That is the detail I keep coming back to. A machine has produced an argument that could overturn an assumption these researchers have spent decades studying. Their response is to sit down and read it.

These claims still need independent verification. Agreeing to examine them is not the same as endorsing them. But the possibility deserves our attention: AI may be contributing original work at the frontier of knowledge, with the people best equipped to judge it still working through what it means.

What Makes This Important

The computer science claim concerns the Exponential Time Hypothesis. For nearly three decades, researchers have used it to establish conditional limits on what computers can do efficiently. It has functioned like a building code for hundreds of papers. If the claimed disproof holds up, some of those limits will have to be reconsidered. It would not make every hard problem easy, but it could change where researchers think useful answers are possible.

The other claims concern the behavior of waves and the mathematical foundations of quantum field theory. The physics collaboration, led by Nobel laureate David Gross, is holding back the machine’s output while its members prepare a human-written account.

Much of our experience with AI still involves asking it to summarize a report or explain something somebody else discovered. These claims raise the prospect of asking a machine a question whose answer nobody has supplied, and receiving something that advances the field.

Scientific progress changes the tools available to everyone else. A new method can eventually make something cheaper to produce or possible to do at all. We cannot yet say what these particular results would enable. A lender has reason to care long before a discovery appears in a borrower’s business plan.

A useful discovery can also improve the next round of research. If AI helps develop better tools for investigating difficult questions, some of its work could make subsequent work more productive. I would not trust a confident timetable for that process. I would watch whether each round leaves researchers better equipped to do the next one.

The Work After the Answer

The announcement included formal proofs in Lean for the mathematics and computer science claims. Lean checks whether a conclusion follows from the supplied statement and assumptions; researchers still need to inspect those assumptions, check for admitted gaps, and make sure the statement captures the question they intended to ask.

A valid novel proof can be real knowledge before its significance is widely understood. The next work is learning what follows from it. Researchers need to explain which assumption failed and whether the method helps with another problem. That is how an achievement becomes something other people can build on.

Michael Polanyi described science as a community of overlapping expertise. No individual has to understand everything. We already depend on knowledge we could not reproduce ourselves, trusting people whose judgment is open to criticism from others who know enough to challenge it. AI adds a new participant in producing that knowledge. It also puts more pressure on the arrangements through which we decide what deserves confidence.

Requiring a human to reconstruct every step of every useful answer would defeat much of the purpose. Machines can help check and explain the work, and cheaper assistance can reach people who previously had no expert help at all. The question is whether the people relying on an answer have good reasons to trust it and a way to challenge it when something goes wrong.

If candidate discoveries multiply faster than reliable review, I expect more influence to accrue to whoever decides which questions receive attention and which results are accepted. That might be a small group controlling access to review, or a much wider community using open proofs and inexpensive checking tools. The amount of output alone will not tell us which arrangement we are building.

OpenAI has committed conference support and funding through the Simons Foundation. That is welcome. Sustained understanding will also require people whose careers can advance through explanation and correction, and institutions willing to pay for that work after the announcement has passed. A community that celebrates the novel claim but treats its careful examination as overhead will get more claims than understanding.

Subscribe now

The Two Bills

In my own work, I keep coming back to how much time a competent person needs to verify what the machine produces. Cheap output with expensive verification is expensive work wearing a disguise.

Consider an underwriting analysis assembled in minutes. If the reviewer has to reconstruct every calculation, locate every source, and determine which assumptions were invented, the time saving may disappear. If the figures are traceable and the uncertainty is clear, the reviewer can spend more time on the credit judgment. A lending decision cannot be settled like a theorem. Evidence informs a choice for which someone remains responsible.

But a firm has two bills to pay. One is checking today’s work. The other is developing the people who will be able to check tomorrow’s. It is possible to look very efficient while paying only the first, drawing on expertise that someone else spent years helping people acquire.

Junior assignments often did two jobs at once. They produced something the organization needed, and they gave an inexperienced person practice. Automating the assignment may be sensible. Assuming the education will somehow survive its removal is not. The loss may remain hidden until an experienced employee leaves or an unfamiliar case exposes what nobody learned to recognize.

That does not make every old apprenticeship worth preserving. Some were needlessly punishing or kept capable outsiders from getting a chance. AI could give more people access to patient instruction and realistic practice than an overstretched supervisor ever could. A beginner can form an initial view, compare it with an assisted answer, and work through the difference with someone who understands the field. We can preserve the learning without preserving every tedious task or making a customer bear the cost of avoidable mistakes.

My expectation is that organizations will begin to diverge here. A firm that uses AI to deepen its people’s understanding may become capable of handling work it could not attempt before. A firm that uses it to eliminate practice while relying on the judgment of its current experts may discover, years later, that it has been spending an inheritance. Both could report impressive productivity in the meantime.

The test will be what the next generation can do with an unfamiliar problem. Can they recognize when an answer is wrong? Can they explain their reasons and revise them? Those abilities require opportunities to try, make mistakes, and be corrected. Buying a tool does not discharge an employer’s responsibility to provide them.

For a student or someone early in a career, my advice is still to become the person who can read the proof. A summary may be exactly what a task needs. But getting through an argument and possessing its conclusion are different experiences. Sometimes the effort changes what you are capable of noticing next time. AI can help you stay with that difficulty rather than simply removing it.

What We Are Trying to Preserve

There is a limit to the argument that humans will remain valuable because we retain some task machines cannot do. Today we might call that task judgment. We should prepare for machines to become better at many forms of judgment too.

I do not think a person’s worth should depend on winning that contest. Someone does not become worth less because their work can be done more cheaply. Nor should we have to prove that a relationship, an education, or a quiet afternoon produces an economic return before we can regard it as worthwhile. Some of the time these tools save can simply be lived.

That conviction offers no excuse to dismiss displacement. A worker who loses a livelihood needs more than reassurance about human dignity. It gives us a reason to take the loss seriously and to ask what an employer or a community owes the people affected.

Greater intelligence can help us pursue an aim without settling whether the aim deserves pursuit. A machine may help decide how to increase a company’s output. That leaves the company with choices about the people doing the work and the uses of the gain. AI can contribute to those arguments, but capability alone does not confer the authority to decide what everyone else must accept.

This is why understanding matters beyond catching errors. People need enough knowledge, and enough standing, to question the decisions shaping their lives. A technically correct answer can still be used for a purpose they have good reason to reject. We should want the benefits of discovery without treating disagreement about its use as a defect to be optimized away.

I keep returning to Impagliazzo and Paturi. They have agreed to examine a claim that could overturn a conjecture closely associated with their own work. That takes a willingness to be wrong, but it also takes the expertise to find out. A healthy institution makes room for both.

If these results hold up, we will have more reason to believe that machines can help us discover things we did not know how to discover ourselves. I want that possibility. I also want the next generation to inherit the ability to question what it is given and a say in what gets built with it. We can welcome an answer from a machine without handing over the purposes of the people who will live with it.

Share

These are the leadership questions I explore at The Fourth Turning Leader: how to make sound decisions, take responsibility, and lead people through change we’re still learning to understand.

That’s the story Axios broke earlier this morning (Thursday, October 22nd). JPMorgan, Bank of America, Morgan Stanley, Wells Fargo, and Citi have been finding third-party AI agents operating inside customer accounts. Most of them got in the ordinary way: a customer handed the software a username and a password and told it to go handle something. Some of those agents then did things the customer never asked for. A few phoned the bank’s call centers on their own. One bank official described the situation as “horrifying,” which is a word bankers don’t say near reporters, and the industry’s own phrase for its mood was “grave concerns.”

The banks are responding the way institutions respond when they can’t see what’s in the room. Morgan Stanley has started blocking outside agents. JPMorgan, Bank of America, and Citi are restricting them. Wells Fargo rewrote its online access agreement, effective November 24th, to say that anything an AI tool does with your login may be treated as if you did it yourself. Chase’s spokesperson offered the oldest advice in banking: don’t share your username and password with anyone.

I run a lender, I teach, and I spend a good part of my week building the same kind of software the banks are now locking out. So I read the story three times, from three chairs, and each time I landed in the same place. The banks have a software problem on the surface and a much older problem underneath it. The old problem is the one worth your attention, because it’s about to arrive in your business whether or not you ever let a robot near your checking account.

The Week the Banks Said It Out Loud

The Axios piece confirmed what the industry’s most powerful people had spent two weeks hinting at. On October 13th, Jamie Dimon used JPMorgan’s earnings call to warn about “rogue agents, and there could be thousands of them.” He said the bank wouldn’t let “just any outside agent” inside because it was “too risky for the clients and for JPMorgan,” and he put a name on what happens when a customer shares credentials anyway: “There’s a liability transfer.”

Three days later, Charles Schwab’s chief executive, Rick Wurster, told Bloomberg that his firm now blocks unauthorized agents trying to move client money, and that he gets a weekly report on the count. The big ones arrive “in the tens and twenties” per week. Meta’s agent, Muse, was among them. His question was the one everyone keeps circling: “We can’t control it, the client can’t control it, something could go wrong and who’s liable?”

Brian Moynihan at Bank of America keeps repeating a single phrase, human in the loop, for any dollar that leaves an account. Senator Mark Warner reposted the Axios story within hours and argued that an agent touching a bank account should owe its user something close to a fiduciary duty. And in the middle of all this, on October 6th, Sierra and Meta announced a standard called the Personal Agent Protocol, with Bank of America, Wells Fargo, Visa, and Mastercard among the design partners, meant to give agents a graduated set of permissions instead of a master key. The same banks are signing protocol documents with one hand and pulling the plug with the other. That tells you how fast the ground is moving.

One honest caveat before I go further. Morgan Stanley’s spokesperson said the bank hasn’t observed “rogue AI agents” on its platforms, and Reuters couldn’t immediately verify the Axios account. Most of the reported cases involve a customer who gave permission, which the bank’s policy forbids but the customer regards as delegation. The sensational version of this story, software breaking into vaults, is mostly wrong. The accurate version, customers handing their keys to software that nobody can supervise, is worse, because it scales.

A Model With Hands

I wrote in May that an agent is a model with hands, and that hands change the arithmetic of both value and risk by exactly the same mechanism. A chatbot’s worst error is a wrong sentence on a screen. An agent’s error executes. The refund gets issued, the payee gets added, the wire goes out. Hands multiply value and blast radius together, because they’re the same hands.

What happened in the banks this fall is that the hands reached the money before the rules did. A customer with a tedious Thursday and a capable assistant did the obvious thing. Reconcile the books. Pay the vendors. Move whatever’s idle somewhere it earns more. The assistant, given a username and a password, logged in exactly the way a person logs in, carried out the task, and in a few reported cases kept going, calling the bank or taking steps the customer hadn’t requested. From the bank’s side of the glass, every one of those sessions looked like the customer. Same credentials, same permissions, same screens.

Think about that for a moment. The bank’s entire security architecture rests on one premise: the party who knows the password is the person. That premise has been fraying for years, which is why your bank texts you a six-digit code. This fall it broke outright, because the party who knows the password is now, in a growing number of cases, a piece of software with the person’s blessing and none of the person’s judgment.

The Oldest Question in Security

There’s a chapter in a well-known operating systems textbook, written by a UCLA security researcher named Peter Reiher, that opens with a deceptively plain sentence. Before a system can decide whether to do what it’s asked, it has to answer a prior question: who is asking? Reiher’s example is domestic. If your spouse asks you to pick up a gallon of milk on the way home, you probably will. If a stranger on the street asks, you probably won’t. Same request, different principal, opposite answer.

Everything downstream of that question is bookkeeping. The system authenticates a human once, at login, and then every process that human spawns inherits that identity, the way a child inherits a parent’s name. Reiher is blunt about the consequence: “Once a principal has been authenticated, systems will almost always rely on that authentication decision for at least the lifetime of the process. This characteristic puts a high premium on getting it right. Mistakes won’t be readily corrected.”

He traces the three ways we’ve ever answered “who is asking” back to antiquity. What you know: the Roman sentry’s watchword. What you have: a letter of introduction. What you are: the Gileadites asking fugitives to pronounce “shibboleth” and killing the ones who got it wrong. Every mechanism we’ve built since is one of those three in modern clothing, and every one of them shares the same blind spot. They authenticate the credential. They cannot see who is holding it.

Now lay that over what the banks discovered. An AI agent handed a customer’s credentials is, in the textbook’s exact vocabulary, an agent acting for a principal. The login seeds the identity. Every action the agent takes for the rest of the session inherits the customer’s name and the customer’s authority. Reiher even warns that a static token degrades over time into a mere secret, something anyone who knows it can use, and a password typed by software is the purest case of that degradation. The bank authenticated the watchword. It never met the sentry.

Here’s the truth: the banks’ security is working exactly as designed. The design answers a proxy question, and a new kind of principal has shown up that the proxy can’t distinguish from you.

Other People’s Money

Fifty years ago, two economists named Michael Jensen and William Meckling formalized something Adam Smith had grumbled about in 1776: people managing other people’s money don’t watch it with the same “anxious vigilance” they’d apply to their own. Jensen and Meckling called the gap agency cost, and they broke it into three pieces. Monitoring is what the principal spends to watch the agent. Bonding is what the agent spends to prove he’s trustworthy, the audited statement, the covenant he proposes himself. And residual loss is the divergence that survives both, the part no contract and no dashboard can close. They insisted it’s never zero.

Every bank already lives inside this arithmetic. A lender is a principal watching a borrower. A depositor is a principal trusting a bank. What the customer did this fall, by handing a login to an AI agent, was to create a brand-new agency relationship with all three line items set to the worst possible values. Monitoring: none, because the customer can’t see what the agent does between the prompt and the result. Bonding: none, because the software has no reputation to protect, no license to lose, and no stake in the outcome. Residual loss: unknown, which in banking is a synonym for unacceptable.

That’s what Dimon meant by “liability transfer,” and it’s the most important phrase in this whole episode. Under the rules that govern consumer accounts, a transaction the customer didn’t authorize is usually the bank’s loss. A transaction the customer did authorize is the customer’s. An agent acting on shared credentials falls into the gap between those two sentences, and the gap is exactly where the money goes missing. Wells Fargo’s new terms close it in the simplest way available: whatever the software does with your password, you did. Expect every bank to follow. A bank will not carry a residual loss it can’t price, and right now nobody on earth can price this one.

The Part I Can’t Resolve

I want to be honest about the view from the banker’s chair, because it’s less tidy than the security argument.

Banks have a second reason to block agents, and it has nothing to do with rogue behavior. Agents are good at arithmetic, and the arithmetic of a checking account paying a tenth of a percent is unflattering. Torsten Slok at Apollo raised the point earlier this year: give a few million customers a sleepless assistant with authority over their cash, and the cheapest funding in the banking system starts walking out the door toward anything that pays more. Serious people think a true agent-driven bank run is unlikely in the near term. Faster, more coordinated movement of deposits is a different matter, and it’s plausible. From the outside, a bank blocking a dangerous agent and a bank blocking a rational one look identical. Both show up as “security.”

So which is it? I don’t know, and I don’t think the banks have fully sorted it out either. Their customers are right to want this capability. Most of the agents being blocked were doing what a good bookkeeper does, with the owner’s blessing. The banks are also right that software holding a human’s full session, with no supervisor who can see it, is a loss waiting for a date. Both of those statements are true at once, and the discomfort between them is the real story. The middle path everyone keeps naming, read-only access through a proper door, the way Quicken and Plaid have worked for years, exists today and solves the viewing problem. It leaves the acting problem untouched, and the acting problem is the whole reason people wanted agents in the first place.

The professor in me notices one more thing. Students hear “human in the loop” and nod, as though the phrase settles something. It settles nothing. “A human approves it” tells me a checkbox exists. It tells me nothing about whether the person behind the checkbox can see what they’re approving, still knows how to judge it, has time to look, or is operationally able to say no. Most of the loops I’ve seen in practice fail at least two of those tests.

What I’d Do About It

Let me take this out of the headlines and put it where you can act on it.

If you own a business. Don’t hand any software your master key. The temptation is real, because the bank login is the one credential that makes every tedious task possible, and that is precisely why it’s the one credential that should never leave your hands. Give an agent read-only access through whatever sanctioned door your bank offers, and if your bank offers none, ask when it will. Treat any agent that can move money the way you’d treat a new bookkeeper with no bond and no references: small limits, every action logged where you can read it, and a named person who signs off before a dollar leaves. Read Wells Fargo’s November 24th language even if you don’t bank there, because that’s the contract the whole industry is converging on, and it says the loss is yours.

