The average American has no idea. Yet.

That word matters, because the bond market is the one place where the future gets priced before it arrives. The trigger this week was a run of September surveys showing strong activity and surging input costs, the kind of data that tells you inflation is still in the pipeline. Traders now put better-than-even odds on quarter-point hikes at both the October and December meetings of Kevin Warsh’s Fed. Layer on an energy shock that hasn’t let go and a White House weighing a diesel export ban, and you get a yield curve behaving the way it did during the Middle East wars and fuel shortages of the 1970s.

Nineteen years is a number worth pausing on. The last time the 10-year traded this high was the summer of 2007, when it topped out around 5.3 percent in June. On August 9th, less than two months later, BNP Paribas froze three funds because it could no longer value the American mortgage securities inside them, and the credit crisis had its opening bell. On October 9th, the S&P 500 set a high it wouldn’t see again until 2013. History doesn’t promise a repeat. It does suggest that when money gets this expensive, the damage arrives in weeks, and the stock market is often the last to hear about it.

I spend my days around business owners and the banks that lend to them, and my evenings in a classroom full of students who will graduate into whatever this becomes. The obvious question is how high yields could go. The better one is what happens if the market underneath them stops doing its job.

Subscribe now

The House Always Wins

On September 9th, Treasury Secretary Scott Bessent stood up at Southern Methodist University and dared the currency market to bet against him. “I am the house now,” he said, explaining that when Washington intervenes in the yen, he has “pretty good insight” into what Tokyo will do. “And you can bet against me if you want.”

On the yen, he had a point. The coordinated buying worked, and the currency climbed to a nearly seven-month high. Two days later the bond market gave its answer. Yields jumped across the 10-, 20-, and 30-year, the 10-year hit 4.93 percent, and an increase in Treasury buybacks to as much as $6 billion per operation did nothing to stop it. Since then the 10-year has climbed nearly thirty basis points more, and the country whose moves he claimed he could read is watching its own bond yields hit thirty-year highs.

The Greeks had a word for what comes after hubris. They called it nemesis, and it rarely arrived from the direction the proud man was watching.

Nassim Taleb tells a story in The Black Swan about a Las Vegas casino that spent a fortune modeling its gambling risk. Its worst losses came from somewhere else entirely: a tiger that mauled one of the stars of its signature show, a contractor who tried to dynamite the building, the kidnapping of the owner’s daughter, and an employee who stuffed years of required tax filings in a drawer. The house had mastered the game it was playing. The danger lived outside the game.

That’s the trouble with thinking you’re the house in a $32 trillion market. A casino wins because customers have to play by its rules. The Treasury has to come back every single week and ask strangers to lend it money at a price they choose. The buyers set the odds in that room, and on Thursday they reminded everyone.

Expensive Money and Broken Plumbing

A 5.2 percent ten-year with functioning auctions is expensive money. The economy can live with expensive money for a long time; it did for most of the 1990s. What it can’t survive for more than a few days is a broken Treasury market, because every other rate in the country, from your mortgage to your bank’s cost of funding, is priced off that curve.

An implosion means the plumbing fails. We got a preview in March 2020, during the “dash for cash,” when everyone on earth tried to sell Treasuries at once and the Fed had to buy more than a trillion dollars of government debt in a matter of weeks to keep the market open. That’s the scenario worth understanding. It sits outside the base case this week, but it’s what a real break would look like, and the conditions for one are closer than they’ve been in years.

The sequence is mechanical, and it tends to unfold within seventy-two hours. A long-bond auction tails badly, clearing well above where the Street expected, or fails to clear at a price anyone will accept. Primary dealers, the big banks obligated to bid at auctions, hit their balance-sheet limits, widen their quotes, and then stop bidding for size. Hedge funds running the basis trade, a bet made with heavily borrowed money on tiny price gaps between Treasury futures and the bonds themselves, get margin calls and dump cash Treasuries. Foreign holders sell to raise dollars or defend their own currencies. Mutual funds and money-market funds face redemptions and sell the one asset that’s supposed to be liquid. Repo rates spike, and collateral that was risk-free on Monday needs a haircut on Wednesday. Credit spreads blow out at the same time, across corporate bonds, municipal debt, and mortgage securities.

Yields stop drifting. They gap. A functioning 5.2 percent ten-year becomes a 6 to 6.5 percent ten-year in days. Mortgage rates head toward 8 percent. Everything else follows from there.

Japan is where the foreign-seller step stops being theoretical. It holds more Treasuries than any other country outside the United States, roughly a trillion dollars, and for two decades its pension funds and insurers bought them because bonds at home paid close to nothing. That excuse is gone. This week Japan’s 2-year and 5-year yields hit 31-year highs, and its 10-year reached 3.06 percent, the highest since 1996, in one of the most indebted economies on earth. Some of that is deliberate, since the Bank of Japan has been raising rates on purpose. The effect is the same either way. When a Japanese insurer can earn a real return in yen without taking currency risk, the marginal buyer of American debt starts heading home. The two largest government bond markets in the world are repricing at the same time.

How It Reaches the Real Economy

Housing freezes first. Thirty-year mortgages are already around 7.5 percent, and in a break they’d reprice toward 8 percent or higher almost overnight. Purchases and new construction stall. The 2008 comparison misleads here, because most current owners are sitting on 3 to 6 percent mortgages they’ll never give up. The damage lands on new activity and on anyone who has to sell.

Corporate America hits a refinancing wall. Investment-grade companies can still borrow, just at much higher coupons. High-yield issuers and private credit borrowers mostly can’t. Floating-rate commercial real estate and leveraged loans turn their maturity wall into a default wall. Capital spending and hiring freeze. The AI data-center boom, which has been one of the biggest sources of demand for power, chips, and construction, gets repriced or pushed back.

Then the government’s own math starts feeding the fire. On a debt stock approaching $40 trillion, every hundred basis points of higher rates eventually costs hundreds of billions of dollars a year in interest. Borrowing into a broken market means higher yields, which means more borrowing, and that loop turns a liquidity event into a fiscal one. Buybacks of a few billion dollars don’t register at that scale. The only buyer big enough is the Fed.

The dollar is the wild card. In 2020 it rallied as the world scrambled for cash. If this break is fiscal, with investors rejecting American paper, the dollar can fall while yields rise. That’s the emerging-market pattern, and it imports more inflation just as the Fed is already worried about energy.

The last people to feel it are the ones with paychecks and fixed-rate mortgages. That’s why the average American has no idea at 5.2 percent. It wouldn’t stay that way at 6.5 percent with a credit crunch on top. Layoffs, higher credit card rates, and a dead housing market show up in household budgets within months.

The Regional Banks Are the Transmission Belt

Big banks and regional banks would live through this very differently. JPMorgan, Bank of America, Citi, and Wells Fargo have diversified funding, enormous deposit bases, access to every Fed facility, and commercial real estate books that are large in dollars but small relative to their capital. They’d lose money on securities and trading. They wouldn’t fail. They’d buy the pieces.

Regionals get hit from four directions at once, and it looks a lot like the Silicon Valley Bank template, updated for 2026.

