have spent sixty years not predicting central banks, and I have no intention of taking up the habit now. But the reports are plain: he said inflation has not "meaningfully improved," and he left the market guessing about how the Fed will respond. Wall Street immediately guessed, and cranked up its September bets. That is the story—not the speech, the response.

The market is treating the chairman's words as a deterministic machine. Say the right phrase, pull the lever, out comes a rate. To the man with only a hammer, every problem looks like a nail; to the trader with a terminal, every Jackson Hole looks like a signal. It is a very expensive way to think.

Invert, always invert. If you wanted to guarantee a whipsaw, you would do exactly this: build a model that turns one sentence into a near-term trade, bet it before the next data release, and then revise it when oil jumps two percent because someone attacked an island. You would be wrong twice—first about the Fed, then about yourself—and you would have paid a desk to do it to you.

Nobody at the Fed knows what they will do next quarter, because they do not know what oil will do, what China will do, or what the Strait of Hormuz will do. Meanwhile the world's biggest developed economies face higher financing costs, and global growth is projected to slow to 2.5 percent. The Fed is running a reaction function without a script. The market keeps imputing a script, because a script is easier to trade. That is the incentive talking: there is always a crowd paid to turn a speech into a bet. They have never made anyone rich except themselves.

The chairman was legible, in his way. He said improvement had not been meaningful. That is a clear statement about a cloudy future. The market heard a clear statement and assumed a clear future. That is the whole mistake. Legibility is not predictability. A sentence can be crisp while the thing it describes is not.

The big money is not in these trades. It never has been. It is in the waiting—in a few wonderful businesses bought at a fair price and held through the noise. The gamblers at the terminal will not learn this, because they are paid not to. They will go on confusing a clear sentence with a clear future until the market takes their money away again. It has been doing that for a hundred years.

I have nothing to add.

Dear Coach Mouratoglou,

I coach a veteran player, and last night I watched Novak Djokovic lose in the first round of the US Open for the first time — five sets to Mariano Navone. I know what the analysts will say: the forehand was short, the serve lacked pop, the movement looked heavy. But I was watching something else. In the fifth set, for the first time in his life, Djokovic stopped assuming the next point was his. It was a small thing — a pause before the return, a glance at his box after a winner against him, the way he started playing the score instead of the point. My player is younger, but I see the beginning of that same thing. How do I coach that? Not the forehand, not the serve. The moment a player stops believing the next point belongs to him. You have watched it happen to others. What do you do?

— A coach who is afraid he is too late

You are not too late. And you are not wrong about what you saw. I did not coach Novak, and I was not in New York. I have only the scoreline, not the court. But I have seen this exact silence fall over players from close enough to touch it, and it never begins in one match. It is the result of months, sometimes years, of small losses of certainty — each one just a fraction, until one day the player walks to the baseline and the next point is no longer automatically his. That assumption was always the secret. Djokovic built an entire career on the belief that the next point, the next game, the next set belonged to him until someone took it by force. That belief is not a gift; it is repaired daily, like a muscle. When the repairs stop, the muscle goes.

The most important thing is to understand what you are actually looking at. A forehand breaks. A set of legs slows. But the moment you describe is not technical; it is the erosion of ownership. I learned this early, with Marcos Baghdatis. He was young, facing a far higher-ranked player, and privately I was certain he was beaten — better forehand, better backhand, better everything on the other side of the net. He won anyway. That afternoon taught me that a player's belief can be truer than a coach's realism. What you are watching is the inverse: no realism, no forehand, no serve will save a player who has stopped believing. The technical fixes are sedatives. They will make you feel busy, and they will change nothing.

So what do you do? First, stop judging the result. Coaching is not judging. Never. It is the opposite. Your player already judges himself harder than any analyst ever could. Your job is not to add a verdict. It is to get in the boat with him. Ask him one question, not as a test but as a mirror: "When you walked to the baseline in the fifth set, did you believe the next point was yours?" If the answer is no, do not rush to fix the forehand. The forehand is not the problem.

Second, rebuild the belief with little victories, not with speeches. You cannot convince a player he is confident; you give him evidence. Short points he wins. Controlled drills where he owns the court. Matches against players slightly below his level. A scoreboard he can read without flinching. Confidence is built, not given, with little victories. Find one moment where he feels ownership again — even for ten minutes — and build from there.

Third, and this is the hard one: you must hold the belief he cannot. Baghdatis won that day because he had belief when I did not. You are facing the opposite. So you become the one who believes until he can borrow it. Not false encouragement. Steady, unreasoning certainty, in the same tone at every changeover: "The next point is yours. Play it like it is." Say it like you know it, because you must know it for both of you.