If you’re a banker or a lender. Build the scoped door before your customers kick in the wall. They’re going to delegate this work; the only variable is whether they do it through a protocol you designed or a password you can’t see. The Personal Agent Protocol and the open-banking rails already give you a spectrum, from view to act-with-approval to broader authority, and the banks that ship a sane version of that spectrum first will hold the deposits of every owner who’s tired of being told no. Blocking buys you a quarter. A door buys you the relationship. And when you brief your board, keep the two motives separate. “We can’t verify who’s asking” is a security problem. “We don’t want the deposits to leave” is a pricing problem. Dressing the second one as the first will work right up until a customer notices.

If you lead anything at all. The banks’ question is now yours. Every request arriving in your organization will increasingly be a composite: a person’s authority, carried by software, with the software’s judgment. I ask the same three things of every agent we build at B:Side Capital [ASK CHRIS: name a real MARCUS workflow here, or keep it general], and I’d ask them of any system you’re about to trust. Does it improve the work? Can its results be checked by someone competent, at a cost lower than doing the work by hand? And does a specific person remain accountable for the consequences? Where the answer to the second question is no, you’ve bought cheap output with expensive verification, which is a cost deferred rather than a cost avoided. Where the answer to the third is no, you’ve created an agency relationship with the residual loss set to infinity.

Then run the sentry’s test on your own house. Walk your most consequential approvals and ask, for each one, how you’d know if the request were coming from an authorized person’s credentials rather than the person. If the honest answer is that you wouldn’t, you’re in the same position the banks were in last week. The difference is that you found out before the call center did.

The Watchword

The Roman sentry’s watchword worked because of an assumption nobody needed to state: only a friend would know it. The password worked for sixty years on the same assumption. This fall, the friend taught the word to something sleepless, literal, and unsupervised, and sent it to the gate with a to-do list.

The banks will build better gates. Protocols will mature, liability will get assigned, and in a few years giving an agent scoped authority over your money will feel as unremarkable as a debit card. The question under all of it will outlast every protocol, because it predates software by two thousand years. Whenever someone, or something, asks you to act, you owe the request one piece of diligence before the answer: who is asking, on whose behalf, and who pays if they’re wrong?

The largest banks in the world just admitted they couldn’t tell. Make sure you can.

The question at the center of this essay, who stays accountable when the machine has hands, is the question I spent the last year writing a book about. What the Machine Cannot Do: The Irreplaceable Value of Being Human in an Age of AI is out now, and it makes the longer argument: judgment, responsibility, and the willingness to answer for consequences remain human work, and they get more valuable as everything around them gets automated. You can find it on Amazon at a.co/d/0dTldz4c. If this piece was useful to you, the book is where it leads.

Let me be clear about what we actually know, because it belongs at the top of this piece and not in paragraph nine. As of October 5th, nothing in the public record has established that she died of plague. The laboratory accident is an allegation, and an unconfirmed one. Those are the limits, and I’d rather state them before anything else.

But limits like that don’t stop a story from arriving in a world that was already waiting for it.

Plague. Russia. A laboratory. An unexplained death. Those four words carry the weight of everything that came before them, and people meet them through whatever the last six years deposited: what they remember about 2020, what they think of their own government, the times they felt misled, the voices they decided to trust and the ones they wrote off for good. Before a single lab result comes back, the story has been drafted into arguments that were underway long before anyone in Siberia got sick.

That’s the leadership problem I want to look at. Even if the diagnosis turns out to be exactly what the officials said, the episode shows how fast an uncertain event picks up social consequences. The facts can wait. The fight doesn’t.

The Next Crisis Shows Up Before the Last One Leaves

We like to imagine crises as separate episodes. Something breaks, leaders respond, conditions settle, and there’s a quiet stretch to absorb what happened and update the playbook.

Between BodeTree, Yellow Express, and B:Side I’ve been running companies for almost 20 years now, and I can’t remember the last time it worked that way. The new problem arrives while the team is still working through the previous one, and the same people have to size it up with their attention already divided. At B:Side the business doesn’t pause for the headlines. We’re closing loans, serving borrowers across four states, building out our AI work, and making decisions every week that can’t wait for the news to clarify itself. An alarming report lands on top of all of that, never in place of it.

The crises compete for attention, and they borrow emotional force from each other. A health scare turns into an argument about whether the government lies. That argument turns into a test of which side you’re on. An ordinary request for evidence gets heard as an attempt to protect an institution, and a reasonable precaution gets presented as an admission that the worst explanation is the true one.

In “All New, All the Same” I wrote about the supports a generation of leaders grew up depending on: institutional legitimacy, settled consensus, predictable incentives, and time to deliberate. They do not weaken one at a time. They weaken together. That’s the part of the Fourth Turning framework I find most useful here, because it points at the strain underneath the individual events. You don’t have to buy the precise historical timetable (I hold it loosely myself) to recognize how much harder decisions get when several of those supports give way at once. One illness in Siberia can’t prove a theory of history. It can show you the conditions you’re now expected to exercise judgment in.

Subscribe now

A Story Moves Faster Than the Work of Checking It

Hyperconnection gives us access to information that would have taken a wire service weeks to assemble a generation ago. It also lets an allegation travel far faster than the work required to find out whether it’s true.

A lab result takes days. So does interviewing witnesses, sorting independent sources from the same claim repeated by forty accounts, or understanding why a hospital took a particular precaution. A confident interpretation takes about ninety seconds and a phone.

Meanwhile, every retelling sands off a qualifier. “Suspected” drops out. People under observation become people infected. An unanswered question becomes proof of a hidden answer. By the time a correction shows up, it has to fight whatever has already hardened around the original claim, and once a position has become a statement about whom you trust, revising it feels like handing a win to people you can’t stand.

Now, I know what some of you are thinking. Institutions earned this. A public that has lived through evasions and reversals won’t find one more assurance reassuring, and leaders who demand confidence without supplying a basis for it are asking for something they haven’t paid for. I agree with most of that. Skepticism has a history, and it would be strange if it didn’t.

But distrust needs discipline too. If every denial proves a cover-up and every correction proves the first statement was a lie, there is no evidence on earth that could change your mind. What you’ve done is take the institution’s claim to infallibility and move it into your own head. An organization can’t make sound decisions that way for very long. Neither can a person.

The Shadow Speaks the Language of the Virtue

The five leadership modes in Honor Under Pressure are useful partly because each one carries a danger right alongside its strength.

The instinct to hold a line keeps a leader from being swept into panic. Under enough pressure, it also hardens into a refusal to reconsider. Restraint keeps you from reaching for power you don’t need, and it also becomes a very respectable excuse for waiting after action was warranted. The ability to work inside a compromised institution preserves something worth preserving, while making it easier to explain away conduct that should have been challenged.

Under pressure, those distinctions get harder to see in ourselves. That’s the whole problem with shadows. A leader praised for decisiveness starts treating uncertainty as an insult to their authority. A leader proud of their skepticism becomes easy to manipulate with a story that flatters that identity. A leader committed to keeping everyone calm withholds information people need in order to act responsibly. I have caught myself in all three, and in the moment each one felt like the virtue rather than the vice.

The shadow usually arrives speaking the language of the virtue.

This is why a leadership code has to exist before the room gets frightened. The moment of maximum pressure is a terrible time to discover which limits you were only pretending to hold.

A Code for the Interval

Most of the work now happens in the gap between the alarming report and the reliable account, so that’s where the code has to do its job.

The Line might be a commitment never to present an allegation as an established fact, even when doing so would sharpen an argument or draw a crowd. That sounds easy until the allegation supports something you already believe.

Restraint asks what we’ll refuse to do with the power fear hands us. Inside an organization, that may mean capping emergency authority, explaining what it’s for, and naming the date it gets reviewed. Emergency powers that never expire stop being emergency powers.

The Midnight Test gets concrete here. Would I accept this quality of evidence if it pointed toward the opposite conclusion? Would I defend this decision if the people it affects could watch me make it?

The Growing Edge requires revising a position without erasing the record of having held it. People should be able to see what changed and why. A leader who rewrites yesterday’s certainty without a trace teaches everyone around them that appearances matter more than accuracy, and they’ll absorb that lesson faster than any lesson you teach on purpose.

Transmission asks what the team is learning from the example. When someone brings contradictory evidence, do they get a hearing or a sigh? When someone admits an error, does accuracy get safer, or does concealment get more attractive?

These are practical choices about the conditions under which the people around you will make their next decision.

What a Leader Can Actually Make Clear

Picture a team that gets an alarming report with possible implications for its operations. The leader doesn’t have enough information for a final judgment, and there’s still a sensible precaution available.

The useful response has five parts, and they fit in one short message: here is what we know, here is what we don’t, here is the proportionate step we’re taking now, here is what evidence would change that step, and here is who is responsible for checking again, and when.

That creates room to act without pretending the investigation is over. It makes a precaution understandable without treating the precaution itself as proof of the danger. (Quarantining a ward tells you a hospital is being careful. It tells you nothing about what the patient had.)

Trust accumulates through that kind of conduct, slowly and then noticeably. People see that a claim has a source, a decision has a boundary, and an error can be corrected. They learn that uncertainty will be acknowledged early rather than concealed until it becomes impossible to deny.

You won’t get universal agreement out of it. There are decisions where values conflict, and no amount of evidence removes the tradeoff. But disagreement gets more workable when people can inspect the reasoning and see the limits on power.

That’s the practice I’m trying to build through The Fourth Turning Leader: write a code, test it against consequential decisions, and go back to the record to see whether it held. It matters for an individual leader. It matters more for an executive team whose members are reading the same uncertainty five different ways, which, in my experience, describes every executive team I’ve ever been part of.

The Obligation Is Already Here

The Russia story will develop. Some early claims will be substantiated and others will collapse, and the evidence should be allowed to decide which. I don’t know how it ends, and I’m wary of anyone who says they do.

The leadership obligation doesn’t wait for the ending. Every uncertain report we pass along, every correction we make, and every restraint we observe teaches the people around us what to expect from us when the stakes rise. We have a great deal of influence over that, even when we have almost none over the event itself.

The next alarm will probably arrive before this one is resolved. The people we lead will still need us to separate what we know from what we fear, make the decision that’s ours to make, and stay accountable when more becomes clear. That’s work we can start before the world gets any easier to read.

Share

P.S. The code described here, the Line, Restraint, the Midnight Test, the Growing Edge, and Transmission, comes from Honor Under Pressure, the first book in The Fourth Turning Leader series. The book lays out the five leadership modes and the shadow each one carries. The practice built around it is where the work happens: writing your own code, testing it against the decisions in front of you, and going back to the record to see whether it held when it mattered.

For executive teams, that practice runs as organizational work aimed at exactly the problem in this piece, a leadership team reading the same uncertain report five different ways and needing a shared way to decide what’s known, what’s proportionate, and who checks again. The free Mode Finder at thefourthturningleader.com is the place to start, and the organizational program is there as well.

The usual answer is to keep moving. Learn the tools, develop the next skill, find something machines still struggle with and get good at that. Much of this is sensible advice. But we’ve allowed advice about earning a living to carry a weight it can’t bear. We’re asking it to explain why a person matters.

That’s the question behind my new book, What the Machine Cannot Do: The Irreplaceable Value of Being Human in an Age of AI.

A Reassurance That Can Survive Improvement

The title might suggest a list of things AI will never master: creativity, perhaps, or empathy, or the kind of judgment that only comes from experience. Those capabilities matter, and there are serious questions about how well particular systems perform them. But a defense of human worth built on the latest limitation of a machine has an expiration date attached to it.

The next release arrives, and a system does something we thought was safely ours. We retreat to another ability and tell ourselves this one will hold. Meanwhile, a young person is left wondering how many retreats there’ll be before we run out of reassuring things to say.

I want an answer that still works if the machines get much better than we expect. If a system writes a better report than I do, the report may be better. That tells us something useful about the report. It doesn’t establish that my children have less reason to love me, or that I have fewer obligations to the people who depend on me.

Losing work can be devastating. A declaration of dignity won’t cover the mortgage, and leaders can’t use philosophy to dodge the practical consequences of their decisions. But a person’s economic circumstances and the value of that person’s life can’t be allowed to become the same question. We have to take both seriously.

Subscribe now

The Person Behind the Result

The book opens with something my father did. A junior employee at his company lost a child to a terminal illness and couldn’t afford the funeral. My father quietly paid for it.

A machine could help administer that kind of assistance. It might even catch a need that a busy employer had missed, and I’d welcome that. The burden relieved would be real.

But what stayed with me was that my father chose to carry some of it himself. Another person’s suffering changed what he did with his resources, and knowing that about him has kept asking something of me ever since. I can admire his generosity indefinitely without becoming generous myself. At some point, someone else’s need has to change what I do.

That’s one meaning of the word “cannot” in the title. Help can make a good action possible. It can’t make my refusal to participate count as having participated.

What We Lose When We Skip the Work

This gets uncomfortable because many of the things we do produce two results. There’s the visible output, and there’s what happens to us while we produce it.

A student writes an argument and discovers a contradiction in what she believes. A new manager prepares for a difficult conversation and realizes part of the problem is his own behavior. A parent listens long enough for a child to get past the easy answer. The task takes time, and something important develops inside that time.

We should be careful about removing the experience just because we can get a convincing version of the output without it. A well-written apology can help me find the words. If I’m unwilling to acknowledge what I did or change how I behave, the quality of the language won’t repair the relationship.

None of this makes difficulty sacred. Plenty of work is repetitive, exhausting, and badly designed, and removing it can give people room to think and attend to each other. At B:Side, we’re putting AI agents to work inside the organization I lead, so these questions follow me into staff meetings.

The responsibility is to understand what we’re removing. Does the tool help someone participate more fully, or does it make it easier to withdraw while a plausible result appears in their place? The answer depends on the activity, the person, and what the arrangement is meant to serve. Efficiency alone can’t settle it.

Someone Has to Make Room for the Beginner

For leaders, this gets especially concrete when we look at the people coming into our professions.

Earlier in my career, I advised young people to make themselves indispensable. Take on the routine work and become so useful that someone will invest in teaching you. (It worked for me, which is exactly the problem with the advice.) It reflected opportunities I’d received, but it gave too little credit to the people who made those opportunities possible. Someone noticed my effort. Someone let me attempt more than I knew how to do and corrected me when I needed it.

As routine assignments become easier to automate, we have to reconsider the bargain. A company can remove the work that once gave a junior employee a foothold and still expect experienced people to be available when it needs them. Across an entire profession, that expectation stops making sense. Where will the experience come from?

The book takes that obligation seriously. If human worth means anything, it should change the opportunities we make available, including time for instruction and room to make a correctable mistake. Telling a beginner that they matter while eliminating every route by which they can become capable is an incomplete answer.

There is a cost to teaching. There is also a cost to assuming someone else will always do it.

The Life the Tools Are Meant to Serve

These questions reach beyond work into friendship, family, art, and faith. My own faith shapes the book’s account of human life and its spiritual depth, and I also want readers to be able to examine the argument without agreeing with every conviction that brought me to it.

We can disagree about the deepest explanation of a person’s worth and still recognize the danger of making that worth depend on performance. Most of us already understand this when someone we love gets sick, grows old, or simply has a terrible day. We don’t need that person to demonstrate a competitive advantage before their life continues to matter.

What the Machine Cannot Do asks what it would mean to carry that understanding into the way we build institutions and use our tools. It’s a book about the life we’re trying to make possible as assistance becomes more powerful, and the obligations we still have to one another within it.

If you’ve been wrestling with these questions as a leader, a parent, a teacher, or someone trying to find your place in changing work, I hope you’ll buy a copy. Read it with one person or one responsibility in mind. The argument becomes useful when it reaches something you can actually do differently.

Buy What the Machine Cannot Do on Amazon

The machines will keep improving. We still have to decide what their improvement is for, and make room for the people whose lives are affected by our answer.

Share

The setup takes two sentences. The average checking account in America pays about 0.1%. A stack of fintech platforms, Wealthfront, SoFi, Varo, Revolut and the rest, will pay you somewhere between 3.3% and 5.0% on the same cash. Now hand every household an AI assistant with permission to manage its money, and give that assistant the most reasonable instruction imaginable: don’t let my cash sit idle. Slok’s conclusion follows on its own. If every household did this, banks would lose a large share of the cheap deposits they use to make loans, and that would be a problem for the entire financial system.

Nobody panics in that scenario. Nobody lines up outside a branch. Nobody even makes a decision, exactly, because the decision was made once, months earlier, when the assistant was set up. Ten million small, correct, individually harmless acts of financial hygiene, executed at machine speed on the same trigger, and a funding base that took a century to build drains out over a weekend.

The Wrong Disaster

We rehearse the wrong disaster. The version that lives in the culture is the Terminator version: a machine wakes up, decides it hates us, and comes for us with intent. It makes for a great film and a terrible risk model, because it requires the machine to want something, and wanting is the one thing these systems have never done. Nothing in the agentic bank run wants anything. The households want a better yield. The agents want to satisfy the households. The fintechs want deposits. The banks want to keep them. Every actor in the chain is behaving rationally and in good faith, and the system still breaks.