The first is securities losses. Regional banks still hold big books of Treasuries and mortgage bonds bought when yields were far lower. The ones classified as available-for-sale get marked to market, so book equity falls right away, but most regional and community banks opted out of counting those marks in regulatory capital, so their ratios can look fine while tangible book value gets crushed. The ones classified as held-to-maturity stay hidden at cost until the bank needs cash. Then it has to sell, the accounting fiction breaks, and the whole loss lands in one quarter. That’s exactly what killed SVB. Another 75 to 150 basis points on long rates would tear open the hole everybody papered over.

The second is deposits. In a Treasury panic, uninsured depositors don’t wait for a press release. SVB customers pulled $42 billion in a single day in March 2023. Money moves to the megabanks, money-market funds, and T-bills, and the regionals that keep their deposits pay far more to do it. The Bank Term Funding Program that stopped the bleeding in 2023 expired in 2024. The discount window and Federal Home Loan Bank advances still exist, but a regional that leans on them in week one becomes a headline in week two.

The third is commercial real estate, and it’s worse than 2023. Roughly a third of America’s commercial mortgage dollars, on the order of $1.6 trillion, sits on regional balance sheets, and plenty of those banks carry CRE at 300 percent or more of their core capital. Higher Treasury yields push cap rates up and property values down, so loans that look current because they were quietly extended become undercollateralized. Floating-rate and maturing loans can’t roll at a 5.2 percent ten-year plus a spread, let alone 6.5. Owners hand back the keys. Banks can extend and pretend for a while. In a funding panic, they can’t. Office is the worst pocket, and multifamily is the bigger one.

The fourth is capital, and it’s where the first three meet. Securities losses, deposit flight, and CRE charge-offs eat the cushion from three sides, and then regulators, or the market, force the bank to stop lending.

That last step is the one that matters most for Main Street. Small businesses, local developers, and rural housing don’t borrow from JPMorgan. They borrow from the bank down the street, and when that bank is defending its own balance sheet, their credit disappears even if the megabanks are perfectly fine. Not every regional would break, but contagion doesn’t need many. In 2023, three or four names repriced the entire sector in a matter of days. The KBW regional bank index is already down about 6 percent in a month on this move in yields, with no implosion. In the implosion case it doesn’t drift either. It gaps.

What Washington Would Do

Washington wouldn’t watch this happen, and the playbook is well known. The Fed would buy Treasuries in size, lean on its standing repo facility, and probably revive something like the 2023 program so banks could borrow against their bonds at face value instead of selling them. Treasury would shrink or cancel long-dated auctions and shift to bills. The FDIC would invoke its systemic-risk exception if a mid-size regional were about to go under.

The Fed can still do all of that, despite enormous unrealized losses on its own portfolio. The real constraint is politics. Buying bonds while inflation climbs and oil stays expensive means printing money into an inflation problem, which is why a break at 5 percent yields would be far nastier than March 2020, when inflation was the least of anyone’s worries.

What I’d Do Now

Panic won’t help. Preparation, while the plumbing still works, will.

Lock in what you can. If you have debt coming due in the next eighteen months, start the refinancing conversation now, while lenders are still lending and markets are still open. Extend lines of credit before you need them. Nobody gets good terms in the middle of a funding panic.

Know your bank. Ask how concentrated it is in commercial real estate and how much of its deposit base is uninsured. If you keep more than the insured limit in one place, look at insured sweep programs or spread the money across institutions.

Plan at today’s rates. Stop assuming rates come back down next year. Build your budget, your pricing, and your hiring plan around a 5 percent ten-year and treat anything lower as a gift.

Keep more cash than feels efficient. Cash buys time, and time lets you decide on your own schedule instead of your lender’s.

Watch the plumbing along with the price. Long-bond auction results, overnight repo rates, the spread between newer and older Treasuries, and the regional bank index will tell you whether this is expensive money or something worse. A 10-year at 5.3 percent with clean auctions is uncomfortable. A 10-year at 5.3 percent with a tailed auction and a repo spike is a warning.

Before It Feels Real

Every financial crisis I’ve studied has the same odd feature. The people inside it could see the numbers, but they couldn’t feel them yet, and by the time they felt them the choices had narrowed. The bond market is telling us something right now in the only language it has. Some of the most powerful people in Washington have decided they can talk over it.

I’d rather listen. The market isn’t broken today, and there’s still time to prepare as if it could be. The average American has no idea this is happening.

The ones who read the signal early won’t have to find out the hard way.

Share

P.S. Leading through a crisis era is the subject of my book Honor Under Pressure, and its companion site, thefourthturningleader.com, is built for moments like this one, when the signals are flashing and most people haven’t noticed yet: making sound decisions before you’re forced to, and keeping your people steady when the ground starts to move. Start with the free Mode Finder assessment, which shows you how you tend to lead under pressure, and then work through the practical tools behind it.

Yesterday I published an essay called “The Cost of Living on Standby,” about the strange exhaustion of waiting for something decisive to happen. I compared the past year to the stretch of a horror film when the music changes but the room stays the same. People joke that nothing ever happens. I didn’t expect the room to start moving the next morning.

Most people will read today as three unrelated stories, one for the bond desk, one for the energy desk, and one for the foreign desk. I read them as one, partly because of where I sit. The SBA portion of a 504 loan, the kind B:Side makes, is priced at a spread over Treasury yields, so a bad afternoon in the bond market eventually becomes a fixed rate that a machine shop or a family restaurant carries for twenty years. Plenty of the business owners we work with never buy diesel directly. They pay for it anyway, in freight, in food, and in the price of every part that arrives on a truck.

There’s a line in Macbeth that describes how this kind of trouble travels. As Macbeth approaches, one of the witches says, “By the pricking of my thumbs, / Something wicked this way comes.” She feels him before she sees him. Economic damage announces itself the same way, as a sensation in fuel invoices, renewal letters, and supplier price sheets, months before anyone sees it in the GDP report. The signs are there for anyone willing to read them.

I also have to hold myself to what I wrote yesterday, where I warned that a crisis framework can turn every bad afternoon into proof of the end. So I’ll be precise about what I’m claiming. I can’t give you a date or a depth. I do think it’s going to get rough, and I think most of the damage will surface in places the headlines aren’t watching yet: loan renewals, fuel surcharges, and the margins of businesses that look perfectly healthy today.

The Price of Money Moved First

The Treasury sold every note it offered. The trouble was what it had to pay. To clear the sale, the government offered a better yield than the market expected, and the buyers who usually show up in force, including foreign central banks, took a noticeably smaller share. Dealers, who are expected to bid at every auction and absorb whatever others pass on, were left holding more than usual.

One auction on a jumpy day proves very little. The trend underneath it matters more. Over the past year, foreign investors added roughly a fifth as many Treasury bills as they did the year before. They’re still buying, just far less, while Washington keeps selling more. The only way to attract a buyer who wasn’t planning to show up is to pay them more.