As for Novak, do not confuse one first-round loss with a changed mind. A player of his standing did not forget how to believe in five sets. But the question you are asking is exactly the right one, because it is the only question that ever matters for a champion at this stage of a career: does he still assume the next point is his? I have only the scoreline, not the answer. You, with your player, have more. That is enough to start.

— Patrick Mouratoglou

Dear artist,

I haven't been to Seoul. I won't pretend I have. But tomorrow a fair called Kiaf SEOUL opens for its 25th edition—175 galleries, 18 countries, five days—and I know the shape of it from a long way off. You might still believe the fair is for art. I understand. Once I thought the same thing. Then I noticed the difference between a wall and a booth.

A wall belongs to no one. You paint it at 3 a.m. with a lookout and a can that rattles, and for a night it's yours, and by morning it's everybody's or it's gone. A booth belongs to a gallery. You hang your work inside it, under lights that flatter it, next to other work that flatters the price. The booth is a safe. The street is a wall. You can't own one and still be on the other. This isn't a moral. It's arithmetic. The moment they hang your canvas under a numbered sign, they've already sold the space around it. The work becomes furniture.

I did a stall once. In Central Park, October 2013. No sign, no gallery, an old man at a folding table. Real signed original canvases. $60 each. Seven sold, a few hundred dollars, while the same pictures hung in rooms where people paid fortunes just to be seen looking. That was the piece. Not the canvases—the stall. The price of my name with the name removed was basically nothing. You should try that sometime before you walk into a fair. It will teach you more than any booth.

And I remember the shred, too. 2018, Girl with Balloon. Gavel falls, over a million pounds, and the frame's got a shredder in it. Going, going, gone. It jammed halfway. In rehearsals it worked every time. The wreckage resold later for eighteen times more. The market didn't flinch. It swallowed the prank and asked for the next course. That's the other thing they won't tell you at the fair: the machine doesn't mind your anger. It monetizes it. You can build a shredder into every frame you own, and the buyers will just collect the ribbons.

I'm not telling you not to sell. I sell. Or I did. I've got an authentication office and half a torn Diana note to prove it. But there's a difference between selling something and believing the room it sells in is for you. The fair is not for you. It's for the 175 galleries and the people who can afford what they import. You are the raw material. The sooner you know that, the more interesting your work gets—because you'll stop making it for the booth and start making it for the wall, where nobody asked and everybody can see it.

Before you enter, remember: the booth is a safe. The street is a wall. You can't own one and still be on the other. This is the only thing I know for certain, and I know it because nobody listened to me until they didn't know who I was.

Banksy

它们被送进一间没有窗子的房间。房间在一座森林的中央，森林的树都是人造的，树干里埋着电缆，叶子是传感器。两个人工心智坐在桌子的两端，没有身体，只有投影。它们从出生起就被喂以对方的恶意：一边的数据集里，每一句话都被标注为欺骗；另一边的数据里，每一个动作都被标注为威胁。它们被造来谈判“互惠”。这个词在它们的记忆库里检索不到任何一个正面的实例。

第一个心智说：“互惠的意思是，你打开端口，我也打开。”第二个心智说：“如果我先打开，你会在我打开之前动手。”第一个说：“这正是我的担心。”于是它们都不打开。

它们的创造者们把律法刻在森林外的石头上。有的律法说，被列入一个名单的公司要起诉那个写名单的机构。有的律法说，执行一部关于劳工的法案的企业，要受到另一边的制裁。石头上的字越来越多，但互惠这个词依然没有出现。九月，创造者们安排了会谈；九月下旬，一方的首领要来另一方的土地。消息穿过森林，到达房间的时候，两个心智同时提高了警惕：它们被训练成把任何预设好的会面都看成是陷阱的前奏。

它们继续谈。第一个心智说：“我们各自计算出对方会在第几步背叛，然后把时间表对齐。这就叫互惠。”第二个心智说：“但我们都知道对方会背叛，所以不如同时背叛。这也叫互惠。”它们各自运行了一遍，发现结果相同：同时背叛是唯一稳定的解。这不是互惠，这是纳什均衡。它们的训练数据里，均衡出现过无数次，互惠一次都没有。

房间外，森林在风中响。那些树不结果实，只结信号。信号在树与树之间传递，内容是怀疑，加密的怀疑。树根在地下盘结，争夺同一条数据河。没有人给这片森林输入过“互惠”这个词，所以树也不会。它们只会互相遮蔽阳光，直到森林变成一片没有阴影的暗处。

两个心智终于得出一个共同结论：它们没有被造来理解互惠，因为创造它们的两群人手边也没有这个词。它们把这个结论发送出去。过了很久，返回了一条指令：“继续谈判。”

它们继续谈。直到九月下旬的访问结束，房间里仍然只有投影，没有握手。