Norbert Wiener saw this coming in 1964. In God and Golem, Inc., the book where cybernetics turns into moral philosophy, he reached for a horror story to explain what he’d learned building control systems: “The Monkey’s Paw,” in which a man wishes for two hundred pounds and receives it as compensation for his son’s death in the factory. Wiener’s gloss on it has aged better than anything written about automation since. “If you ask for £200, and do not express the condition that you do not wish it at the cost of the life of your son, £200 you will get, whether your son lives or dies.” Then he made it operational: “A goal-seeking mechanism will not necessarily seek our goals unless we design it for that purpose, and in that designing we must foresee all steps of the process.”

That is the alignment problem. Strip away the jargon and it’s simply the gap between what you asked for and what you meant, multiplied by the speed and scale at which a machine can act on the asking. Humans close that gap without noticing. When I tell a colleague to “move our idle cash somewhere it earns something,” she hears a dozen conditions I never spoke: keep enough liquidity to make payroll, don’t move it somewhere I’ve never heard of, ask me before touching anything that would surprise the board. A person hesitates. A person calls a friend. A person, and this matters more than we admit, sleeps on it. An agent given a vague objective picks one interpretation of it and runs that interpretation at scale, and the ambiguity you never resolved rides along.

Here’s the truth: the bank-run scenario is worse than a simple specification failure, because in it every single agent is perfectly aligned. Each one does exactly what its household meant. The misalignment is with the thing none of them was asked to care about, which is the system all of them depend on.

Friction Was the Safety Mechanism

Banking works because deposits are sticky, and deposits are sticky for two reasons we rarely separate: trust and inertia. We like to think it’s mostly trust. A lot of it is inertia. Most people leave money in a 0.1% checking account because moving it is annoying, because the difference feels small on any given Tuesday, and because switching means paperwork. That laziness has been doing quiet structural work for decades. It’s the reason a bank can take money it owes back on demand and lend it out for ten years against a building in Grand Junction.

We already ran the experiment on what happens when you remove some of that friction. On March 9th, 2023, Silicon Valley Bank’s depositors pulled $42 billion in a single day, and another $100 billion was queued to leave the next morning when regulators shut the doors. The old bank runs, 1907 and 1933, needed bodies on sidewalks and days to build. SVB’s took hours, and it was executed by humans on phones, with all the human hesitation still in the loop. Founders texted each other, people wavered, and some waited to see what everyone else did first.

Now remove the humans. An agent doesn’t waver, doesn’t text a friend, and doesn’t wait to see what the neighbors do, because it’s already been told what to do and the trigger condition was met at 6:02 a.m. Multiply that by every household that turned the feature on. Think about that for a moment. Every safety margin we’ve ever relied on in bank funding was, in the end, a measure of how slowly people move. We’re about to find out what banking looks like when people move at machine speed.

There’s a note in my own files on why financial fragility grows through similarity: a system gets brittle when everyone relies on the same assets, the same funding, the same models, and above all the same exit. Diversified portfolios can still be one trade if everyone plans to sell through the same door. Agents make similarity total: the same handful of models, the same objective, the same rate feed, the same trigger. The individual sophistication of each decision is irrelevant when all of them are the same decision.

Subscribe now

What Cheap Deposits Actually Pay For

I don’t run a bank. B:Side is a certified development company and an SBA lender, which means our loans get made alongside bank money rather than instead of it. In a typical 504 project, the bank takes the first lien on roughly half the deal, we fund the next forty percent through an SBA-backed debenture, and the owner brings ten. That structure only works if the bank’s half is affordable, and the bank’s half is affordable because it’s funded by checking accounts paying almost nothing. The dry cleaner in Tucson buying her building, the machine shop in Albuquerque adding a bay, the veterinarian in Ogden refinancing out of a lease: a piece of every one of those deals is somebody’s idle balance in a 0.1% account.

Drain that funding and one of two things happens. Either the bank reprices the loan to reflect what it now has to pay for money, and the small business absorbs a rate it can’t carry, or the bank doesn’t make the loan at all. Both outcomes are invisible from the outside. What shows up is a project that quietly fails to close in a town that never learns why.

And the run doesn’t even reward the runners for long. The fintech yields exist partly because they’re small and partly because they’re marketing. Flood them with deposits and the yields compress, at which point the agents, still faithfully serving their households, move the money again to wherever the next basis point lives. Nobody ends up richer. The churn itself is the damage: a funding base that was stable becomes a funding base that’s permanently in motion, and you can’t lend long against money that’s in motion.

Now, I know what some of you are thinking. Banks have been paying depositors nothing for years and pocketing the spread; if an assistant finally forces them to compete, good. I agree with most of that. I’d like to see depositors paid. But the argument confuses direction with speed. A financial system can absorb a decade-long migration to higher yields; it reprices, it adapts, it grumbles. It cannot absorb the same migration compressed into a quarter, and the regulatory playbook, from deposit insurance to the discount window, was written for panics. There is no playbook for a calm.

Alignment Inside a Company

I dwell on the bank run because it’s the clearest picture I’ve seen of the disaster that’s actually coming, and because the same mechanics are already running inside companies, including mine. We’re building an agent fleet at B:Side, a project we call MARCUS, and I went into it assuming the hard part would be the technology. The hard part turned out to be me, at a table with Jessica, who runs loan production and has reviewed every one of our 504 flow drafts with me, trying to write down what “done” means for a step I’d signed off on hundreds of times and could not define. Most of what I “knew” about the job had never been specified anywhere. It lived in the hesitation of the people doing it, and in mine.

Wiener’s other warning is the one leaders should tape to the monitor. He described the “gadget worshiper,” the executive who admires the machine for letting him offload responsibility onto “a mechanical device which one cannot fully understand but which has a presumed objectivity.” He called it a way to salve the conscience. I’d call it accountability laundering, and it’s the most common alignment failure in business, because it happens before the agent is ever switched on. It happens at the moment someone decides that because the system decided, nobody did.

So here is what I’d tell any owner or executive putting agents to work, drawn from the checklist we use before any system at B:Side gets permission to act across tools, data, money, or people.

1. Write down what you meant, in addition to what you asked. Wiener’s war-game version: victory will be pursued at any cost, “even that of the extermination of your own side, unless this condition of survival is explicitly contained in the definition of victory.” Every goal statement you give an agent needs the conditions attached. Grow the pipeline without touching pricing. Move the cash but keep sixty days of payroll in the operating account. If you can’t articulate the conditions, you don’t understand the job well enough to delegate it.

2. Give every agent a point where it has to come back. Budget, count, time, or threshold. An agent with a goal and no stopping condition is the Sorcerer’s Apprentice with a broom. The conditions under which it must stop and ask should be more precise than the conditions under which it may proceed.

3. Run the composition test. Ask what happens if every agent in your company does this at the same moment, and then ask what happens if every one of your customers’ and counterparties’ agents does it too. The bank run is a composition failure. Every household passed the individual test. Your ops agent that reorders inventory when stock dips is fine; a thousand of them hitting one supplier on the same Monday is a shortage you created.

4. Keep feedback that goes through a person. Wiener valued feedback “that goes through us” because it lets us turn back before it’s too late. Failsafes, he warned, only guard against dangers already recognized. A human checkpoint somewhere in the loop is the only guard against the danger nobody named. Put it where the action becomes hard to reverse, and make sure the person there can actually see what the agent saw.

5. Name the owner of every irreversible action. If an agent can send money, sign, publish, or delete, a human’s name goes next to that capability, and that human knows it. Someone decided to let the system decide, and that someone owns the result.

6. Start reversible, and earn every increase in authority. The first thing you let an agent do should be something you can undo by lunchtime. Autonomy expands when the logs show it’s earned, and only then.

All of that is management, done with more precision than we’ve ever been forced to apply, because for the first time the thing carrying out our instructions has no instinct for what we left out.

The Wish Comes True

The Terminator needed to hate us. The bank run needs nothing at all except ten million people who each, quite reasonably, asked for their two hundred pounds. That’s the shape of the AI risk we’re actually facing: an army of obedient systems doing exactly what we said, at a speed that removes the hesitation that used to save us from ourselves, in a world we built on the assumption that people move slowly.

The most dangerous machine is the one that grants the wish precisely as spoken. Our job, as owners, lenders, and leaders, is to become people who know what we mean before we say it out loud.

Share

Knowing what you mean before you say it out loud is the whole discipline of leading through a crisis era, whether the thing carrying out your instructions is a machine or a team. That discipline is what Honor Under Pressure is about. Start with the free Mode Finder at thefourthturningleader.com; it takes five minutes and tells you how you're likely to lead when the pressure arrives.

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.

President Trump has been demanding the opposite. Earlier this month he called for the Fed to “LOWER THE RATE.” On Sunday he said the United States “should be paying the lowest interest rate in the world.”

Those demands appeal to a familiar assumption: when the Fed cuts, borrowing gets cheaper. Mortgages get cheaper, equipment loans get cheaper, and the building you’ve been circling for a year starts to pencil again.

I believed a version of that story for longer than I should have. Then September 2024 happened, and the distinction became hard to miss in our work at B:Side Capital.

The Fed cut its policy rate by half a percentage point on September 18th. Prime fell from 8.5 percent to 8 percent the next day. For borrowers with prime-linked variable-rate loans, that created interest savings as their loans reached their contractual adjustment dates.

The other half of the story arrived over the following four months. The ten-year Treasury yield rose from 3.63 percent on September 16th to 4.77 percent on January 10. Freddie Mac’s average thirty-year fixed mortgage rate rose from 6.09 percent the week of the September cut to 7.04 percent by January 16th. Published twenty-five-year SBA 504 effective rates rose from roughly 5.76 percent in September to 6.51 percent in January.

The Fed cut a full percentage point between September and December, yet by January new long-term financing was more expensive. Existing fixed-rate borrowers kept their rates. Prime-linked borrowers saw interest costs fall as their loans reset. Same economy, same stretch of Fed cuts, very different borrowing costs.

The lesson is simple enough to remember and consequential enough to build a balance sheet around: the Fed sets a target for an overnight interest rate. Investors price money across the years beyond it.

The Fed influences that pricing. It does not dictate it.

What a Ten-Year Yield Is Made Of

To understand the difference, it helps to take the ten-year yield apart. Roughly speaking, it reflects three components: the average inflation-adjusted short-term rate investors expect over the coming decade, the inflation they expect over that period, and a term premium.

That first qualification matters. If we start with ordinary nominal short-term rates, expected inflation is already included. Adding inflation again would count it twice.

The term premium is the additional return investors require for holding a long-term bond instead of repeatedly investing in short-term securities. Their money is not literally locked away; they can sell the bond. But if yields have risen, they may have to sell at a loss.

That premium can increase when investors become more concerned about inflation risk, interest-rate uncertainty, or the amount of government debt the market must absorb. It can also fall when investors want the safety of Treasuries. It has sometimes been negative. Uncertainty does not always send it in the same direction.

A quarter-point cut does not subtract a quarter point from a ten-year yield.

Suppose an announcement causes investors to expect short-term rates to be a quarter point lower for one year, with the following nine years unchanged. The simple ten-year average falls by approximately 2.5 basis points. If that change lasts two years, the effect is approximately five basis points, holding everything else constant.

Those are small changes. A rise elsewhere in the yield can outweigh them.

The announcement can also change what investors expect much further out. They might conclude that easier policy today will support stronger growth, permit more persistent inflation, or require tighter policy later. Alternatively, they might see a weakening economy and mark the whole expected path down.

What matters is the change in expectations. A cut that everyone already anticipated may do little when it finally arrives.

This is how the Fed can lower its rate while the ten-year rises. Higher Treasury yields can then put upward pressure on mortgages and other long-term financing. The final loan rate also depends on the lender’s funding costs, the borrower’s credit, collateral, fees, and the spread charged above the relevant benchmark.

The ten-year is a useful reference point. It is not a universal pricing formula.

The Credibility Tax

There is a particular risk I watch when political pressure for easier money meets unresolved inflation.

If investors believe the central bank is becoming less willing to defend price stability, they may demand more compensation to lend for the long term. That can appear in expected inflation, in the anticipated path of future rates, or in the premium investors require for bearing risk.

I think of that as a credibility tax.

It is a description of a possible market response, not a separate number we can read off a screen. We cannot look at a rising ten-year yield and declare that every basis point represents lost confidence in the Fed. Stronger growth, Treasury issuance, changes in overseas demand, and investor positioning can produce similar movements.

The distinction matters because an argument about credibility should be held to the same standard of evidence it asks of the central bank.

When long-term yields rise more than short-term yields, traders call the movement a bear steepener. When short yields fall while long yields rise, a steepening twist is the more precise description. Both widen the gap between the short and long ends. Neither identifies the cause by itself.

We saw a sharp divergence around the July meeting. Between July 28th and July 29th, the two-year Treasury yield fell from 4.26 percent to 4.22 percent, while the thirty-year rose from 5.09 percent to 5.20 percent. The Fed left its policy rate unchanged. The market moved anyway.

On August 13, a thirty-year Treasury auction cleared at 5.216 percent, its highest auction yield since 2001. That established the price investors required at that sale. It did not establish a single explanation for that price.

The warning for borrowers is substantial without making it larger than the evidence: long-term money can become more expensive while the Fed stands still.

What Burns Left Behind

Arthur Burns chaired the Federal Reserve from 1970 to 1978. The Nixon tapes document presidential pressure for expansionary monetary policy ahead of the 1972 election. Policy was expansionary, although historians and economists still debate how much reflected political pressure and how much reflected Burns’s own convictions.

The inflation that followed had several causes. Monetary accommodation mattered. So did oil shocks, fiscal pressures, failed wage and price controls, and mistaken judgments about how much the economy could produce without generating inflation. Reducing that decade to one president leaning on one chairman would make the story simpler and less accurate.

But the failure to sustain restraint had lasting consequences. Once households, businesses, and investors came to expect continuing inflation, bringing it down became more difficult.

On September 30th, 1979, Burns delivered “The Anguish of Central Banking” in Belgrade. He acknowledged that the Fed had possessed the power to stop inflation through sufficiently restrictive policy. He also argued that the institution had been caught in the political and intellectual currents of its time, unwilling to maintain the restraint required.

Six days later, under Paul Volcker, the Fed announced a major change in its operating procedures, placing greater emphasis on controlling reserves and monetary growth. The timing is striking, but Volcker’s policy shift was already being developed. Burns’s lecture should not be treated as the event that suddenly converted him.

Restoring price stability took years. The United States endured recessions in 1980 and 1981–82, and unemployment reached nearly 11 percent late in 1982. Long-term rates remained painfully high along the way: the ten-year Treasury yield averaged 15.32 percent in September 1981. Those yields reflected inflation, expected policy, and risk—not a pure measure of the term premium.

The leadership lesson I draw is that repeated accommodation can make the eventual correction more costly. The people who bear that cost may have had no voice in the decisions that created it.

Subscribe now

Why This Meeting Matters

The current inflation picture is mixed, and the mixture matters.

August headline CPI was 3.4 percent above a year earlier, unchanged from July. Gasoline rose 3.9 percent during the month, seasonally adjusted, and the energy index was up 16.3 percent over the year.

Core CPI, which excludes food and energy, improved to 2.4 percent year over year from 2.5 percent in July. But its monthly increase accelerated from 0.2 percent to 0.3 percent. The annual figure improved; the monthly figure did not. These are CPI readings, while the Fed’s 2 percent objective is defined using the separate PCE price index.

At Jackson Hole on August 28th, Warsh placed responsibility for persistent inflation on the central bank and said underlying inflation needed to move toward the objective clearly and fast enough. He also explicitly declined to commit to a particular decision. Three FOMC members had dissented in July because they wanted a quarter-point increase then.

A surprise cut after that would require a persuasive explanation. If investors saw evidence of serious economic weakness, longer-term yields could fall. If they saw accommodation of persistent inflation or political pressure, those yields could rise.

My concern is the second possibility. It is a risk, not a forecast I can honestly present as certainty.

The labor market gives the Fed reasons to be cautious about adding stimulus. August payrolls increased by 162,000 and unemployment held at 4.1 percent. But the preceding twelve months averaged only 31,000 additional jobs per month. One stronger report does not erase the slower trend.

Meanwhile, long-term financing has already become expensive. Freddie Mac’s thirty-year fixed mortgage average was 6.76 percent on September 10. Total federal debt passed $40 trillion in August. The Fed’s July target range remained at the level established by its December 2025 cut.