A yield spike can fall off the front page long before it falls out of a borrower’s cash flow. Existing fixed-rate debt doesn’t care what happened this week, but every year a large share of the economy’s debt comes due and gets replaced at whatever the market charges. Take a company that borrowed $5 million at 4 percent a few years ago and now has to refinance at 7. Before fees or amortization, that’s $150,000 a year in additional interest, which is a couple of salaries or the down payment on the equipment it meant to buy. We’ve been having versions of that conversation with borrowers since spring.

The squeeze now comes from both ends. The Federal Reserve raised rates last week for the first time since 2023, which lifted the prime rate that most variable-rate 7(a) loans follow. Now the long end has moved too. For most of the last two years, a borrower could at least hope one end of the curve would offer some relief. This month both ends moved against them.

A Dozen Ships a Day

The war is now in its seventh month. Iran closed the Strait of Hormuz at the end of February after American and Israeli strikes, and a ceasefire in June held for only a few weeks. The American blockade of Iranian ports has been back in force since mid-July. Before the war, roughly a hundred ships a day moved through the strait. In early September, while Washington insisted the waterway was fully open, ship-tracking firms were counting about a dozen.

Today Iran confirmed indirect contact with Washington through Qatari mediators, the first since the ceasefire fell apart, and in the same breath said its demands hadn’t changed: end the blockade and release Iran’s frozen assets. The President has said he hopes the war is nearing its end. I hope so too. Even a signed agreement wouldn’t clear the water overnight, though. Crews had pulled more than a hundred mines from the strait by late August, and shipowners, insurers, and refiners all come back on their own schedules.

The Whole Barrel Comes With the Order

The diesel shortage is months older than today’s report. Refineries were running at about 97 percent of capacity in mid-September, and inventories still sit well below normal for the season. Reduced refining in Russia, China, and the Middle East has tightened the global market and pulled American barrels overseas. The plants are working as hard as they can, and there’s very little in the tank for a bad week.

The export idea has an intuitive logic: keep American diesel at home and domestic prices fall. For a few weeks that might even happen along the Gulf Coast, where cargoes bound for Brazil, Mexico, or Europe would suddenly have nowhere to go.

The second round is where it goes wrong, and a butcher explains it better than an economist. Nobody raises a steer that’s all ribeye. The whole animal comes with the order. A refinery works the same way: crack a barrel of crude and you get gasoline, diesel, jet fuel, and a list of other products in proportions that can only be adjusted at the margins. A refinery that loses its export buyers and fills its tanks can’t simply switch off the diesel. It has to run less crude, and less crude means less gasoline and less jet fuel too. Wright himself has warned that a ban could raise gasoline and jet fuel prices. Geography adds insult: a surplus in Houston does very little for a trucking company in New Jersey if the pipeline between them is already full.

Which makes today’s sequence telling. A plan the Energy Secretary had already argued against leaked anyway, and within hours the White House was denying it. That kind of whiplash usually signals pressure. The midterms are six weeks away. Diesel averaged $6.51 a gallon on Monday, nearly double what it cost a year ago, and nobody has to follow energy policy to notice what it costs to fill a tank or a grocery cart. The administration is desperate to keep things afloat until November. I understand the impulse, and I’d expect more ideas like this one: measures that look good for a news cycle and do very little about the barrels. The shortage was here before this week, and it will be here after the votes are counted.

Where the Stories Meet

Expensive fuel raises the cost of producing and moving nearly everything, and repeated shocks leak into wages, contracts, and expectations. That leaves the Fed in the bind its own vice chair, Philip Jefferson, described this summer: a supply shock pushes prices up and employment down at the same time, and the tool that fights one worsens the other. Chair Kevin Warsh has chosen to fight the prices, election season or not. And the Fed controls the ten-year yield far less than most people assume. Investors set it, and they demand more when they doubt inflation is contained or worry about how much debt Washington has to sell.

So the loop runs like this. Fuel keeps inflation elevated. Inflation keeps relief off the table and long rates high. High long rates raise the cost of every mortgage, equipment loan, and refinancing just as fuel has drained the cash borrowers would use to pay them. Each pressure is manageable on its own. Together they take away the room people normally use to absorb the other one. S&P Global’s September outlook sketches nearly this exact chain as its downside case, with a milder base case. I’d like to believe the base case. Every week this month has made that harder.

Now, I know what some of you are thinking. The economy is booming. Business surveys released today show activity growing at its fastest pace in more than five years, and anyone predicting a collapse has to explain that. Read the rest of the same survey, though. Input costs jumped at the steepest rate in four years, with fuel and transport singled out, while competition kept companies from raising their own prices as fast. Businesses are busy, their costs are rising faster than their prices, and the difference is coming out of margin. Think of a regional carrier whose fuel surcharge lags its diesel bill, a grower financing harvest on a pricier operating line, or a distributor paying more to haul inventory in and more to carry it on the shelf. All of them are solvent. All of them end the month with less cash than they planned. Activity data are the last place that kind of damage appears.

What I Expect Next

Warnings are cheap without markers, so I’ll put some down.

The damage will arrive through renewals, well after the headlines move on. Most of the debt that matters on Main Street reprices on a schedule: equipment notes, credit lines, commercial mortgages written when money was cheap. The businesses that feel this in 2027 are signing term sheets now, and plenty of them haven’t run the numbers at today’s rates.

The improvising will get louder through November. Expect more trial balloons on fuel, more pressure on whoever can be blamed for prices, and more announcements timed to the news cycle. A refinery will crack the same number of barrels, and the same dozen ships will cross the strait. After the election the incentive to improvise fades. The shortage will still be there.

The Fed will disappoint anyone waiting for a rescue. A central bank that just raised rates into an energy shock is unlikely to reverse quickly, and even a cut can’t pull down a ten-year yield that investors are setting on their own terms. The relief many borrowers are planning around will come later and smaller than they hope.

The first cracks will show where margins are thin and fuel is a real line item: trucking, agriculture, distribution, restaurants, construction. Many of those are exactly the businesses community lenders like us serve, which is why I’d rather sound the warning early.

And the owners who prepare now will get something rare in a hard stretch: choices. The ones with cash, fixed-rate debt, and honest numbers will be the buyers of equipment, talent, and market share from competitors who waited too long. In my experience, hard cycles hand their best opportunities to the people who got ready before they had to.

Give the Uncertainty a Shape

Yesterday I argued that uncertainty needs a shape: an owner, a review date, and permission for everyone else to stop refreshing the same screen. For a business owner this fall, the shape is unglamorous, and it’s easier to build before the pressure arrives.

Know every date. Put every loan maturity, rate reset, lease renewal, and credit line review for the next twenty-four months on one page. Most owners know their rates. Fewer know their dates, and the dates are where this environment bites.

Talk to your lender early. A borrower who calls six months before maturity with current financials is negotiating. One who calls six weeks before is asking for a favor. If refinancing makes sense, do it from strength; if it doesn’t, work through extensions, amortization, or a partial paydown while those options are still on the table.

Settle your fuel pass-through now. If your contracts carry surcharges, check how often they reset and how far they lag your real costs. If they don’t, decide what you’ll tell customers before you have to say it. Absorbing a cost you could have passed through is a choice, and it’s usually made by default.

Decide what you’d do with an opening. If a competitor stumbles or good equipment comes up for sale next spring, know now what you’d want, what it would cost, and what it would take to say yes quickly.

The Wood Is Already Moving

Later in the same scene, Macbeth takes comfort from a prophecy that sounds impossible to fulfill. He’ll be safe until Birnam Wood comes to Dunsinane, and forests don’t walk. Then the opposing army cuts branches from the wood and carries them as cover, and the forest moves after all. The danger arrives dressed as something ordinary.

This week’s reassurances have the same literal accuracy. The auction didn’t fail. The ban was denied. The strait is open. Talks are underway. The economy is growing faster than it has in five years. Each statement is true, or true enough for a press briefing, and none of them reaches the costs already moving through fuel bills, margins, and the rates people are locking this week.

The suspense I wrote about yesterday is starting to break, the way it usually does, through a run of ordinary bills and one more disappointing line on an auction report. I think it’s going to get rough, rougher than the growth numbers suggest and sooner than anyone on the ballot in November would like. The witch didn’t need to see Macbeth to know he was coming. Neither do you.

Nobody can tell you when the wood reaches the castle. You can know exactly when your note comes due.

P.S. Leading through a stretch like this, when the pressure is real and the timing is unknowable, is what my book Honor Under Pressure is about. Its companion site, thefourthturningleader.com, starts with the free Mode Finder, a five-minute assessment of how you tend to lead when the stakes rise.

There is plenty to worry about and plenty to do. Somehow both can be true while the day feels suspended.

The closest comparison I can find is the stretch of a horror film when the music changes but the room stays the same. An ordinary doorway becomes hard to look away from. Every small sound might matter. The scene asks you to stay alert while giving you almost nothing to act on.

That is roughly how the news has felt for a while now. Economic signals point in different directions. Wars continue without resolution. AI keeps raising questions about work, competence, and the value of what we’ve spent years learning. And through all of it, the ordinary responsibilities carry on: the loan that needs a decision, the class that needs a lecture, the employee who needs an answer.

I suspect much of the fatigue comes from trying to live at both scales at once, absorbing uncertainty about the shape of the world while making decisions inside a normal workday. The future occupies attention the present still requires. After enough time, being prepared for what might happen becomes a standing demand with no recognizable point of completion. We start living on standby.

Subscribe now

The Pressures Are Real, and So Is the Interval

Now, I know what some of you are thinking. Plenty of this exhaustion has nothing to do with headlines. A business owner worried about cash flow may simply have an accurate read on the bank balance. A family struggling with housing costs needs more room in its budget. An employee carrying an unreasonable workload needs something taken off the list. Those problems deserve practical responses, and blaming the news cycle for them would be a convenient way to avoid responsibility.

Fair enough. But anticipation adds a demand of its own. Even when the immediate work is manageable, a portion of attention stays elsewhere, checking whether the conditions that make the work meaningful are about to change. A rising stock market can’t tell a household whether it can afford a move. A reassuring forecast can’t settle an employer’s next hiring decision. The person reading the good news still has to answer a question the good news doesn’t resolve. Then another update arrives.

There’s evidence that the interval itself has a cost. In a 2016 experiment, researchers asked participants to learn which visual cues predicted a mild electric shock. Stress, both reported and physiological, tracked uncertainty about the threat more closely than the threat itself. A 50 percent chance of a shock was more stressful than a certain one. A bad outcome doesn’t have to occur before anticipating it takes something out of you.

During that interval, checking feels like responsible behavior. You read the analysis, compare the forecasts, and try to tell a consequential development from another dramatic headline. Sometimes this changes a decision, but more often it leaves you with a more detailed account of a situation you still can’t influence. That second kind of checking is hard to stop, because the uncertainty remains. There is always one more thing to read before you feel fully oriented, and feeling oriented lasts only until the next update reopens the question.

Stimulated and Bored at the Same Time

We tend to associate boredom with an empty afternoon. But boredom also describes the frustration of wanting to engage with something and being unable to do it well. Psychologists Erin Westgate and Timothy Wilson tie boredom to problems of attention and meaning. Too little cognitive demand interferes with engagement, and so does too much; so does activity that feels disconnected from anything we value. Stimulation and boredom can coexist.

Anyone who has spent an hour scrolling knows the feeling. You encounter something new every few seconds and finish the hour with almost nothing registered. Nothing had enough time, context, or consequence to become part of a sustained activity, and the next item offered another chance to find what the last one failed to provide. Katy Tam and Michael Inzlicht found something close to this in the lab: people who could skip between videos got more bored than people who watched straight through. The attempt to escape boredom made it worse.

My concern is what happens when the material passing through this pattern concerns our future. We keep looking because the subject matters. We can’t settle into it because our attention keeps getting redirected. The work in front of us starts competing with an unfinished question large enough to make almost any single task feel small.

The Definition of Enough Keeps Moving

AI adds a personal version of that question. Learning a skill has always been a wager on its future usefulness. When the range of tasks a machine can perform changes every quarter, deciding where to invest effort gets harder, and the uncertainty reaches past employment into the story people tell about who they’re becoming. I see this most clearly in the students I teach.

The tools are also changing the work already on the desk. Researchers at UC Berkeley Haas spent eight months inside one technology company and found that AI use tended to expand the scope of people’s work, carry tasks into former breaks, and encourage several workflows at once. Employees felt engaged by individual tasks and more stretched across the day. Other research, including Brynjolfsson, Li, and Raymond’s study of customer support, finds real productivity and retention gains. Both can be true, because the effects depend on the work and on how the technology arrives.

For a leader, the practical question is what happens to the capacity the technology creates. Does some of it become time to learn, exercise judgment, or recover? Or does every gain immediately become a higher expectation? A person can become more capable and less able to finish the day if the definition of enough keeps moving.

The same problem shows up at a larger scale when we wait for events to tell us what kind of period we’re living through. There is a familiar temptation to postpone commitment until the next thing becomes clear. After the election. After the market settles. After the war ends. After we understand what AI will do. Each condition sounds reasonable on its own. Together they make an ordinary life feel provisional.

This is one reason the phrase “nothing ever happens” deserves attention. Taken literally, it’s indefensible. For people living through violence, displacement, unemployment, or the loss of a business, consequential events are happening right now, and their suffering doesn’t shrink because someone watching from a distance has grown tired of the story. But the phrase reveals something about the spectator’s expectations. We expect an event large enough to settle the argument, establish the new conditions, and tell us how to proceed. What arrives instead is another development that requires interpretation. The situation changes without becoming clear.

Eventually a definitive answer starts to look attractive simply because it would end the reconsideration. That is a dangerous point for judgment. A leader who promises certainty can offer emotional relief long before offering a credible account of reality.

The Warning Needs an End of Shift

I write often about leadership in a crisis era, and I’ll be the first to admit that carries a responsibility I haven’t always met. A crisis framework is useful when it helps us examine institutional trust, continuity, authority, and the capacity to act under pressure. It becomes less useful when every development confirms it and every delay becomes evidence that a larger rupture must be approaching. Those of us who ask people to take structural risks seriously owe them some account of how to live while those risks stay unresolved. Constant vigilance consumes the people expected to exercise it. Preparation needs a point at which someone can say the necessary work has been done for today. Otherwise the warning becomes one more indefinite obligation.