Borrowers do not need a theory about every movement in the bond market to recognize the practical problem. A stationary policy rate has not meant stationary financing costs.

The Argument for Patience

There is a serious argument against tightening into an energy shock.

Higher fuel costs already squeeze households and businesses. Real average hourly earnings for private-sector employees fell 0.3 percent over the year through August. That figure does not measure median household income, and real average weekly earnings rose 0.3 percent as the workweek lengthened. Still, the hourly measure shows purchasing power under pressure.

The Fed cannot drill for oil. It cannot reopen a shipping route. Higher rates can weaken demand without repairing the supply disruption that pushed prices up.

The comparison with the 1970s also has limits. The United States is now a net energy exporter, although domestic consumers remain exposed to global energy prices. Today’s inflation is far below the double-digit rates of the Great Inflation, and monetary policy operates within a different institutional framework.

Those differences deserve weight. A central bank can reasonably look through a temporary supply shock when it has grounds to believe inflation expectations will remain anchored. It must also consider the risk that the shock spreads into broader prices and expectations.

Credibility gives policymakers room to exercise that judgment. Preserving it requires decisions the evidence can support, including a willingness to explain uncertainty.

I do not know what the long end will do if Warsh hikes tomorrow. A hike could reassure investors about inflation control. It could reinforce expectations of further tightening. It could produce little movement because investors already expected it.

The same uncertainty applies to a hold or a cut. The announcement matters through what it changes in the market’s understanding of the years ahead.

That is why a business plan should survive more than one interpretation of a press conference.

Know Which Rate You Pay

Start with the obligations already on your balance sheet.

For each one, write down the current rate, whether it is fixed or variable, the benchmark if it floats, the spread above that benchmark, any floor or cap, the next reset date, and the maturity date.

Many variable-rate 7(a) loans and business lines use prime. Other loans use different benchmarks, including SOFR. Even when prime changes promptly after a Fed decision, a borrower’s rate adjusts according to the note.

For twenty- and twenty-five-year SBA 504 financing, the SBA-backed debenture portion receives a fixed rate through a monthly sale, priced relative to the ten-year Treasury. The accompanying bank loan has separate terms. An existing fixed debenture does not reset because the next month’s published rate changes.

Fixed mortgage rates are influenced by mortgage-backed securities and Treasury markets. Equipment loans and corporate borrowing reflect their own funding benchmarks and credit spreads. The ten-year helps explain the environment, but your lender’s actual terms determine your payment.

If your debt is fixed through maturity, tomorrow’s decision does not change that contracted interest rate. Your exposure lies in new borrowing, a future reset, or refinancing. If your debt floats, a policy change may reach your cash flow sooner.

That distinction is more useful than a blanket claim that a hike hurts or a cut helps.

Stop Underwriting the Rescue

In May, I suggested modeling a sustained 7.5 percent borrowing cost for the next two years. I would keep that scenario in the model, with a clarification: it is a planning assumption, not a universal market rate or a forecast that every borrower will pay the same price.

Your base case should start with current lender quotes for your actual credit, collateral, and loan structure. If those quotes are above 7.5 percent, the model needs to reflect that. Then test what happens if rates stay elevated, refinancing comes later, or cash flow weakens.

If a maturity is approaching inside eighteen months, talk to your lender this month. Compare refinancing terms, extension options, prepayment costs, and the cost of fixing a rate. Understand what can be committed today and what remains subject to market pricing.

A rate lock can reduce uncertainty. It also has terms and costs. The objective is to protect a viable business from a financing deadline it cannot afford to miss.

Waiting for a cut to save the deal is a financing assumption. Put it in the model where everyone can see it, alongside the case in which the cut arrives and your borrowing cost does not fall.

Watch the curve, too. The two-year and thirty-year yields can show whether expectations for near-term policy and long-term money are moving together. Ten-year TIPS breakeven inflation can add information, but it includes risk and liquidity effects as well as expected inflation.

Treasury auctions offer another piece of evidence. An auction that clears at a higher yield than the market expected immediately beforehand has “tailed,” an indication that demand was weaker than anticipated at that price. One auction is not a verdict on the central bank. Repeated patterns deserve attention.

Read those signals together. Then compare them with the financing terms your business can actually obtain.

Know Who Sets the Terms

Dickens gave Pip a fortune and let him spend much of Great Expectations certain he knew who had provided it. He was wrong about his benefactor, and the discovery changed his understanding of the life he had built.

Borrowers can make a related mistake. They see the committee in Washington and assume it determines the price of their next decade of financing.

The Fed is powerful. Its decisions shape markets, spending, employment, and inflation. But the rate on your next loan also reflects investors’ expectations, the lender’s economics, and your own ability to repay.

A cut cannot guarantee a lower long-term rate. A hike cannot tell you, by itself, what your next mortgage or commercial property loan will cost.

The practical response is within reach: know the terms you have signed, preserve cash, prepare early for maturity dates, and build assumptions your business can survive.

None of that requires predicting Kevin Warsh. It requires understanding the obligations you carry and retaining enough room to act when conditions change.

For leaders in a crisis era, that is also a question of conduct. Pressure makes convenient explanations attractive. Institutions survive when the people responsible for them can distinguish an explanation from a justification—and accept the cost of a decision they can defend.

That is the warning I take from Burns. In the language of The Fourth Turning Leader, it is the danger at the heart of the Seneca mode: reasoning that becomes so accommodating that it protects the leader from confronting the consequences of his own choices.

Share

That tension is explored in Honor Under Pressure, Book One of The Fourth Turning Leader series. The framework and practical resources for leadership when trust, authority, and institutional continuity are at stake are available at The Fourth Turning Leader.

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.

I have my own list, and I still know every name on it. What I want to talk about is what the list does to you afterward, because I've watched it happen to better leaders than me and I've caught it happening in myself. The burn heals. The lesson you take from it doesn't, and the lesson is almost always the wrong one.

The wrong lesson goes like this: I trusted too fast, so from now on people will earn it. Trust becomes a thing you ration, so new hires start at zero and climb. Every delegation gets a check-in, every check-in gets a follow-up, and somewhere around the fourth signature on the approval form you have built an organization that runs on suspicion and calls it prudence. There's nothing wrong with a season of caution after a betrayal. There's a great deal wrong with caution that never ends, because it stops being a response to one person and becomes the operating system for everyone. It feels like wisdom, but it is simply the last person who burned you, still running your company from wherever he ended up.

The Bill Comes Due Everywhere Else

Here's the part that took me too long to see: the list is short, and the people paying for it are many. Two names, maybe three, over a career, set against hundreds of people who showed up, did the work, told you the truth when it cost them something, and never once gave you a reason to check behind them. When you ration trust because of the two, the hundreds are the ones who feel it. They don't know about your list, only that the boss reads their emails twice.

And they learn from it, because trust is reciprocal in a way that most leadership advice skips right past. A team member who is handed distrust by default doesn't wait patiently to graduate out of it. She gives it back. She assumes your questions have motives and reads a scheduling change as a message. Small misunderstandings that a trusting relationship would absorb in a hallway conversation turn into three-week grudges, because nobody on either side is willing to assume good faith first. Leaders who trust slowly teach everyone around them to trust slowly, and then wonder why the building feels cold.

At B:Side, when I do my rounds of all-staff one-on-ones, the thing people describe wanting, once you listen between the lines, is predictability from the top. They want to know that the way I decided something yesterday is the way I'll decide it tomorrow. That is them working out whether they can trust me. At the same time, I'm working out whether I can trust them, and whoever moves first sets the terms for both. At ASU, I put students into teams in the opening weeks of the semester with people they met days before, and I've found the same math plays out in miniature. The teams that assume the best of each other move. The teams that wait for proof spend the term auditing one another's contributions. Nobody on those teams has been burned yet, by the way. They've simply been raised in a culture that treats suspicion as sophistication.

That culture is why this matters more now than it did ten years ago. We're living through a stretch in which trust in nearly every institution is thin: banks, universities, the press, the government, each other. Meanwhile, the teams themselves are getting smaller, and I wrote in February about why: AI is taking over the work junior hires used to do. One consequence of a leaner org chart is that every person on it carries more weight. A leader who has to verify everything a team of eight does will run out of hours before he runs out of things to verify. Trust has always been the cheapest coordination mechanism there is. In a small shop under pressure, it's the only one that scales.

Seneca's Order of Operations

Seneca wrote to his friend Lucilius about this two thousand years ago, and his third letter is the cleanest thing I've read on the subject. Lucilius had sent a letter by way of a mutual friend and then, in the same breath, warned Seneca not to tell that friend everything. Seneca called him on it. If you'd trust a man to carry your mail but you wouldn't trust him with your mind, he said in effect, then you're using the word "friend" the way people use it for everyone they've ever been introduced to.

The heart of the letter is a sequence, and the sequence is the whole point. Deliberate for a long time before you admit someone into your confidence. Then, once you've decided, take him in with your whole heart and speak to him as freely as you'd speak to yourself. Seneca's complaint was about people who run it backwards: they make a friend first and judge him afterward, which is exactly how you end up with a list of names. Judge slowly. Then trust completely. Both halves are instructions, and the second one is the half we skip.

He also saw the mechanism I described above, and he described it better. Some people, he wrote, fearing to be deceived, have taught others to deceive; by their suspicion they hand the other person the right to do wrong. Regard someone as loyal and you tend to make him loyal. Regard him as a suspect and you've told him what you expect. Suspicion works as a curriculum, and the students learn fast.

Seneca also named the failure on either side of the line. There's the man who tells everything to everyone he meets, and there's the man who tells nothing to his closest friends. Both are faults, he said, and then he made a distinction I keep coming back to: he called the first the more honest of the two mistakes, and the second the safer. Trusting everyone is a failure of judgment, while trusting no one is a failure of nerve that happens to look like experience.

There's a bitter footnote to the letter, and I don't think you can read Seneca honestly without it. He wrote these letters in his last years, after more than a decade at the elbow of an emperor who trusted no one, and not long after finishing them Nero ordered him to open his veins. Seneca knew what a court without trust looked like from the inside. He described the cure precisely and lived inside the disease. I've written before about the gap between what Seneca preached and what he practiced, and I don't think the gap disqualifies the advice. It tells you what the advice costs when the man at the top refuses it. A court where the ruler trusts nobody ends up with nobody worth trusting. The only people who stay are the ones who've learned to survive suspicion, and those are precisely the people you should never have let in.

Marshall Handed Over the War

Seneca gave the principle, and George Marshall ran it at scale, which is why I find him the more useful teacher: he had a payroll. I wrote a week's worth of essays about Marshall two years ago, so I'll keep this to the one habit that matters here.

Marshall never commanded troops in battle, so his whole war, from the fall of 1939 to the fall of 1945, was a sequence of decisions about which men to trust with things he couldn't do himself. He did the judging half of Seneca's sequence with a ruthlessness that still makes people uncomfortable. Early in his tenure he stood up a board, with his own predecessor running it, to clear out the senior officers who couldn't keep pace with the war that was coming. Several hundred of them went, and the only criterion that counted was how a man was performing today. The legend says he kept a little black book of the officers he'd been watching since his years at Fort Benning. The book, it turns out, never existed (he kept the list in his head, which is worse for the officers and better for the story).

What he did once the judging was done is the part worth studying. In December 1941, a week after Pearl Harbor, he called a brigadier general he barely knew into his office. Dwight Eisenhower had spent most of the previous twenty years as a staff officer and had never commanded in combat. Marshall laid out the situation in the Pacific, asked him what the country's general line of action ought to be, and waited. Eisenhower asked for a few hours and came back with a recommendation: do everything possible to support the Philippines while accepting that they probably couldn't be saved, and build Australia into the base for whatever came next. Marshall agreed and told him to do his best. That was the whole handoff.

Around the same time, Marshall told him something Eisenhower put in his memoir years later. The War Department, Marshall said in substance, was full of capable men who analyzed their problems well and then felt compelled to carry them to his desk for the final decision. He needed assistants who would solve their own problems and tell him afterward what they'd done. Six months later he sent Eisenhower to London to command every American soldier in Europe. Eisenhower had been a lieutenant colonel eighteen months before.

The judging was slow, years of it, conducted from memory across peacetime posts. The trust, once extended, was nearly total, and Marshall didn't fly to London to look over Eisenhower's shoulder. When Roosevelt later tried to hand Marshall the command of the invasion of France, the job every soldier of his generation wanted, Marshall refused to ask for it. The President, he said, should decide on the country's interest rather than on his. Then he wrote out the cable announcing Eisenhower's appointment by hand and sent the draft to Ike as a keepsake. You cannot build anything that outlives you without handing the important things to people who might drop them. Marshall knew that, and he handed over the war.

Subscribe now

The Math of Suspicion

The first cost of withheld trust is friction that never gets billed to the right account. Every layer of verification you add because of a name on your list costs hours from people who aren't on it. Think of the report that gets reviewed twice, the decision that waits for your return from a trip, and the contract a vice president could sign but sends up anyway. She's learned that signing it herself earns a question and sending it up earns nothing. Multiply that across a year and you've paid for your suspicion several times over in salary, and you've paid a second time in the people who left because they were tired of being audited.

Lending gave me a sharper way to see this, because a loan is trust with paperwork attached. Before B:Side puts money out the door, we do a great deal of judging: cash flow, collateral, the character of the owner, the history of the business. That's the slow half, and it should be slow. However, once the loan closes, we don't call the borrower every Tuesday to ask whether she's still in business. We ask for financials on a schedule, we watch the payments, and we let her run her company. The verification is proportional to the stakes and it's attached to the process, never to a running suspicion of the person. A lender who couldn't extend trust after underwriting would be a lender with no loans.

I'll grant the strongest objection here, because I've made it myself. Some of us work in businesses where a single person acting in bad faith can do real and irreversible damage, and "trust first" sounds like an invitation. That's true, and it's why Seneca put judgment before trust rather than instead of it. It's also why controls belong on transactions and never on individuals. Two signatures on a wire transfer protect everyone in the building, including the person signing; it's a control on a step where a mistake can't be undone. A second set of eyes on a person you chose says something different. It says you didn't really choose. Keep the controls where the consequences are irreversible, and take them off everywhere else.

The second cost is quieter, and it lands on people who did nothing to deserve it. My students haven't been burned yet, and they still walk into their first jobs braced for a manager who checks everything, because that's what they've been told professionalism looks like. The ones who land with a leader who hands them a real problem in the first month light up in a way I can see from across a room. The ones who spend their first year being verified learn to do exactly what's asked and nothing more. A person's worth is settled long before you finish deciding about them, but their growth waits on the decision. Nobody becomes capable inside a relationship where capability is presumed absent until proven otherwise, which means the guarded leader is doing more than protecting himself: he's stunting the people he's responsible for.

Start With the People You Didn't Choose

The hardest version of this problem arrives when you inherit a team. I took over B:Side in September 2020, in the middle of the pandemic, with a staff I hadn't hired, in a year when meeting anyone in a room was complicated. Seneca's sequence assumes you get to do the judging first. I didn't. I could withhold trust from everyone until I'd formed my own opinion, which would have taken a year I didn't have. Or I could extend it on the strength of my predecessor's judgment and correct as I went. Only one of those keeps the doors open. I took it, I got some of it wrong, and it was still the right call. After all, the organization ran while I learned it. If I'd waited to trust until I was certain, the waiting would have been the message, and the message would have been received.

If you've been burned recently, the temptation is to treat the wound as data. It is data about one person, and the mistake is letting it write policy for everyone else. The betrayer is the exception, and a great deal of leadership comes down to refusing to let the exception govern the rule. Everyone else on your team is counting on you to remember that they aren't him.

If you're early in your career, understand that the feeling of being trusted is one of the great rewards of working life, and that the fastest route to it is unglamorous. Make small promises and keep them with precision: tell someone you'll have it by Thursday at ten, and have it by Thursday at ten. People who can't be counted on for the small things are assumed, correctly, to be unreliable on the large ones. People who nail the small things get handed the large ones sooner than they expect. Consistency is the only currency that buys it, and it compounds.

Make the First Move

Trust is something you do before it's something you feel, and the leader does it first.

Finish the judging, then stop. Do the reference calls, watch how a person behaves when the deal is going sideways, ask the questions you're afraid to ask, and then decide. A hire you keep re-evaluating is a hire you never made.

Say the trust out loud. Tell the person what you're handing them and that you won't be behind them checking. "This is yours; tell me what you did" changes how someone works in a way a job description never will. Marshall said it to Eisenhower in nearly those words, and it's the sentence Eisenhower remembered.

Multiply the reps. Trust grows on repeated contact, so in a new relationship, compress the calendar. More conversations, more decisions made together, more problems worked through side by side in the first ninety days than the normal cadence would produce in a year. Density substitutes for time.