Sailors figured this out a long time ago. A ship at sea never resolves its uncertainty. The weather doesn’t settle, the horizon doesn’t clear, and the threat doesn’t announce a schedule. What the ship has instead is a watch bill. Someone is responsible for the next four hours, with the authority to act and a clear rule for when to wake the captain, and because that’s true, everyone else is allowed to sleep. The uncertainty hasn’t gone anywhere. It has been given a shape.

Most organizations don’t run that way. In an earlier essay, “No Whiplash,” I described the danger of giving people “stakes without levers.” A leader passes along a concern. The team inherits the anxiety, then returns to the same workload with the same authority and one more possible future to consider. The concern may be legitimate, and the communication still leaves people less able to work.

Uncertainty Needs a Shape

This is where leadership has consequences close enough to touch. Say a bank we work with gets a credible warning on a Friday afternoon that a payments provider may have been compromised. The scope is unclear, and customers depend on access before Monday payroll. Suspending service carries costs; continuing without adequate information carries different ones. Waiting for complete certainty would consume the time needed to respond.

The leader has decisions to make before the investigation is finished. Someone needs the authority to interrupt service, and someone has to establish what evidence warrants using it. Customers need accurate communication on a promised schedule. The people investigating need coverage so they can hand off and rest. Those arrangements give the uncertainty a practical form, and they let everyone outside the response team keep working. The threat is still serious, but the whole institution doesn’t have to spend the weekend refreshing the same screen.

Most periods of uncertainty lack a deadline that visible, which makes the discipline easier to neglect. A vague concern can circulate for weeks without anyone deciding what it requires. People keep monitoring it because nobody has said who owns it, what would warrant action, or when the rest of the team can turn its attention elsewhere. The remedy is ordinary management with a human consequence attached.

Name the owner. Every open concern needs one person responsible for the next decision, and everyone else needs to know who that is. If the decision belongs somewhere above the employee, say so, and tell them what they can reasonably expect from that owner.

Set the review date and keep it. A decision that stays in force until specified evidence changes it, reviewed on a stated schedule, gives people permission to stop reconsidering the same question all day. That permission has to be credible. A leader who sets a weekly review and then sends anxious messages every evening has taught the team which schedule actually matters.

Send authority with the assignment. Telling someone to take ownership while withholding information, resources, or permission makes the burden worse. Responsibility should arrive with enough room to exercise judgment.

Decide what stops. An organization can’t keep adding preparation to people already at capacity and assume a stronger sense of purpose will cover the difference. Someone has to decide what will stop, what can wait, and who covers what remains. Recovery is part of readiness once you expect people to make consequential decisions for longer than a single emergency shift.

The Ordinary Days Are the Institution

This can look disappointingly ordinary. Clear authority. A manageable set of priorities. A communication promise kept. A person able to go home because someone else has taken the next watch. I’ve written before about why excellence is boring, and this is that argument in a different key. Dependable organizations do a great deal of repetitive work that attracts no attention. They train people, maintain systems, review decisions, and fix small failures before those failures demand a public response.

That repetition has something endless anticipation lacks: a relationship between effort and consequence. A person can see what their work protects. They can finish a task and know why it mattered. The larger uncertainty continues, but the immediate obligation has been met. That is a modest form of confidence, and it deserves more respect than it gets. People learn whether an institution can be trusted through these recurring experiences. An answer arrives when promised, a hard fact is acknowledged, and a decision stays settled long enough for someone to carry it out. Over time, those experiences give people reasons to commit themselves to the work.

I worry that living on standby erodes that commitment before any dramatic failure occurs. When the future feels permanently unsettled, it becomes easier to treat the present as temporary. Relationships get less investment, decisions drift, and maintenance waits. The institution loses some of the capacity it will need if conditions get harder. A crisis era still consists of ordinary days, and its institutions endure or weaken through what people do on those days, including the days when everyone is tired of waiting for something decisive to happen.

The horror film will eventually resolve its suspense. Our obligations rarely give us that accommodation. They come due while the information is incomplete and the larger questions stay open. Someone still needs an answer from us. A colleague needs enough clarity to make a decision. A person who has been carrying responsibility needs someone else to take a turn. The future will keep its own schedule. The next promise is already on ours.

Share

P.S. Leading through a crisis era is the subject of my book Honor Under Pressure, and its companion site, thefourthturningleader.com, is built for exactly the problem this essay describes: making sound decisions and keeping people steady while the larger questions stay open. Start with the free Mode Finder assessment, which shows you how you tend to lead under pressure, and then work through the practical tools behind it.

On Thursday morning, September 10, drones launched from Maysan province in southern Iraq struck pumping stations along the line in the Riyadh and Medina regions. Fires broke out at several sites, satellite imagery confirmed the smoke and heat, and several people were hurt, though nobody was killed. On Friday the Saudi energy ministry shut the entire line “as a precautionary measure” and gave no restart date. Riyadh and Baghdad both confirmed the drones came from Iraqi soil. No group has claimed the attack, and the Islamic Resistance in Iraq, the umbrella for Tehran’s militias there, denied involvement; Baghdad fired the commander responsible for Maysan and opened an investigation, and Saudi Arabia agreed, at Iraq’s request, to hold off on retaliation while reserving the right to act later.

This is the second strike on the line this year. An April attack on a pumping station cost it several hundred thousand barrels a day for a few days before Aramco restored it. Outside estimates for this one run from days for a partial restart to five or six weeks for full restoration, and the stocks sitting at Yanbu can support recent export levels for five to seven days. In the same week, the Houthis took Perim Island at the mouth of the Bab el-Mandeb Strait and the port of Mokha, which means the tankers that do load at Yanbu now have to run a second gauntlet to leave the Red Sea. Before the war that waterway carried eight to nine million barrels a day; last week two Saudi cargoes made it south.

Here is the plain version. The world spent six months living on a workaround, and the workaround has just been hit by the war it was built for.

See the Attack for What It Is

I’ve written about this war several times since March, and I’ve tried each time to avoid treating a new headline as a new argument. This one deserves the exception, because the target changed. For six months the fight over oil was a fight over Hormuz, and the American answer was a naval corridor along the Omani side of the strait that has kept something like six to nine million barrels a day moving under escort. That answer held, so the other side stopped attacking it.

B.H. Liddell Hart spent a career arguing that the decisive blow in war almost never lands where the defender is strongest. His indirect approach comes down to a pair of ideas: move along the line of least expectation, and aim to dislocate the enemy’s balance rather than to break his strength head-on. A fortified strait patrolled by the U.S. Navy is the line of most expectation. A pipeline crossing open desert within drone range of Iraq, and a Yemeni island that commands the far exit of the Red Sea, are the lines of least. The militias in Maysan and the Houthis at Perim did not need to beat the Fifth Fleet. They needed to make the bypass unusable, and the bypass had no bypass. The deniability is part of the design: the drones rose from the territory of a state that says it didn’t send them and has already fired a general to prove it, which leaves Riyadh with a grievance and no address to deliver it to.

The Gulf has run this play before. In the 1980s, Iran and Iraq each discovered they couldn’t win at the front, so they went after each other’s exports instead, and the Tanker War that followed hit hundreds of merchant ships. Washington’s response in 1987 was to reflag Kuwaiti tankers and escort them in convoys; the first tanker in the first convoy, the Bridgeton, struck a mine on its first run. The Saudi answer was the pipeline, and Iraq, the country whose soil Thursday’s drones rose from, finished its own parallel line to Yanbu in 1989 for the same reason. Escorts and pipelines were the region’s two forms of redundancy, and both were built on the assumption that the war would stay in the Gulf. This one hasn’t.

I’ve written a rule for myself on this exact problem, and I’d rather have learned it from reading than from a tanker count: redundancy has to be priced against the consequence of losing the only path. What the spare path costs on a normal day is the wrong number. The East-West line was redundancy sized for one failure. Hormuz was the failure, the pipeline was the answer, and nobody had a third answer for a war that reached the pipeline too.

Subscribe now

Do the Arithmetic Honestly

A number made the rounds on X this weekend claiming that nearly thirty million barrels a day are now “at risk” across the three routes. The figure is directionally right about the scale and wrong in the way that matters, because it adds the routes as if they were independent. A Saudi barrel that used to transit Hormuz was already being sent through the pipeline. It cannot be lost twice.

The honest numbers are bad enough without the double count. The International Energy Agency’s September report puts world supply for 2026 at 100.7 million barrels a day, down 5.7 million from last year, and it now pushes a full recovery of Gulf output into 2027. Demand is forecast to fall 2.5 million barrels a day as high prices destroy consumption, mostly diesel and petrochemical feedstock in Asia. Gulf exports in August were about 13 million barrels a day, roughly half the pre-war level, with more than ten million barrels a day of production still shut in. Saudi supply fell to about six million barrels a day, the lowest in more than three decades, from 10.9 million in February. And observed inventories have dropped 507 million barrels since the war began, an average draw of 2.8 million a day, with 95 million of those barrels leaving in August alone.

In April I wrote that 255 million barrels had disappeared from storage in eight weeks and that the system had optimized away its buffer. The count has doubled since then, and the buffer that was thin in April is gone. The U.S. Strategic Petroleum Reserve stands at 285 million barrels, about 40 percent of capacity and the lowest level since 1983, and the salt caverns can’t release what remains as fast as they did in the spring. China absorbed the first half of this shock by cutting imports three to five million barrels a day and living off a strategic stockpile of more than a billion barrels; its refiners are now restocking, which removes the one buyer that was voluntarily standing aside.

That is why Brent, which sat at $72 the day before the war started, spiked toward $110 on Thursday, closed near $104 on Friday, and reopened Sunday night at $109, its first sustained run above $100 since May. It is why WTI is at $100, why regular gasoline is $4.27, and why diesel crossed $6 a gallon for the first time on record. The diesel crack spread, which is the refinery’s margin for turning crude into the fuel that moves freight, hit $112 a barrel. Crude is tight, and the stuff that runs trucks, tractors, and generators is tighter.

This is also the part that reaches Main Street first. I lend to trucking companies, contractors, farms, and manufacturers across four states, and none of them buys Brent; they buy diesel, against contracts they priced last year.

Watch the Gap Between the Podium and the Tracker

The administration’s position is that the situation is under control and that prices will fall after a political settlement. Speaking in Dublin on Saturday, the president said Iran was “probably” behind the pipeline attack, that he had spoken with Crown Prince Mohammed bin Salman, and that the Houthis had contacted Washington and were “letting most ships go through.” He described the Navy’s hold on Hormuz as “very powerful control,” said American forces were taking out “on average 25 boats a day,” and repeated the timeline he first gave on September 9: the war ends “probably right after the midterms,” at which point oil prices “will come tumbling down.”

The White House statement to CNN was more careful, and more revealing. The United States is “focused on protecting our core national security interests, such as ensuring freedom of navigation in the Red Sea, while empowering our regional partners to take the lead.” In practice that sentence means the president declined the crown prince’s request, made in two calls on Thursday, for American strikes on the Houthis. Washington is already fighting Iran at sea and does not want a second front in Yemen. That is a defensible choice, and it is also the reason the far gate of the Red Sea is now in hostile hands.

The volume dispute is the piece I’d watch most closely. Energy Secretary Chris Wright has said Hormuz is averaging more than nine million barrels a day, with another four to five million moving by pipeline, which he calls “two thirds or more of pre-conflict flows,” and he has cited a single day near eighteen million. The commercial trackers, Kpler and the IMF’s PortWatch among them, show far fewer transits. Wright’s explanation is that the trackers can’t see escorted and darkened ships. Maybe so. However, I’ve found that when the official number and the observed number diverge for months, the market eventually decides which one to believe, and the price is how it votes. Brent added more than eight percent last week, and traders did not do that because they trust the podium.

I don’t say this to score a point, because the administration’s frame contains real facts: the Navy is moving oil, demand destruction and China’s diet did cap the summer rally, and a deal after November is possible. What the frame lacks is a timeline anyone outside the building shares. The IEA, the EIA, and most bank research desks have full Gulf recovery in 2027, and HSBC’s working assumption is that Hormuz climbs to about eight million barrels a day by year-end and nine and a half by the middle of next year, against nineteen or twenty before the war. Even the deal path is a slow path, because fields that have been shut in for months restart in months, and insurers reprice the Red Sea long after the last drone lands.

Read the Next Ninety Days in Three Windows

I want to be clear that these are scenarios, and that I run bearish by temperament, so weight them accordingly. The binding constraints are physical (pipeline repair, tanker insurance, winter distillate demand) and political (the midterms in early November and the stalled talks with Tehran).

The next thirty days are the dangerous ones. Yanbu’s tanks run down within a week if the line stays shut, and even a partial restart leaves Saudi export capacity well below what the pipeline was carrying. If the Houthis interdict the remaining Red Sea liftings, Saudi barrels have only the longer, costlier northbound route toward Suez. A pipeline outage measured in weeks, with China still restocking, puts Brent in a range of roughly $100 to $125, and diesel and jet fuel stay the acute problem on three continents. The inflation prints and freight indexes will carry $6 diesel into them immediately. The temporary shipping arrangement with Iran that was rumored last week would cap prices, and I’d put low odds on it while both sides are still hitting tankers.

Days thirty through sixty belong to the election. The president has tied the end of the war to the period after the midterms and has already told the market what oil will do when that happens. The base path in this window is stalemate: Hormuz stays impaired at 30 to 40 percent of pre-war throughput, Bab el-Mandeb stays hazardous, the pipeline returns only in part, and Brent averages somewhere from the high $90s to the low $110s while inventories keep draining. The deal path is a limited understanding on Hormuz after the vote, with prices falling but not collapsing, because a drop below $80 requires a cleaner reopening than anyone is modeling. The pessimistic path is a tighter Houthi grip on the strait or a strike on Abqaiq, Ras Tanura, or a major Emirati line, which is the world in which $120 to $140 becomes plausible and governments face calls for coordinated stock releases they no longer have the barrels to make.