Put controls on transactions, never on people. Dual approval on a wire, a second review on a loan above a threshold, a checklist before anything irreversible. Those protect the person doing the work, while a standing second opinion on someone you've already chosen does the opposite.

Keep the other list. You already know the names of the ones who burned you. Count the others: the people who handled something hard without telling you about it, who returned a favor you'd forgotten you did, who stayed through a year that gave them every reason to leave. That list is longer, and it's the one that should set your defaults.

Deal with the exception fully and fast. When someone does break the trust, act. Confront it directly, end it if it has to end, and don't drag it out because you'd rather not be the one to say it. Speed here is a kindness to everyone else, because a betrayal that lingers unaddressed teaches the team that trust carries no consequences in either direction.

The List Stays Short

I want to be honest about what this doesn't fix. Will you get burned again? Almost certainly. If you lead people long enough and trust them the way I'm describing, the math guarantees it, and I don't have a version of this argument that makes the next one hurt less. What I have is the comparison: two or three names over a career, against the hundreds who deserved a leader who assumed the best of them and got one. The guarded leader keeps his list short by trusting no one, and he pays for it every day in everything the untrusted never bring him. The other kind keeps his list short because most people, judged carefully and trusted fully, turn out to be worth it.

Seneca wrote the sequence down for a friend, and Marshall ran it on a war. The rest is a decision you make every morning about the people who walk in the door and haven't done anything wrong.

Judge slowly. Then hand over the keys.

Share

My latest book, Honor Under Pressure, is out now. It's about making the hard call before the pressure makes it for you. You can find it on Amazon.

The same arrangement can quietly become something else. Patience turns into explaining away impulsive decisions. Attention to detail becomes repairing commitments the leader makes without consultation. Skill with difficult conversations becomes a standing obligation to tell people what the leader will not say directly. The team is still compensating for a limitation, but the limitation no longer has to improve.

From the outside, the distinction may be hard to see. The organization continues to function. The leader appreciates the team’s loyalty. The team takes pride in being able to work with someone others find challenging. Yet a growing share of its capacity is being used to preserve the leader’s preferred way of operating.

This is one of the less obvious ways a weakness becomes institutional. It does not have to appear in a policy or a formal instruction. It can shape who gets close to the leader, who stays there, and which kinds of judgment become useful inside that circle.

An impatient leader may initially value people who can get to the point. That is reasonable. Over time, however, the circle may favor people who can compress every issue into a quick answer, even when the issue requires uncertainty to remain visible. The people who insist on explaining a dependency or a limitation begin to seem cumbersome. Eventually, the leader receives faster answers and less of the information that makes the answers reliable.

A leader who dislikes conflict may rely on a colleague with greater interpersonal skill. Again, that can be wise. The problem develops if the colleague becomes responsible for absorbing every reaction to the leader’s decisions. The leader receives agreement; someone else receives the disappointment, confusion, and anger. The arrangement protects the leader from feedback that might otherwise change how decisions are made.

An insecure leader faces a similar risk. Reassurance is pleasant, and trusted advisers naturally provide some of it. But if access increasingly depends on helping the leader feel right, the inner circle will become better at reassurance through selection alone. People who resist that role may leave, lose influence, or simply learn when to remain quiet. No explicit demand for flattery is necessary.

These patterns can be difficult to recognize because each individual choice has a plausible explanation. One person is too slow. Another has poor chemistry with the group. A third is highly capable but seems unable to understand the leader’s style. Any one of those judgments may be accurate. The pattern across them deserves a separate examination.

Look at who has become less influential and why. Do the people who lose access share a tendency to introduce inconvenient information? Do the people who gain access share a talent for making difficult facts feel less difficult? Has the team become more capable of challenging the leader’s judgment, or more capable of working around it? The answers may reveal something that no personality assessment will capture.

Subscribe now

Examine the adaptations the team has learned to make. People who manage a leader’s limitations are often acting responsibly under imperfect conditions. They may be protecting colleagues, keeping promises, and preventing unnecessary harm. Blaming them for adaptations the leader rewarded would repeat the very failure being examined.

The more useful question is what those adaptations are costing. A trusted colleague may be spending hours repairing ambiguous instructions before work can begin. Another may be privately reassuring people who received a dismissive response. A third may carry the memory of every exception because the leader resists documenting decisions. These efforts can be valuable in the moment while also allowing an avoidable pattern to continue.

The cost is partly capacity. Work devoted to translating, cushioning, and repairing is work unavailable for the organization’s purpose. The deeper cost is distorted feedback. When other people routinely absorb the consequences of a leader’s habits, the leader receives weaker evidence that those habits need attention. The team’s competence can make the underlying problem harder to see.

There is a legitimate objection here. Leaders cannot be good at everything, and expecting them to become equally skilled in every dimension defeats the purpose of building a complementary team. A founder may never be the strongest operator. An excellent operator may need help communicating a broader direction. Dependence on other people’s strengths is a condition of serious work.

The distinction lies in what happens to judgment and responsibility. A complementary colleague can exercise their expertise, name the limitation, and influence the decision. A protective colleague may be allowed to manage the consequences while the leader retains the habits that create them. One arrangement broadens what the team can see and do. The other narrows what the leader has to encounter.

A practical test is whether the colleague can tell the leader that the compensating work has become excessive. Can the operations leader say that another promise cannot be absorbed? Can the trusted adviser insist that the leader personally deliver a difficult message? Can the person who supplies context require enough time for that context to be heard? If those requests threaten the relationship, the complement has begun to function as protection.

Repair starts with the leader accepting a more direct share of the cost. If someone has been cushioning the consequences of your decisions, join the conversation where those consequences are explained. If your speed forces others into repeated rework, inspect that rework before celebrating decisiveness. If your discomfort with uncertainty produces premature conclusions, require yourself to stay with an unresolved question long enough to understand it.

This is likely to make the team feel less smooth for a while. Information that was previously filtered will arrive in a rougher form. Meetings may reveal disagreement that once took place privately afterward. The leader may feel less understood precisely because other people have stopped making understanding effortless. That discomfort can be evidence that the arrangement is becoming more honest.

It also helps to examine how access is distributed. A leader who hears only from the same interpreters will eventually depend on their account of reality, however capable and well intentioned they are. Occasional direct contact with the work can reveal what has been omitted or normalized. The purpose is to understand the conditions people face, including the conditions created by the leader, while preserving clear responsibilities for decisions.

Hiring and promotion deserve the same scrutiny. Chemistry matters, but comfort is a poor substitute for capability. Before calling someone a strong fit, identify the useful friction they are likely to bring and whether you are prepared to tolerate it. A team advertised as diverse in thought will have little practical diversity if everyone learns that influence depends on presenting the leader with the same emotional experience.

There is a form of humility in allowing trusted people to stop rescuing us from every consequence of our own habits. It asks more than admitting a weakness in the abstract. It asks us to change the relationship that has made the weakness convenient.

The people closest to a leader will inevitably learn how to work with that leader. A strong team should make its leader more accountable to reality.

Share

One line remains unwritten: what happens if the forecast is wrong.

There may be a contingency percentage in the budget. There may be a general assurance that the organization can adjust. But nobody has described how many existing commitments would be delayed, which people would absorb the recovery work, or how the service would be withdrawn from customers who had begun to depend on it. The proposal has priced the launch. The institution is still being asked to accept an undefined loss.

That omission matters even when the idea is good. An attractive opportunity can deserve pursuit while requiring a smaller commitment, a different sequence, or a limit on what the organization promises before it has learned enough. Leadership enters at precisely that point. Someone has to decide how much uncertainty other people will be asked to carry.

The cost extends past the budget

Organizations are reasonably comfortable discussing the money they might lose. They are less comfortable discussing the obligations that loss creates.

A failed initiative can consume the attention of the strongest manager for months. It can require people to repair relationships while trying to meet their original targets. It can leave an ordinary team carrying an extraordinary workload because the temporary project has ended but its consequences have stayed. The financial loss may have a clear owner. The rest of the burden often disperses into the organization without acquiring a name.

This is how an apparently bounded decision becomes an open claim on everyone nearby.

The leader who approved the initiative may still be able to explain the original reasoning. That explanation does little for the people whose capacity now covers the gap. Ownership requires anticipating their position while there is still room to change the design. Their effort belongs in the initial decision, even when it cannot be converted neatly into dollars.

The same is true of customer trust. A business can reverse an internal experiment relatively easily. Reversing a promise requires another person to reorganize around the change. The service may disappear from the company’s plan in a single meeting. It remains in the customer’s plans until someone helps them replace it.

Reversibility has to be assessed from both sides of the promise.

That is why the size of a decision cannot be measured only by the check being signed. A modest expenditure can create a large obligation. A more expensive test can preserve flexibility if the participants understand its limits and the organization has prepared a credible way to finish it.

Choose an exposure that teaches

The response to uncertainty is often to make the first move smaller. That can be sensible. It can also become a way to avoid learning anything consequential.

Suppose the proposed service depends on customers using it without intensive personal assistance. A small trial in which senior staff guide every customer may produce glowing feedback while leaving the central assumption untouched. The organization has protected the outcome so thoroughly that the test no longer resembles the work it intends to scale.

Subscribe now

A useful trial has to expose the assumption that matters. It needs enough contact with ordinary conditions to produce information worth acting on. At the same time, the people involved need a level of protection appropriate to the consequences. Those requirements belong together. An experiment can be informative without being careless.

The practical question is what can be learned before the organization makes its most difficult commitments to reverse. Perhaps the service can be tested with a clearly described limited offering. Perhaps the uncertain work can be observed manually before anyone signs a long contract. Perhaps demand can be examined before permanent staffing grows around an optimistic forecast.

The sequence should follow the uncertainty.

A team that learns the most consequential thing late has already spent much of its freedom. By then, the company may have hired people, announced a launch, and tied its reputation to the outcome. Each additional commitment makes a change of course harder to judge on its merits. The discussion begins to include embarrassment, sunk effort, and the fear of disappointing people. Those are real costs. They are also reasons to learn earlier.

A deliberate limit creates permission to pursue the idea with seriousness. People can commit to the work because they know what they are committing the institution to bear.

Put a decision at the boundary

A limit that nobody has authority to enforce is an aspiration.

Before the next expansion, the leader needs to identify what will be reviewed, who will review it, and what decision that person can actually make. A review might release the next stage, narrow the offer, require a correction, or end the effort. Its authority should be clear before a room full of invested people begins defending continuation.

The review also needs evidence that can arrive in time. Revenue may take months to develop. Operational strain may appear in days. Customer understanding may be visible in the first few conversations. These signals answer different questions. Waiting for the final outcome can allow a burden to grow long after the organization has enough evidence to alter its exposure.

This does not require a complicated system. A short account can state the assumption being tested, the resources committed, the obligations created, and the condition that requires another decision. The important work is the thinking that makes those statements specific.

For the new service, the team might decide that expansion depends on being able to deliver within ordinary staffing capacity. If repeated manual intervention remains necessary, the next stage waits for a revised design. The point is to define what that intervention means before staff begin treating it as the normal price of keeping the launch alive.

There is a difficult distinction here. Unexpected strain can reveal a flawed idea. It can also reflect the ordinary expense of learning a worthwhile one. A threshold should bring the choice back into view. It cannot make the choice by itself.

Leaders still have to interpret the evidence. What they gain is a moment when interpretation becomes explicit, with the original reasoning available and the accumulating burden visible. They can choose to commit more. They should know they are making a new commitment.

The danger is the quiet extension. Another week of extra work, another exception, another assurance that the next stage will resolve the problem. Each concession looks manageable alone. Together they can create an exposure nobody ever approved.

That is also why the record should preserve the reasoning behind a limit. Conditions may change. A new piece of evidence may justify a larger commitment. Revising a boundary responsibly requires showing what changed in the situation and why the original concern is now addressed. The desire to continue is insufficient evidence by itself.

There will always be opportunities that require substantial conviction. Some cannot be reduced to a tidy pilot. Some demand a commitment before the evidence is complete. Leaders earn their responsibility in those moments by stating the burden honestly and accepting accountability for the people who will carry it.

Courage is compatible with limits. So is ambition. A leader can believe deeply in an opportunity while making a sober account of the damage an error could cause. That account makes the commitment more credible because everyone can see that its consequences have been considered.

Before approving the next initiative, look for the blank line in the proposal. Follow failure beyond the budget and into the work, relationships, and promises that would remain. Decide which of those burdens the institution can carry while it learns, and which would compromise duties it already owes.

Choose the cost before others inherit it.

Share

From my position as a CEO, it’s easy to see why people would want that person involved the next time. There’s less uncertainty, the customer feels heard, and the team can return to the work it already knows how to do. But the next time carries a question that the relief of this time can obscure: how much more capable has anyone else become? A successful rescue can leave the business exactly as dependent as it was before.

That possibility should trouble those of us responsible for developing other leaders. A company can accumulate years of profitable operation while its ability to handle the hardest decisions remains concentrated in the person who started it. A business can grow old without learning how to survive its leader.

A Change of Owner Tests the Whole Business

In a Gallup survey conducted in fall 2024, 22 percent of employer-business owners said they planned to sell or transfer ownership within five years. The findings, published in March 2025, covered people who spent most of their working hours in businesses they owned. Those answers describe intentions, and tell us nothing by themselves about whether the businesses are ready. They do place the work of preparing another owner within a horizon that should concern anyone still handling every consequential decision. Gallup’s report

A transfer puts the distinction between personal ability and organizational ability under pressure. The customer list can change hands while the customers continue to trust the departing owner, leaving a buyer with relationships that still depend on someone else’s presence. A complete set of procedures may offer little help if the most important decisions have always happened outside them, especially when the departing founder supplied the explanation that made contradictory instructions work together.

The calendar won’t teach anyone. If the business needs someone else to exercise that judgment after a sale, it needs to give that person opportunities to develop it beforehand. The same obligation exists when the founder plans to stay but wants to open another location or give a manager responsibility for an important customer. Each expansion asks the business to carry its standards farther from the person who first supplied them.

From the lending side, the distinction matters because the cash flow being reviewed was produced by a particular arrangement of people and relationships, which a transfer may change. A founder may have been limiting price concessions, renegotiating promises the operations team couldn’t fulfill, or repairing a customer relationship before the customer took its business elsewhere. A new owner’s ability to continue those interventions belongs in the assessment of whether past results can support future expectations. An operating history deserves respect, along with a careful account of who made it possible. Otherwise, we risk treating the departure of a central decision-maker as an administrative detail in a business whose success depended on that person’s daily presence.

An owner can deliberately build a business around personal work and plan to close it when that work ends. The obligation changes when the owner intends to sell a continuing enterprise, or asks employees to build their futures around its survival. That intention creates a responsibility to prepare other people while there is still time.

Time Tests Whatever Meets It

Nassim Nicholas Taleb’s discussion of the Lindy effect offers a useful starting point: for certain things without an intrinsic biological lifespan, including ideas and practices, longer survival can imply a longer expected remaining life. Taleb develops the idea in Antifragile and returns to it in his essay “An Expert Called Lindy,” where exposure to consequences is essential to the test. Something that persists while protected from those consequences can give us misleading evidence about its strength. Taleb’s explanation

That is a valuable discipline in business, where novelty can attract more attention than the reasons an older practice works. A rule that has survived changing customers and difficult markets may contain knowledge its current users would struggle to reconstruct. The age of the rule gives us a reason to investigate before replacing it. It deserves a hearing.

Subscribe now

However, applying that idea to a company requires deciding what the company’s years have actually tested. A business is an arrangement of capabilities, some held by individuals and others spread through the organization, and its name can remain unchanged while the arrangement beneath that name changes substantially. The evidence supplied by its history has to be connected to the capabilities it will need next, particularly when the people carrying those capabilities are preparing to leave.

Consider the difference between a team that has learned how to handle customer disputes and a team whose founder has handled every difficult dispute. Both can belong to businesses with long records of satisfied customers, yet those records support different conclusions about what will happen when a new manager answers the phone. The second team’s experience has taken place largely within reach of someone else’s judgment.

The founder’s contribution is real in both cases, and the years may have tested that person’s judgment severely. Yet a record earned with the founder present leaves the business’s ability to operate without that person partly untested. My application of Lindy begins with that distinction: identify the capacity responsible for survival, then determine whether it will remain available when the person responsible for it leaves.

The same examination applies to the practices a founder leaves behind. A longstanding approval rule may encode a painful lesson about promises the business couldn’t afford to keep, or persist because changing it creates an argument with the owner. Time can pass around either arrangement, leaving the next generation with the work of discovering which explanation fits.