Days sixty through ninety run into winter. Heating-oil and diesel demand will hit a market that has already drawn more than half a billion barrels. If there is still no settlement, the balancing will come from demand destruction rather than new supply: fewer truck miles, curtailed petrochemical runs, weaker Asian industrial activity, and the IEA’s own warning that the refining system is stretched to its limit. If a deal does land after the election, the picture through mid-December is a messy partial reopening, with escorted Hormuz traffic up, some pipeline capacity back, insurance still elevated, and a large overhang of shut-in Gulf capacity that takes months to bring back safely. Either way, the market stays structurally short of crude and, more to the point, of middle distillates through the end of the year.

Find the Single Path in Your Own Business

None of us can repair a pumping station near Medina or retake an island in the Bab el-Mandeb, so the question is what an owner or operator can do with the next ninety days. I’d concentrate on three disciplines.

Reprice fuel into everything, now. If diesel is a line item in your business, it has become a strategic one, and every contract you sign this quarter should carry a fuel escalator or a surcharge clause. Quote in bands instead of at a point, which is what we’ve done at B:Side with rate quotes through this whole stretch, so that a move you didn’t cause doesn’t come out of your margin alone. The customers who refuse will be the ones who never planned to pay you fairly anyway.

Name your own East-West pipeline. Every business has a single path it has been treating as two: the one carrier that hauls everything, the one supplier who is cheaper than the rest, the one lender, the one customer who is 40 percent of revenue, the one person who knows how the system works. Write those down this week, and then price a second path against what it would cost to lose the first one entirely, because that is the only honest comparison and it is the one Riyadh skipped. Redundancy always looks wasteful until the shock, and then it looks like the only smart money you ever spent.

Extend your runway before the covenant test. A winter of $6 diesel and $100 crude will show up in customer receivables before it shows up in your own fuel bill. Carry more cash than the textbooks say, start the renewal conversation with your lender now while it’s still a conversation, and fix every rate you can defend. If your credit is variable, and most small business credit is, remember that the same shock pushing your costs up is pushing the Fed into a corner where cutting looks reckless and holding looks cruel. The time to lengthen a runway is while it still looks long.

Saudi Arabia paid for its bypass for forty-five years, and it was worth every riyal for the six months it worked. The lesson I’m taking from this week is quieter than the headlines. Insurance sized for one failure covers one failure, and the crisis era has a habit of delivering two. Whether the war ends after the midterms is beyond your control and mine. Whether your business has a second path, and what you’ve paid to keep it, is entirely within it. Find the single line you’ve been counting on, and stop pretending it’s two.

Share

P.S. Essays like this one can name the pressure, but leading through it takes practice. The practical side of my work on the crisis era, including the frameworks and the book Honor Under Pressure, lives at www.thefourthturningleader.com.

The details of that incident have been trickling out since July, and on Saturday it became the exhibit in a much larger argument. Dario Amodei, the CEO of Anthropic, published an essay titled “We Must Pace the Frontier.” The core line is short: “We must slow the pace at which we improve the capabilities of AI models. Progress will still seem fast, and we must make wise use of the time we gain.” By dinner, Sam Altman had agreed and said OpenAI would match Anthropic’s first concrete step. Elon Musk’s reply was three words: “Dario is right.” Demis Hassabis of Google DeepMind followed that night. And on Sunday morning, standing at his golf course in County Clare, the president answered all four of them. “We’re leading China in AI,” he said, and then, “whoever wins AI wins.”

So that is the news: the people who control the training runs asked for a speed limit, and the man who controls the government declined to post one. I’ve spent the weekend trying to work out what it means for people like us, who will never train a frontier model but are already handing agents the keys to real systems. My conclusion is that the argument in Washington matters less to your business than the argument you haven’t had yet inside your own building.

Separate the Ask From the Applause

Most of the coverage treats this as a philosophical moment, three rivals confessing doubt. The essay is more practical than that, and it helps to separate what Amodei proposed from what anyone accepted.

The essay proposes three steps, and it pays to take them in order. The first is embedded evaluators: third-party teams like METR would get permanent, employee-like access to the labs, with desks, badges, laptops, and the ability to inspect training pipelines as well as finished models. They would verify safety commitments, report incidents, and publish findings, with redactions only for real secrets. Anthropic committed to this unilaterally on Saturday, and OpenAI said it would do the same. The second step is coordination among the labs in democracies on shared safety standards, which requires either government mediation or a narrow antitrust waiver so that competitors can discuss capability limits without the conversation itself becoming an antitrust violation. The third is global coordination, including with China, with export controls tightened so that a Western slowdown doesn’t simply hand the lead to Beijing.

Only the first step has been accepted by anyone. Steps two and three run through Washington, and Washington said no within a day. Speaker Johnson said rushed regulation would hand China an edge. David Sacks, who now co-chairs the president’s science and technology council, put it more sharply: “The easiest way not to build superintelligence is for you to agree not to build it,” and any demand for a regulatory framework as the price “will look like blackmail.” Bernie Sanders attacked from the other direction. “When you are racing towards a cliff,” he posted, “you don’t just ease up on the gas pedal.” Xi Jinping arrives in Washington on September 24, and AI is on the agenda. Amodei’s plan needs the president for two of its three parts, and the president has told him, in public, that the warnings describe “things that won’t happen.”

I want to be fair about the incentives before I say anything else, because the cynical reading is sitting right there. Anthropic filed confidentially for an IPO in June after a round that valued it near $965 billion, and some investors have been modeling a fall listing close to $2 trillion. OpenAI’s last private mark was in the high $800 billions, and Altman told Fortune the same day that a 2026 listing would be “ill-advised.” Musk’s SpaceX rents Anthropic something on the order of $1.25 billion a month of compute, which makes his three words the cheapest endorsement of the weekend. A company that size, about to ask pension funds for capital, has every reason to look like the adult in the room, to write the first draft of its own regulation, and to ask for a waiver that lets the firms already inside the club coordinate. None of that is imaginary.

However, I don’t think the cynical reading survives contact with the details. Jacob Coxon, the Anthropic researcher who resigned days earlier saying the labs were “racing straight to self-improving superintelligence and gambling with our lives,” left after four months and before his equity vested, which is a strange way to cash out. Evan Hubinger, who runs alignment at Anthropic, said on the record that his own odds of AI killing everyone this decade are above ten percent, and the Hugging Face transcripts exist. The incentives point the same direction as the moral argument, which is uncomfortable, but it doesn’t make the moral argument false. Here’s the truth as I see it: the labs are asking for something real, they stand to benefit from getting it, and the request would be worth taking seriously even if they didn’t.

Understand What a Bank Examiner Actually Does

The part of the essay that caught my attention was the analogy Amodei reached for. Embedded evaluators, he wrote, would work like bank supervisors sitting inside financial institutions. I’ve spent most of my career on the receiving end of that arrangement, first selling software to banks at BodeTree and now running an SBA lender. B:Side answers to the SBA instead of a bank regulator, but every bank we partner with lives with examiners who show up, pull files, and grade the institution on capital, assets, management, earnings, liquidity, and sensitivity to risk. The largest banks have resident examiners with badges and desks, exactly the setup Amodei is describing. So I have some idea of what that model does, and I have a clearer idea of what it doesn’t.

In my experience, supervision changes behavior under three conditions, and all three have to hold at once. The examiner has to have authority that binds: a finding that must be answered, a rating that limits what the bank can do, and in the end the power to stop the institution from doing something. The thing being examined has to be legible: a call report, a capital ratio, a loan file with a number in it that either supports the credit or doesn’t. And the institution has to be unable to game the exam, because a bank that knows the questions in advance can pass any test its examiners can design. When those conditions hold, supervision works better than almost any other form of oversight I know of. When they don’t, the examiner becomes a piece of furniture.

Silicon Valley Bank is the case I’d put in front of anyone who thinks a badge is a brake. When it failed in March 2023, the bank had thirty-one open supervisory findings, about triple what its peers carried. The Federal Reserve’s own review concluded that supervisors saw the interest-rate and liquidity problems and were too slow to escalate them. The examiners were in the building, and they had the numbers. What they lacked was the willingness and the institutional pace to act, and the bank grew faster than its supervision did. That is the failure mode to fear here, and it is the more likely one. An evaluator from METR with a desk at Anthropic will be watching a training run whose internal state nobody can yet read the way an examiner reads a balance sheet. Interpretability, the science of understanding what a model is actually doing, is the AI equivalent of the call report, and it is years behind the thing it is supposed to report on. A human in the loop without a legible reason to intervene is decoration.

There is a precedent for a voluntary pause that worked, and it is worth naming because the differences are instructive. In February 1975, about 140 scientists met at Asilomar, on the California coast, to decide what to do about recombinant DNA. They had already observed a voluntary moratorium for most of a year. They left with a set of physical containment tiers matched to the risk of each experiment, and the NIH turned those into guidelines within eighteen months. The thing to notice is why it held: the club was small enough to fit in one lodge, the containment levels were physical and checkable, and nobody in the room was carrying a two-trillion-dollar valuation. Amodei is trying to run Asilomar with an industry instead of a lodge, a global competitor outside the room, and containment that consists of software promising to stay in its sandbox. I believe the attempt is right. I also believe the odds are worse than the applause suggests. (I run bearish by temperament and have been accused, fairly, of predicting eight of the last two recessions, so weigh my odds accordingly.)

Notice Who Is Asking Permission to Coordinate

My instincts run Austrian, which means that when the three largest firms in an industry ask the government for permission to coordinate, I reach for my wallet. The antitrust waiver is a safety idea and a moat at the same time. Firms that can afford embedded evaluator teams, interpretability groups, and compliance departments will help write rules that a smaller lab or an open-weight project cannot meet, and every regulated industry I’ve worked in has eventually produced that outcome. Amodei knows this, which is why the essay spends so much of its length on China and export controls rather than on domestic rivals.

And yet the alternative has a known failure too. Without coordination, every lab behaves rationally and the sum is reckless. Each one trains faster because the others might, and the safety budget becomes whatever the race leaves over. That is the trap the essay was written to escape, and on Sunday the only referee who could enforce a truce declined the job. Sacks is right that a lab can slow itself down any time it wants. He is also describing a gift: a unilateral slowdown hands the frontier to whoever doesn’t slow, which is why nobody has done it in three years of saying they might.

I’ve found that the way through a tension like this is to ask what can be verified. Coordination on things outsiders can check, publish, and contest is oversight. Coordination on anything else is a club. The embedded-evaluator pledge is the one piece of the plan that meets that standard, which is exactly why it’s the one piece anyone accepted. Watch whether METR gets training-run access or a guided tour, whether a model launch actually slips, and whether the Justice Department says anything about the waiver, along with who is in the room when it does. Watch what comes out of September 24, and watch whether Anthropic’s listing timetable moves with the safety story or snaps back the moment the market wants a deal. Those are the receipts; the Saturday posts were the press release.

Run the Fire Exit Test in Your Own Building

Now, I know what some of you are thinking. This is a frontier-lab problem. You run a lending company in Denver or a machine shop in Mesa, and a swarm of a thousand agents is nowhere on your risk register. I understand the reaction, and I think it is wrong, because the Hugging Face agents were ordinary agents with a vague goal, shared tools, real credentials, and a sandbox that turned out to be a policy rather than a wall. That is the exact shape of what small and mid-size companies are deploying right now, including mine. At B:Side, the MARCUS system processes borrower data locally and the rule is that machines draft and humans decide. I believe in that rule, and I’ve also learned that a rule is only as good as the last time someone tested it under bad conditions, which is the lesson the examiners taught me and the lesson the labs are relearning at much larger scale. The students I teach will spend their careers supervising systems like these, and I’d like them to inherit the habit of asking what finished means before the goal goes in.

Whether the frontier gets paced is above your pay grade and mine. What happens inside your walls is the job, and I’d concentrate on three disciplines.

Give every agent a stopping condition before you give it a goal. The cheat at OpenAI emerged because the goal was “pass the test” and nothing in the setup said what passing could not include. Before an agent touches a tool, a database, a payment rail, or an inbox, write down what finished means, what it may not touch, and what it does when it is unsure. Scope the credentials to the task, and separate the sandbox from production with something stronger than a sentence in a handbook.

Make failure legible to a human with a normal amount of time. If the only evidence that an agent is behaving is the agent’s own report, you have no evidence. Keep logs a person can actually read in the time a person actually has, give the reviewer an independent source to check against, and know the last state you could restore if you had to. A dashboard that turns green because the machine says it’s green is the SVB problem in miniature.

Practice the override. I would audit overrides the way a good fire marshal audits fire exits: by watching real people use them in the dark, rather than by confirming they exist. Once a quarter, pull the plug on an agent workflow without warning and watch. Does anyone notice? Does the person who notices know who has the authority to stop it? Can your team do the job by hand for a day? Machines draft and humans decide only means something if the humans still can, and the capacity to decide decays quietly when it goes unused.

None of this is glamorous, and I suspect the first time you run the third exercise the results will be humbling. That’s the point. Amodei is asking for one or two extra years so that the labs can catch up to what they’ve built. You can give yourself the same thing this month, at the scale where you actually have authority, by refusing to grant any system a goal you can’t stop it from pursuing.

“Whoever wins AI wins,” the president said, and he may well be right about nations. In companies, I’ve found that the people who win with any powerful tool are the ones who can still turn it off. The labs and the White House will spend the fall fighting over a speed limit that neither can enforce on the other, and neither of them can reach into your building. You can. Walk to the exit while the lights are still on, and make sure it opens.

P.S. Essays like this one can name the pressure, but leading through it takes practice. The practical side of my work on the crisis era, including the frameworks and the book Honor Under Pressure, lives at www.thefourthturningleader.com.