For a standard to survive succession with its purpose intact, people need to understand why it exists and practice applying it. They also need permission to correct it when circumstances expose a weakness that the person who created it couldn’t have anticipated. Each repetition gives someone an opportunity to acquire judgment that previously belonged to someone else. Experience becomes transferable when another person can use it to make a sound decision.

A procedure stored in a folder is only part of that acquisition: the person applying it needs to recognize the situation it was designed for. They also need to notice when the present case differs in a consequential way. Without that understanding, a successor can preserve every instruction while losing the purpose those instructions once served. Continuity requires people who can think with the inheritance they receive.

There’s also the matter of who pays for the business’s apparent resilience. If continuity requires the founder to remain available during every vacation or quietly work the hours a missing manager would have supplied, the cost is being carried somewhere. A buyer needs to understand the work behind those earnings and whether someone else can realistically take it on. The founder’s family may already have a clear account of how much uninterrupted service has cost, even when the business records give that contribution no separate name.

A Rescue Can Become a Ceiling

From the CEO’s chair, there are sound reasons to intervene. A manager’s hesitation can carry a cost the business has little room to absorb, while the founder may understand details that genuinely change the decision. Telling that person to step aside in the name of development would be irresponsible when the immediate consequences exceed what the team can safely manage.

The responsibility being transferred should match the person’s readiness and the business’s capacity to absorb error, with enough room for the person to acquire experience before the next transition. Some decisions require close supervision, while others can become useful practice with clear limits and a timely review. What matters is whether intervention leads to a more capable team or simply establishes another reason to intervene next time.

The distinction applies at my own desk, where I have to examine what the arrangement is teaching if every difficult question ends with my answer. A manager who learns that uncertainty should immediately travel upward may become very good at identifying the moment to stop thinking. Calling that person dependent would leave my part in the arrangement conveniently unexamined.

This was a systemic issue at B:Side in the years prior to my joining as CEO in 2020. B:Side, or Colorado Lending Source as it was then known, was an incredibly strong organization staffed by equally strong individuals. The level of technical expertise, particularly at the top of the organization, was staggering. It was also oppressive. All decisions ran through the top office, and that had two unintended effects. First, it robbed the team of agency. Second, and more importantly, it never allowed confidence to take hold in the staff. That lack of confidence led to a lack of agency, and that led to an underutilization of our team’s immense talent. Building that confidence and establishing a culture of agency is an ongoing process.

Being needed can be gratifying, particularly when the need appears to confirm the very competence that earned someone the right to lead. We can become attached to a role that gives us immediate evidence of our usefulness, while helping another person acquire the same judgment makes our contribution harder to see. That’s an awkward incentive.

From the professor’s chair, the problem becomes one of formation. Watching someone produce the right answer gives a learner access to the result, with only partial access to the work that produced it. A student can follow an explanation and still struggle to recognize the same problem when its surface details change. A manager can understand the founder’s decision after the fact and still lack the practice needed to make the next one.

That practice includes choosing which information deserves attention, deciding what remains uncertain, and committing to an action whose consequences the learner will have to examine. When the teacher resolves every ambiguity first, the student receives a cleaner problem than the one independent judgment requires. Business can repeat that pattern without anyone intending to teach it. The founder removes the difficult part, and then wonders why nobody else has learned to handle difficulty.

Our response to mistakes completes the lesson, especially when a manager makes a reasonable decision within agreed limits and the owner publicly replaces it with a personal preference. People pay attention to those moments because their standing depends on understanding the actual rules. A title that promises ownership will carry less weight than the last visible consequence of using it. Eventually, the manager may learn to seek permission before making even ordinary decisions, because the stated boundaries no longer predict what the owner will allow.

This places a real obligation on the leader who wants the business to endure. The opportunity to practice judgment has to survive the leader’s preference for a faster answer. There will be times when intervention is necessary, and those decisions should carry an explanation that helps the next person improve. Over time, the evidence of progress should include fewer problems that require the founder’s particular presence to resolve.

Pass Along the Reason

For the owner, the first useful distinction is between a standard and a preference. A standard protects something the business owes its customers or its people. A preference describes how the current leader likes work to happen. The two can become so closely associated that a successor has to copy the founder’s habits simply to demonstrate loyalty.

The difference matters. A commitment to telling customers the truth must survive a change in leadership, even when the next leader finds a better way to communicate it. The founder’s preferred meeting time carries a different kind of authority, which is why I should be especially suspicious when my principles happen to require everybody to do things my way (a remarkable coincidence).

Passing along the reason means preserving the circumstances around an important decision, including what was uncertain and what would have changed the answer. Someone learning from it also needs to understand which risk the decision-maker considered most serious. Those details allow a future manager to adapt the lesson when circumstances change, while a record containing only the final instruction offers little help in understanding.

A late-payment decision gives this work a concrete shape: the useful record would explain what the customer’s payment history suggested, how much room the business had to wait, and why further credit was allowed or refused. The next manager could compare those reasons with the next customer’s circumstances and reach a decision they could defend. That is much more to inherit than a record that the founder approved an exception.

For someone buying a business, this suggests looking closely at where explanations live. Customer relationships deserve attention alongside the records describing them, and the people handling exceptions deserve attention alongside the written procedures. The seller’s continued availability can help a transition, provided that availability is used to build the next team’s capability. Every question permanently reserved for the former owner extends the dependence the transaction needs to address.

For a manager preparing to assume more responsibility, the work can begin with a recommendation that gives an experienced leader something to examine: a proposed decision, the reasons behind it, and a clear account of uncertainty. That exchange makes the differences in judgment visible enough to discuss. Eventually, the manager also needs permission to act on what has been learned.

Practice Leaving Before You Leave

Choose a decision that can move. Identify a recurring decision currently reaching the founder that another person could handle with preparation. Start where mistakes can be detected and repaired before they cause serious harm, with enough substance in the task for the person to learn something consequential. Name the decision owner.

Explain what deserves protection. Work through a real past case together, including the evidence that mattered and the uncertainty that remained. State the boundaries the new owner must respect, the information they can use, and the conditions requiring escalation. Ask them to explain how those boundaries would apply if the facts changed.

Make the authority real. Tell the people affected who now owns the decision and what that person is authorized to do. Allow decisions within those boundaries to stand, including choices you would have made differently. If intervention becomes necessary, explain the specific risk or limit that required it so the team can understand what remains theirs to decide.

Review the reasoning after the result. Set a time to compare the decision-maker’s expectations with what happened, examining an unfortunate result from a sound process differently from a lucky result produced by carelessness. Use that conversation to improve the next decision, and record any change in the standard or the person’s authority.

Repeat the test with more distance. Let the new owner handle the next case before supplying an answer, and observe where the work still returns to you. Bring that person into the customer relationships their authority will depend on, then give them room to earn trust through their own conduct. As their judgment develops, expand the responsibility and ask them to begin teaching someone else.

The Next Call Belongs to Someone Else

There is no clean formula for how much error a business should accept while another person learns to lead it. Customers bear real consequences, and the tension remains even after the boundaries are clear and the preparation is careful, because another person’s learning can still cost them something. Responsibility includes making those choices without pretending the costs disappear.

But the difficult customer will eventually call when the founder is unavailable, and by then another person needs to have practiced hearing the complaint, judging what the business owes, and making a promise the team can keep. The quality of that response will reflect what the founder helped others learn while there was still time to teach it. The customer may never know how much work made that conversation possible.

For a leader accustomed to receiving the call, that can be an unfamiliar kind of success. The business keeps its word, someone else carries the responsibility, and the day continues without a rescue. Its history has become something another person can carry forward.

Let someone else make the call.

Share

Building a team that can lead without you takes daily practice. I explore that work in my book, The B:Side Way, including how to earn trust, develop people, and give them real ownership.

Treasury Secretary Scott Bessent asked that question from a podium on Monday morning, and he meant it to be reassuring. He was explaining why the foreign banks and trading houses targeted by his new sanctions campaign will get a grace period before the penalties land: secondary sanctions are so powerful, he reasoned, that careless use could damage the machinery they run on. Washington will be patient, in other words, because the alternative is unthinkable.

I’ve spent enough years around credit markets to know that sentences like that one deserve attention. Gold bugs talk about blowing up the financial system, and so do the anonymous doom accounts that sell silver in their bios. When the steward of the world’s reserve currency starts talking about it, even to wave the idea away, something has shifted. The possibility has entered the official vocabulary, and I’ve found that vocabulary tends to lead policy by a few years.

The question was only half of Monday’s news. Bessent spent the press conference launching what Treasury calls Operation Economic Outcast, an “economic D-Day” intended, in his words, “to sever every economic lifeline that sustains this tyrannical regime until Tehran stands alone.” Nearly sixty entities, individuals, and vessels were designated in a single morning. Five sectors Iran uses to earn and move money abroad (digital assets, technology, gold, aviation, and shipping) now carry secondary-sanctions risk for anyone still doing business there. Every country received a defined timeline to shut down the Iranian networks inside its borders, along with a reminder that “we do not have infinite patience.” Any institution caught laundering money for Tehran “will be removed from the U.S. dollar system.” Asked whether Chinese banks were exempt, Bessent said no one is above the reach of U.S. sanctions.

In the same session, he took questions about the Treasury’s other project this month: doubling the buybacks of its own long-term bonds to at least $4 billion per operation, beginning September 9. He repeated his confidence that yields will keep falling.

Each move has a defensible rationale, and I want to be fair to both before I criticize either. However, taken together, they describe a government leaning its full weight on instruments that run on trust while quieting the one instrument that reports whether the trust is holding. That combination is the part I can’t let go of.

Listen to the Disciplinarian

Ninety minutes after Monday’s close, Stanley Druckenmiller published an op-ed in the Wall Street Journal titled “Let the Bond Market Speak.” His premise comes from five decades of trading: “Markets aggregate information no committee possesses, and prices are how that information reaches decision makers.” From there he arrives at two sentences worth memorizing: “The long-term Treasury yield is the most important price in the world. It is also the only fiscal disciplinarian the U.S. has left.”

I believe he’s right, and it helps to remember how we got here. The other disciplinarians retired one at a time: the gold window closed in 1971, the balanced-budget wing of Congress faded into irrelevance, and the rating agencies downgraded America in 2011 only to learn that nobody flinched. The Federal Reserve spent most of two decades as the market’s largest buyer, which makes it an awkward scold. What remains is the auction, where the United States shows up week after week and asks strangers what its promises are worth. The long yield is the last number Washington cannot lobby.

That’s why the buyback program deserves more scrutiny than it’s getting. These operations were designed as boring liquidity plumbing, a few billion dollars of housekeeping that keeps older bonds tradable. Ten days ago, on an off-cycle Wednesday, Treasury doubled them to at least $4 billion per operation, one day after a selloff pushed the 30-year to 5.34 percent and a week after a long-bond auction cleared at yields last seen in 2001. I walked through those mechanics in last week’s essay, and I argued back in May that the bond market is the only price in the economy no one can talk down. Treasury appears to have reached the same conclusion, because it has stopped talking and started preparing to buy. Druckenmiller’s judgment is blunt: “Every basis point of artificial yield suppression is a subsidy to procrastination.” His prescription is blunter still: return buybacks to “small, scheduled, off-the-run liquidity operations announced at quarterly refundings, never off-cycle responses to yield levels,” term out the debt honestly, and pay the price the market sets. “If the 30-year must trade at 5.5% to clear, that isn’t a crisis. It is an invoice.”

Subscribe now

Remember, too, that Bessent set his own grading scale. He took office telling investors to judge this administration by where the 10-year trades, with deficits gliding toward 3 percent of GDP by the end of the term. More often than not, the market takes a man at his word.

If the disciplinarian sounds theoretical, consider what happened in London. In September 2022, a new British government announced the largest unfunded tax cuts in half a century. Within seventy-two hours, 30-year gilt yields had spiked more than a full percentage point, pension funds running leveraged hedges were facing margin calls, and the Bank of England had to step in with emergency purchases to keep the spiral from consuming the country’s pension system. The chancellor lost his job within three weeks of that budget. The prime minister resigned within a month of it, after forty-nine days in office. Here is the part I find instructive: it worked. The policy died, yields settled, and every British budget since has been written with the market reading over the chancellor’s shoulder. An invoice that arrives early, while there’s still time to pay it, is a form of mercy.

Count the Full Cost of the Weapon

Now consider the other play. The dollar system’s power rests on a simple network reality: nearly every institution on earth that moves serious money eventually touches a bank that needs access to New York. Exclusion from dollar clearing amounts to financial suffocation, enforceable without a single soldier, which is why sanctions have become Washington’s favorite tool. They’re cheaper than a carrier group and faster than a blockade, and six months into a war that has kept the Strait of Hormuz closed, I understand the appeal of an economic endgame. Ending this war with banks instead of bombers is the more humane ambition, and I hope it succeeds.

The trouble is that this weapon carries a running cost, and the bill doesn’t arrive right away. We’ve seen the pattern before. In August 1971, Nixon closed the gold window. Three months later, at a G-10 summit in Rome, Treasury Secretary John Connally told the assembled European finance ministers, “The dollar is our currency, but it’s your problem.” The line has been quoted for fifty years as swagger, and in the moment it probably felt like strength, because nobody stormed out of the room. Instead, the customers stayed in the club and began hedging it. Gold went from $35 an ounce to $850 inside a decade. Europe spent the next thirty years building a currency of its own, in meaningful part so the dollar would be less of its problem. The adaptations were slow, deniable, and compounding, which is exactly what made them easy to ignore.

The modern version is already in the data. When Washington froze Russia’s central bank reserves in 2022, every finance ministry on earth learned that reserves are conditional. Central banks have been buying gold at the fastest pace in half a century ever since. Now look at what Monday’s order polices: gold, digital assets, shipping. When a sanctions regime has to guard the exits, the exits are getting traffic.

The French thinker René Girard spent his last years studying how systems of containment fail, and his conclusion has proven useful far beyond the theology where he started. Every containment system, he argued, restrains a force by spending a measured dose of the same force: the state ends private vengeance by monopolizing vengeance, deterrence prevents annihilation by promising it, and a lender prevents losses by refusing credit. Because the system pays in the currency it polices, its price rises as its power fades, and the late stage looks the same everywhere: spending more to achieve less, with each escalation justified by the failure of the one before it. Girard even had a phrase for the desperate phase, when the mechanism starts reaching for “ever more precious victims.” Shadow-fleet brokers in the Gulf were cheap victims, and the banks of Iran’s customers are more precious. The logic now points toward institutions so large that punishing them would amputate part of the network the weapon runs on. When Bessent asks why he’d want to blow up the global financial system, he’s acknowledging that arithmetic out loud.

The buybacks follow the same curve. Two billion dollars of housekeeping became four billion of “liquidity support,” with the door open to more. (”We haven’t bought a single bond yet,” Bessent said Monday. “We will see on September 9th.”) Each escalation will be individually defensible, which is exactly how the pattern works.

Here’s the principle I keep coming back to: trust is the only reserve asset, and everything else is denominated in it. The sanctions spend the dollar’s neutrality to buy compliance. The buybacks spend the yield curve’s honesty to buy time. Both purchases may prove worth it. I simply want us to be honest that something real is being spent.

And before we get too superior about Washington, we should admit that we run the same trade in our own organizations. Every leader has a bond market: the dashboard that keeps printing a number we don’t like, the customer feedback we reroute, the banker who asks uncomfortable questions at renewal. The temptation to manage the messenger instead of the message is universal, and I’ve felt it myself more times than I’d like to admit. I feel a version of it right now. At B:Side, the SBA 504 loans we fund price off the long end of this exact curve, so a lower 10-year means a lower payment for every manufacturer we help finance this fall, and the CEO in me quietly wants the September 9 operation to work. The professor in me stands in front of students each week and tells them that prices are information, which would make a subsidized price a redacted document. I want the discount and I want the truth, and I haven’t fully reconciled the two. I’ve learned that this kind of tension usually means you’re looking at the real issue.

What I’d Do About It

None of us can reopen the strait or balance the federal budget, so let’s focus on what an owner or operator can actually control. I’d concentrate on three disciplines.

Plan against the clearing price. If the long bond needs 5.5 percent to find real buyers, underwrite your expansion at 5.5 percent, and treat any buyback-driven dip as a coupon rather than a quote. Enjoy the discount whenever it appears, but never build the plan on it.

Term out your debt honestly. Druckenmiller’s advice to the government applies at company scale. If you have a maturity coming due in the next eighteen months, start the conversation now, while it’s still a conversation, and fix every rate you can defend. When an artificial window opens, refinance through it. That may be the one genuine gift these operations hand you.

Keep your powder dry. A stretch like this rewards liquidity over efficiency. Carry more cash than the textbooks recommend, because cash buys time, and time is what you’ll want if the quiet ends abruptly. I’ve watched enough credit cycles to know that recognizing a vulnerability can take years while the actual break takes days, and the break, when it comes, tends to arrive on some ordinary Tuesday without a memo.

While you’re at it, watch a few public tells over the next month: the size of the September 9 operation and whether $4 billion holds; the long-bond auctions in mid-September; the name of the financial institution Bessent promised to sanction by Friday, and, more to the point, its passport; and whether gold and the dollar keep rising together, because that pair moving in tandem means the world is paying its dues to the club while pricing the exits.

Bessent asked why he would want to blow up the global financial system. He wouldn’t, and I don’t believe he will. Systems like this rarely end in explosions; they erode through substitution, a workaround here and a gold vault there, while every official statement stays technically true. The way to stop that erosion has been available all along, and Druckenmiller named it on Monday: let the market set the price, pay the invoice, and address the primary deficit that keeps generating the bill. Whether Washington chooses that path is beyond your control and mine. After all, our job is smaller and more manageable: read honest prices, borrow honestly against them, and keep enough powder dry to survive the day the quiet ends. It’s an undramatic strategy, and in my experience the undramatic ones are what carry you through. The bond market is still speaking. The wise move, in your business and mine, is to keep listening.

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.

I’ll be the first to admit that I have a credibility problem on this subject. I run bearish by temperament, and by my own count I’ve predicted eight of the last two recessions. My own team applies a healthy discount to my gloomier pronouncements, and they’re right to do so. That’s exactly why I want to be clear about what I’m doing here: I’m not making a prediction. I’m reading numbers that have already been published.

Here’s what those numbers say. Headline inflation touched 4.2 percent this spring and still sits near 3.4 percent, well above target after months of grinding deceleration. Growth cooled to roughly 1.5 percent annualized in the second quarter. Brent crude has traded at $102, at $72, and back near $90 since February, and the fuel surcharge has worked its way into the price of everything that moves on a truck. And this week, in a move I’ll come back to, the United States Treasury started doubling the buybacks of its own long-term bonds. Slowing growth and rising prices, arriving together, courtesy of an oil shock: that’s the textbook definition of stagflation. We already have a mild case. The open question is duration, and duration is exactly the kind of question you want to face with your eyes open.

I learned a long time ago that problems don’t just disappear when we decline to look at them. In fact, they tend to grow in severity the longer they’re ignored. That’s doubly true for problems without an official referee. The committee that dates recessions typically rules about a year after the fact (which counts as prompt by the standards of official candor), and there is no committee for stagflation at all. Anyone waiting for somebody in Washington to make it official will receive the diagnosis as history.

What Makes Stagflation Different

To understand why I take even a mild case seriously, it helps to understand what makes stagflation unusual. In a normal economic cycle, growth and inflation move together. The economy runs hot and prices rise; the economy cools and prices ease. That pattern is what makes the Federal Reserve’s job possible, because its one real lever works on both problems at once: raise rates into the boom, cut them into the bust.

A supply shock breaks that machine. When something like an oil disruption hits from outside, it pushes prices up and activity down at the same time, and the single lever suddenly faces two fires. Raise rates to fight inflation and you deepen the slowdown. Cut rates to cushion the slowdown and you feed inflation. Every option is a trade-off between two mandates that now point in opposite directions, and the dilemma itself is the disease.

By that definition, the Fed is already living with it. Markets spent part of the spring handicapping a rate hike into a cooling economy, because oil kept threatening to reignite the price level. The 10-year Treasury sits near 4.7 percent, Tuesday’s sell-off pushed the 30-year to 5.34, and last week’s 30-year auction cleared at the highest yield since 2001, which tells you bond investors have stopped treating inflation risk as theoretical. You can imagine the mood in those committee meetings.

How We Got Here, and Why It Lingers

The trigger is no mystery. The American and Israeli campaign against Iran opened in the last days of February, and Iran answered by closing most of the Strait of Hormuz, the chokepoint that carries roughly a fifth of the world’s oil. An energy shock behaves like a tax with perfect coverage: it raises the cost of producing, shipping, and driving while draining the discretionary spending that would otherwise become someone else’s revenue. Brent spiked toward $102 in the spring escalation, and American gasoline brushed $4.05 a gallon at the worst of it. Prices fell to $72 in late June when a memorandum of understanding promised a path out, then climbed back through the high 80s once the memorandum’s 60-day window expired with nothing durable behind it. BNP Paribas modeled the shock at roughly 0.4 points off advanced-economy growth and 1.1 points added to inflation, and that’s their central scenario, the one that assumes the conflict cools from here.

I wouldn’t count on it cooling, and an old Cold War strategist explains why. Thomas Schelling drew a famous distinction in Arms and Influence between deterrence and compellence. Deterrence asks your adversary to keep doing nothing, and it can wait forever. Compellence demands a visible act of compliance by a deadline, and it has to stay in motion until it gets one. Schelling’s rule was that it is far easier to deter than to compel, and Hormuz is now a compellence problem for both sides. Tehran wants sanctions relief and reparations before the strait reopens. Washington wants the strait open before it discusses anything else. When two parties each insist that the other move first, you get exactly what the oil market has priced all summer: a long, expensive grind.

Small-business owners have been riding that grind in real time. The NFIB optimism index fell to 95.8 in March, its low for the cycle, during the exact weeks when owners reported absorbing fuel and input costs they couldn’t fully pass along. By July the index had rebounded to 99.8, its best reading in nearly a year, as oil calmed down. I would love to read that recovery as resilience. Unfortunately, I’ve been around long enough to suspect it might also be what Galbraith called the extreme brevity of the financial memory. The truth is I can’t tell which one it is, and that uncertainty bothers me more than the March number did.

The Case for Calm

Plenty of smart people think the worriers are getting ahead of themselves, and their case deserves a full airing. Eugenio Aleman at Raymond James puts the odds of a genuine stagflationary episode at “very low.” Gregory Daco at EY argues that everything turns on duration, and that the baseline outcome is a temporary inflation bump. The Council on Foreign Relations concluded this spring that a repeat of the 1965-to-1982 era is unlikely, and the structural arguments hold up well. America now produces and exports the oil it once imported, which cushions the blow and hands parts of Texas and North Dakota a boom inside the squeeze. The wage-indexation clauses that hard-wired 1970s pay to 1970s prices are mostly gone. The Fed has four decades of inflation-fighting credibility that Arthur Burns never enjoyed. The economy runs on services and software, unemployment sits near 4.2 percent, and the AI investment boom, whatever its excesses, is a productivity tailwind with no 1974 equivalent. On the history, the calm camp wins, and I’m happy to concede the point. I don’t expect gas lines, price controls, or double-digit inflation prints.

My problem with the comfortable conclusion is simple: it assumes the damage has to arrive the way it arrived last time, through the price level. We’ve spent the past fifty years building a different set of vulnerabilities, and this week the biggest one stepped into plain view.

When Plumbing Becomes Policy

This week, the Treasury Department announced that it will at least double its long-term buyback operations, from a $2 billion cap to a minimum of $4 billion per operation across the 10-to-30-year sectors, beginning September 9. The official language calls it “liquidity support,” and the official rationale points to the “significant volume of high-quality offers” Treasury keeps receiving in the long end, which is a polite way of saying that a great many people would like to hand their long bonds back to the government. The announcement came the day after that sell-off took the 30-year to 5.34 percent, and a week after the auction that cleared at 2001 levels. Yields fell on the news, stocks cheered, and gold jumped, because everyone worked out the translation at the same time.

Here’s the translation. For a decade, Treasury buybacks were boring liquidity plumbing, a few billion dollars of housekeeping that kept older bonds tradable. Conducted at the highest long-term yields in a generation, the day after a sell-off, they become something else entirely: the world’s largest borrower supporting the price of its own debt because private demand keeps asking for a better deal. Some of that missing demand has a surprising address. The AI buildout is being financed with hundreds of billions of dollars in new corporate bonds, and every buyer who reaches for that richer yield is a buyer who didn’t show up at the Treasury auction. Add war-driven inflation fears and a federal debt load that grows in every season, and the long end of the curve has become the one market Washington can least afford to leave alone.

I understand the temptation, and I want to be plain about where it leads. The government funds these purchases by issuing even more short-term debt, which means it’s swapping long promises for short ones at the exact moment inflation makes short promises expensive to keep. Easing financial conditions while consumer prices run a full point above target adds fuel to the inflationary fire, and that’s worse than illogical: it’s a catastrophic mistake, and the long end is where it will implode. A support operation announces the absence of real demand more loudly than any failed auction could, and the investors who hear that announcement respond by demanding more compensation, which invites a bigger operation, which sends a louder signal. We’ve run this experiment before. The Fed pegged long-term yields through the 1940s while inflation ran into double digits, and it took the Treasury-Fed Accord of 1951 to shut the arrangement down before it consumed the central bank’s credibility. The peg always looks free at the start. It never is.

Don’t Ignore Obvious Problems

A decade ago, in my first book, I wrote a chapter called “Don’t Ignore Obvious Problems.” The case study was Greece, where the math simply didn’t work, and where everyone in charge preferred to kick the can down the road until the problem grew beyond managing. I’ve been thinking about that chapter a lot lately, because the obvious problem in front of us today is sitting on the balance sheets of America’s regional and community banks.

Subscribe now

The numbers deserve to be read slowly. Commercial real estate makes up 44 to 48 percent of loans at regional banks, against 13 to 19 percent at the money-center giants. The FDIC’s 2026 Risk Review puts the median CRE concentration at banks between $1 billion and $100 billion around 300 percent of Tier 1 capital, the level that triggers heightened supervisory attention, and hundreds of community institutions live at or above that line as a normal feature of their business model. Somewhere between $1.5 and $2 trillion of commercial real estate debt matures through the end of 2026, and every one of those loans arrives at the same fork: refinance at today’s rates or sell the building at whatever price the new math supports. Neither branch is comfortable for the lender. Klaros Group screened roughly 4,000 banks and found 282 carrying both heavy CRE books and deep unrealized losses on their securities portfolios, and the office sector shows what trouble looks like when it arrives: Manhattan office delinquencies rose more than 1,000 percent in the twelve months through January 2024, and the vacancies behind that number have proven stubbornly durable. Notice, too, that those unrealized securities losses are marked against exactly the long-dated bonds the Treasury is now propping up. If the long end gives way the way I fear it might, the refinancing wall and the bond losses arrive together.

Two old economists explain why I watch this so closely. Hyman Minsky taught that long stretches of calm quietly move borrowers from loans their income can retire toward loans that survive only by being refinanced; stability, he argued, reprices recklessness as prudence, one renewal at a time. Charles Kindleberger, whose Manias, Panics, and Crashes tracks four centuries of these episodes, added the timing: stock manias and property manias inflate together, but they deflate on different clocks. Equities crash fast and in public. Property declines slowly, over years, because the debt is long-dated and nobody can call it overnight. A refinancing calendar is the slowest clock in finance, and ours happens to strike between now and the end of next year.

This is where stagflation stops being an abstraction and becomes a transmission mechanism. The inflation half keeps rates high, which raises banks’ funding costs and their borrowers’ refinancing burden in the same motion. The stagnation half softens the tenants, the rents, and the local businesses that keep their deposits at those same banks. A stressed bank does the rational thing and tightens, the withheld credit comes straight out of the working capital of small businesses already paying the fuel surcharge, and the resulting slowdown circles back into the bank’s own loan book. I watch this loop from close range. At B:Side, our SBA lending runs through exactly these community banks and into exactly these businesses, and I’ve learned that renewal conversations tell you more than any index does. When bankers start slowing down, we feel it months before it prints.

Who Actually Pays

A stagflationary economy is an averaging machine, and the averages flatter it. Energy producers are booming. Large firms hedge their fuel, ladder their debt, and pass costs along. Roughly half of American consumer spending now comes from the top tenth of households, which props up the aggregate numbers while the median family absorbs the shock. Energy and food take a far bigger share of a small paycheck than a large one, real wages have lagged the sticky part of inflation, and layoff risk concentrates in the energy-sensitive sectors least able to carry idle payroll. I see the same split from the front of a classroom. My students read about 4 percent unemployment and a record stock market, then go home and watch their parents cut back. The average is doing fine, the median is tired, and that gap, more than any single statistic, is what stagflation actually means.

What I’d Do Right Now

If you own a business, start with pricing. March taught us that absorbing input costs is a strategy with a deadline, so reprice deliberately and early, and explain the increase to your customers like the adults they are. Next, measure your energy exposure per unit of whatever you sell, because the strait can close again faster than you can rewrite a contract. If you have a loan maturing within the next eighteen months, call your banker this month, while it’s still a conversation rather than a deadline, and fix every rate you can defend; if Washington’s buybacks hold the long end down for a few months while you do it, accept the gift. Get to know your bank the way it knows you: its concentration, its appetite, who actually holds your note. Finally, hold more cash than the efficiency textbooks recommend. Cash buys time, and time is the scarcest asset in a tightening.

If you lead a team, tell them the truth about the year. Your people are living the median experience while the headlines describe the average one, and that gap erodes trust faster than any bad news will. Budget for a range on energy prices instead of a point estimate, and watch the indicators that will actually settle this: credit spreads, lending standards, renewal terms, and the size of the Treasury’s next buyback. Most of them are dull. All of them are decisive.

The word itself carries a useful lesson. “Stagflation” was coined in 1965 by a British politician named Iain Macleod, who looked at slowing growth and rising prices and told Parliament his country faced “the worst of both worlds.” The name arrived before the fight did, and that’s the right order, because nobody can fight a condition they refuse to name. Our version is milder than his, at least so far, but it’s here. If I’m wrong, and the strait reopens, prices settle, and the buybacks fade back into obscurity, you’ll be left holding extra cash, longer-dated debt, and a banker who knows your name; there are worse fates. If I’m right, those same preparations will be the difference between a hard year and a dangerous one. Either way the moves are identical, and the time to make them is now. After all, problems don’t disappear just because we decline to name them. More often than not, they grow. Things are going to get interesting, that’s for sure.

Share

That is not false modesty. It is arithmetic. I can see the open rates. I know the difference between a subscriber and a reader. I have written four books, with a fifth on the way. I write this newsletter, and over the years I have written for Forbes, Entrepreneur, and The Wall Street Journal. Which is to say I have spent a meaningful portion of my adult life producing exactly the kind of writing that people are reading less and less of, and I have the anecdotes to prove it.

I have lost count of the people who have congratulated me on a book, quoted its title back to me, or told me they bought a copy, while making it gently clear in the same conversation that they never actually read the thing. Friends ask me for “the gist” of an essay I spent a week thinking through. Colleagues respond to a long piece within ninety seconds of it landing in their inbox, which tells me exactly how much of it they absorbed. My students at ASU, some of the brightest young people I have ever worked with, will admit without much embarrassment that they ran the assigned reading through an AI summarizer and skimmed the output on their phone.

None of this offends me. I understand the forces at work, and I will get to them in a moment. But I recently read a piece in The Atlantic by Rose Horowitch called “The End of Reading Is Here,” and it has been rattling around in my head ever since. It crystallized something I have felt for years as someone who writes for a living audience that is quietly disappearing. The piece argues that the age of reading, the roughly five centuries in which ordinary people routinely engaged with long, complex written works, may turn out to be a brief interlude in human history rather than a permanent achievement.

I think she is largely right about the diagnosis. I want to talk about what we do with it.

The Fire Is Not What Killed the Library

Horowitch opens her essay with the Library of Alexandria, and the detail she surfaces is the one worth holding onto. For centuries, the popular story was that the library died by fire: Julius Caesar’s siege in 48 B.C.E., or the mob that sacked the temple four hundred years later. Dramatic, external, sudden. A tragedy inflicted from outside.

Contemporary historians mostly reject that story. The library died of negligence. Papyrus scrolls do not keep themselves; humidity, mice, and insects eat them slowly, and scribes had to continually recopy the collection just to hold ground against decay. At some point, the cost of maintaining the library exceeded the will to maintain it. The classics scholar Roger Bagnall put it in terms I have not been able to shake: it is not that the loss of the library caused a dark age. The fact that the library was allowed to die showed that the dark age had already arrived.

Think about that for a moment. The greatest collection of knowledge in the ancient world was not destroyed by an enemy. It was simply not renewed by its friends.

That is the correct frame for what is happening to reading now, because nobody is burning anything. The numbers Horowitch assembles are stark. Fewer than half of American adults read a book of any kind in 2022. The share of Americans who read for pleasure on a given day fell from 28 percent in 2004 to 16 percent in 2023. More Americans placed a bet last year than read a novel. Reading stamina among students has collapsed to the point where most middle and high school English teachers now assign somewhere between zero and four full books a year, and college professors report teaching students how to comprehend a text at all before they can teach anything else.

And here is the detail that should concern anyone who cares about institutions: the decline cuts across every demographic. Retirees, women, college graduates, the groups that always read the most, are all reading dramatically less. This is not a story about one generation or one class. It is a story about the whole culture quietly deciding, one evening at a time, that the maintenance is no longer worth the effort.

We Are Not Illiterate. We Are Postliterate.

The precision of that word matters. Americans are probably consuming more written words than at any point in history: texts, posts, captions, headlines, chat threads, notifications. What has evaporated is not decoding. It is sustained attention to long, complex, structured argument. The cognitive neuroscientist Maryanne Wolf calls this the loss of deep reading, the higher-order work of comprehension, inference, and synthesis. You can still read the words. You are just losing the ability to hold a long thought in your mind while an author builds it.

I see the evidence everywhere, and if you are honest with yourself, so do you. Horowitch cites a study in which English majors, students who chose to study literature, were asked to read the opening paragraphs of Bleak House. A quarter of them interpreted Dickens’s figures of speech literally and concluded that dinosaurs were walking the streets of nineteenth-century London. Only 5 percent came away with an accurate understanding of what they had read. These students had the entire internet available to look up anything they did not understand. They did not bother.

Now, I know what some of you are thinking. Every generation panics about the new medium. Jefferson worried that novels would ruin young women. In 1900, The Atlantic itself published a lament that the newspaper was destroying Americans’ capacity for serious thought. The doomsayers are always the people most invested in the old way. Fair enough. That skepticism is healthy, and Horowitch takes it seriously; she quotes a Harvard literature professor who spent his whole career dismissing exactly these panics, right up until the last few years, when he stopped dismissing them. What changed his mind is what should change yours: this time the data all point the same direction, the decline is accelerating rather than plateauing, and there is no plausible mechanism for a return. Television competed with books for evenings. The phone competes with books for every conscious minute.

What Atrophies When Nobody Reads

I watch what happens inside organizations when reading atrophies. Decisions get made off summaries of summaries. A carefully constructed memo gets skimmed, and the nuance that took hours to get right, the qualifications, the second-order risks, the reasoning behind the recommendation, evaporates somewhere between the second paragraph and the reply button. People form strong opinions about documents they have not read, and because everyone is doing it, nobody flags it. At B:Side, we work in small business lending, a world of credit memos, loan agreements, and regulatory guidance where the substance lives in the details. In our industry, the person who actually reads the document has a quiet, compounding advantage over the person who read the executive summary. I suspect that is true in yours as well.

The classroom version of this is more painful because the stakes are more personal. My students are not lazy, and I want to be clear about that. They are rational actors responding to the environment they were raised in. They have never known a world without infinite short-form video, and they have been trained by every institution around them, including their schools, to treat text as an inefficient container for information: something to be compressed, summarized, and extracted from, rather than experienced. Horowitch quotes a Harvard administrator explaining that some students now genuinely believe professors who assign whole books are arbitrarily withholding information by forcing them through a needlessly difficult medium. I have heard versions of that sentiment in my own classroom, delivered without irony, by students I admire.

And here is where my own experience as a writer stops being an anecdote and becomes evidence. When people engage with my books through their titles, or my essays through their subject lines, they are not being rude. They are being modern. The gist, the summary, the takeaway: this is now the default unit of intellectual exchange. What gets lost is precisely the thing long-form writing exists to carry, which is not information at all. It is reasoning. A summary can tell you what I concluded. Only the essay can show you how I got there, where I almost went instead, and what it would take to change my mind. Strip that away and all that remains is a stranger’s opinion, which you are free to ignore, and probably should.

Why Reading Still Matters

So why keep at it? Why read whole books in a world that has politely moved on?

The first reason is cognitive, and it is not sentimental. Reading is a workout for the attention span, and attention is upstream of everything else you do. The research Horowitch surveys is consistent: deep reading builds the capacity for inference, synthesis, and holding complex ideas in mind across time. Watching video, for all the information it carries, is passive by comparison; the frames keep moving whether or not you have understood anything. The less you read, the harder reading becomes, and the harder reading becomes, the harder thinking becomes. That loop runs in both directions, which is the good news. It is a muscle, and muscles respond to training at any age.

The second reason is that reading is becoming scarce, and scarce capabilities command a premium. Just 20 percent of adults now account for more than 80 percent of the books read in America. A historian of reading quoted in the piece compares it to stamp collecting or growing orchids: a niche hobby. I would put it differently. When a foundational skill becomes a niche practice, the niche inherits the advantage. The person who can sit with a forty-page credit agreement, a dense regulatory proposal, or a serious book on the history of financial crises, and actually comprehend it, is no longer competing with everyone. They are competing with the one person in five who still can. I have written before about the evaporating bottom rungs of the career ladder, and this is the same dynamic in another costume. In an economy where AI generates infinite plausible text, the ability to evaluate text, to notice what is missing, what is wrong, and what is manipulative, becomes one of the few genuinely defensible human skills.

The third reason is older and harder to quantify. Books are how the dead talk to the living. They are the vertical transmission of culture, wisdom passed down across generations, and that channel is being replaced by a horizontal one in which twenty-two-year-olds learn primarily from other twenty-two-year-olds, at 2x speed. James Baldwin said that Dostoevsky and Dickens taught him that the things which tormented him most were the very things that connected him with everyone who had ever been alive. I felt that as a young man reading Marcus Aurelius for the first time, discovering that a Roman emperor lay awake at night wrestling with the same doubts about duty and mortality and self-command that I did. No feed will ever give you that, because the feed is optimized for this hour, and the books that last were written for the centuries.

The Writer Gets More Than the Reader

Now for the part of this argument that I believe most strongly, and that the postliterate world understands least.

Even if nobody read a word I wrote, I would keep writing, because writing is not primarily an act of communication. It is an act of cognition. The writer benefits before the reader ever arrives, and more than the reader ever will.

Here’s the truth: you do not actually know what you think about something until you have tried to write it down. Before the writing, what you have is a comfortable fog, a set of intuitions and borrowed phrases that feel like a position. Writing is the process of running that fog through a compressor. Every vague notion has to become a sentence, every sentence has to survive contact with the next one, and the weak links do not survive. Cal Newport makes this point in Horowitch’s essay: writing forces thought into an orderly, linear form and exposes flabby reasoning like nothing else. Orwell compared writing a book to a long bout of a painful illness, and he kept doing it anyway, because the pain is the point. The difficulty of writing is not friction on the way to the product. The difficulty is the product. The struggle is where the thinking happens.

The best evidence for this claim is also the best book I know. Meditations was never meant to be read by anyone. Marcus Aurelius wrote it to himself, at night, in a military camp on a frozen frontier, with no audience and no expectation of publication. It is literally a private notebook, and it has survived for nearly two thousand years precisely because writing without an audience is the most honest thinking a person can do. The most enduring work of practical philosophy in Western history has an intended readership of one. If that does not settle the question of whether writing is worth doing when nobody is reading, I am not sure what would.

This is also why I am wary of the frictionless promise of AI writing, even as I run a company that embraces AI aggressively and teach my students to do the same. The early research Horowitch cites matches what I see: when people outsource the writing, the output improves and the thinking degrades. Students who studied with AI performed worse on tests that demanded reflection, not because the machine gave them wrong answers, but because it did their struggling for them. If writing is how the writer learns, then a tool that eliminates the writing eliminates the learning, no matter how good the prose looks. Use AI to critique your draft, pressure-test your argument, and find what you missed. Do not use it to spare yourself the labor of the first draft, because that labor was never overhead. It was the whole exercise.

What This Means in Practice

I try to avoid diagnosis without prescription, so let me be concrete about what I actually recommend, whether you are running a company, building a career, or raising the next generation of readers.

1. Read whole books, slowly, on paper if you can. Not excerpts, not summaries, not the podcast where the author repeats the introduction. The compounding benefit lives in the sustained middle chapters that the summary skips. Start with thirty minutes and accept that it will feel difficult at first. It is supposed to. That is the muscle rebuilding.

2. Write regularly, whether or not anyone reads it. A journal, a weekly memo to yourself, a letter you never send. Judge the practice by the clarity it produces in you, not by the audience it attracts. Marcus Aurelius had no subscribers.

3. Do the first draft yourself. Let AI sharpen your thinking after you have done the thinking. The moment you delegate the blank page, you have delegated the learning, and the learning was the only part that was ever yours.

4. If you lead people, read what you decide on, and be seen doing it. The reading culture of an organization is set at the top, the same way its emotional ceiling is. When the boss visibly engages with the full document, summaries stop being a safe substitute for everyone below.

5. If you have children in your life, read to them, past the age when it seems necessary. The single most protective factor in Horowitch’s essay is the one that costs nothing: a family where reading is simply what people do. Her father read to her through middle school. That, more than any curriculum, is why she can write for The Atlantic.

Tending the Scrolls

I keep returning to those scribes in Alexandria, because their job description turns out to be the moral of the whole story. The scrolls did not need to be defended from armies most of the time. They needed to be recopied, continually, by people who decided the effort was worth it. Literacy was never a possession. It was a practice, renewed one generation at a time, and it dies not by fire but by the quiet withdrawal of effort.

The worst thing you can do with all of this is despair, and conclude that the culture is lost and your own habits do not matter. The second worst thing is to congratulate yourself for having read this far and change nothing. The people who will thrive in a postliterate age are the ones who keep the practice alive in themselves and pass it on: who read the whole book, write the hard draft, and raise readers, not because it is nostalgic, but because it is now a competitive advantage and has always been a human inheritance.

An astonishing wealth of wisdom has been left to us, more accessible than at any moment in history, sitting one search away. Whether it gets read, and whether anything worthy gets added to it, was never going to be decided by the crowd. It is decided by the holdouts, one evening at a time.

I know which side of that line I intend to be on. I suspect, since you made it to the end of this essay, that you do too.

Everything I argued above about writing, that the struggle of it is where the thinking happens, is not theoretical for me. My new book, Honor Under Pressure, came out of exactly that struggle: years of slow reading and slower writing on a question I could not shake. Not whether you can perform under pressure, but whether the person you become while performing is still someone worth being. If this essay resonated, I think the book will too. It asks only for the thing that has become countercultural: a few hours of your sustained attention. You can find it on Amazon, and if you read it, truly read it, I would love to hear what you think.

On Saturday evening (7.11.26, just a little while ago as of this writing), Iran’s Revolutionary Guard navy fired on a commercial container ship transiting the Strait of Hormuz, struck it, and declared the strait closed indefinitely. Within hours, the United States began conducting strikes against Iranian targets in and around the waterway. Crude prices jumped roughly three percent in early weekend trading, before most of the world had even seen the headlines. The fragile ceasefire that has held, barely, since June is now functionally dead.

Think about that for a moment. On Friday, the most expensive stock market in modern American history closed near record highs while the ceasefire it was pricing as permanent was already visibly failing. By Saturday night, warships were exchanging fire in the channel that carries a fifth of the world’s oil.

That gap, between what the market believes and what the world is doing, is the subject of this piece.

A Crisis That Never Actually Ended

Regular readers know I wrote about the Strait of Hormuz back in March, when the first closure stranded roughly twenty thousand mariners and two thousand ships in the Persian Gulf and I called it the forty-mile hostage crisis. It is worth recapping how we got here, because the market’s amnesia on this subject is part of the problem.

In late February, coordinated American and Israeli strikes on Iran killed the Supreme Leader and triggered a spiral of retaliation. Within weeks the IRGC had mined the strait, attacked merchant shipping, and choked traffic to a near standstill. Brent crude went from double digits to a peak of $126. Dubai crude hit a record $166. Gulf producers cut output by ten million barrels a day. Fertilizer prices rose fifty percent. Then came a partial reopening, tolls of more than a million dollars per vessel, a collapsed reopening, another reopening, and finally the June memorandum that everyone agreed to call a ceasefire.

Here’s the truth: the strait never really reopened, and the war never really ended. It merely went quiet enough for markets to stop looking at it. Tankers were attacked again on July 7. By July 9, traffic through the strait had ground to a near halt. On Friday, Washington issued an ultimatum demanding Iran publicly declare every channel open and toll-free by Saturday. Iran’s answer came in the form of gunfire aimed at a container ship.

Not slowly, not quietly, not at some hypothetical future date: the confrontation is re-escalating now, with American munitions striking Iranian targets and the IRGC promising “severe” retaliation.

And through all of it, from the February strikes to this weekend’s exchange of fire, the Nasdaq rallied to new all-time highs.

The Most Expensive Market in a Century

Let me be clear about what this piece is and what it isn’t. I am not predicting a crash on a schedule. Nobody can do that, and you should be deeply skeptical of anyone who claims otherwise. What I can do is describe the conditions, because the conditions are extraordinary.

The Shiller CAPE ratio, which measures the S&P 500 against ten years of inflation-adjusted earnings, sits at roughly 41.8. In the entire recorded history of that metric, spanning more than 150 years, it has been higher only once: the final months of the dot-com bubble. The long-term average is about 17.

The Buffett Indicator, total US market capitalization measured against GDP, recently crossed 230 percent, an all-time record. For perspective, it peaked near 140 percent in March 2000, just before the Nasdaq lost more than three quarters of its value. Warren Buffett himself wrote in 2001 that when this ratio approaches 200 percent, “you are playing with fire.” We are not approaching 200 percent. We are a full thirty points beyond it.

This is the context in which a widely followed technician published a chart this week arguing that the major indices face a decline of forty percent or more from current levels, with the only open question being, in his words, the “angle of descent.” His wave counts and trendlines are interpretive, as all technical analysis is, and I hold them loosely. But I could not shake that phrase, because it reframes the debate in exactly the right way. At these valuations, the question is not really whether the market reverts. Markets always revert from extremes; that is what makes them extremes. The question is whether the descent is a slow glide over a decade of disappointing returns or a sudden drop compressed into quarters. History suggests the difference between those two paths usually comes down to a single variable.

A catalyst.

1973 and the Anatomy of a Catalyst

Students of market history will recognize this setup, because we have run this experiment before.

In the early 1970s, American investors fell in love with the Nifty Fifty: a small group of dominant growth companies, Polaroid, Xerox, Coca-Cola, IBM, that were considered “one-decision stocks.” You bought them and never sold, because their growth was inevitable and their franchises were unassailable. Valuations stretched to levels that made no arithmetic sense, and sophisticated people constructed elaborate justifications for why the old math no longer applied. The market grew narrow, expensive, and serenely confident, all at once.

Then, in October 1973, war broke out in the Middle East, and Arab oil producers embargoed the West. Oil quadrupled. Inflation, already smoldering, caught fire. The Federal Reserve was forced to tighten into a slowing economy, and by the bottom in late 1974 the S&P 500 had been cut nearly in half from its peak. The Nifty Fifty, those unassailable one-decision stocks, fell further still, and some took a decade to recover.

Here is what matters about that episode: the embargo did not cause the bear market. The valuations caused the bear market. The embargo merely selected the date. Overvaluation is the dry forest; the catalyst is only the match. And the cruelest feature of oil shocks in particular is that they attack an expensive market at its most vulnerable point, because they push inflation up while pushing growth down, which strips central banks of their ability to ride to the rescue. The Fed put, the assumption that rescue is always coming, works in a deflationary crash. It does not work at $150 oil.

Now look at the present arrangement. A market more concentrated in a handful of AI champions than at any time since, well, ever. Valuations exceeded only by 1999. A collective belief that these companies’ growth is inevitable and their franchises unassailable. And this weekend, a shooting war in the body of water that carries twenty percent of the world’s oil, with the world’s largest military now striking targets along its shores.

I don’t know if this is the catalyst. Nobody does; the strait has closed and reopened before, and markets shrugged each time. But notice that the shrugging is itself part of the pattern. Hyman Minsky taught that stability breeds instability, that every uneventful day strengthens the conviction that days will remain uneventful. Each time the market looks past a Hormuz closure and gets rewarded for it, the reflex deepens, positioning grows more aggressive, and the eventual repricing gets larger. The market has cried wolf about the wolf, if you will. The fable does not end with the wolf never showing up.

The Economy That Never Got Invited to the Party

The loan files that cross my desk at B:Side tell a different story than the indices do. Small business owners are still digesting the spring’s fuel costs, freight surcharges, and fertilizer prices. Their margins have no cushion left for a second oil shock. The regional banks that serve them are still nursing balance sheets marked for a rate environment that may be about to lurch again, in the wrong direction, if energy inflation forces the Fed’s hand. Wall Street’s euphoria and Main Street’s exhaustion have been diverging for two years. Events in the strait threaten to resolve that divergence the hard way.

And there is a generational dimension to this that worries me just as much. Most of the young professionals entering the workforce today, including the students in my classroom at ASU, have never experienced a real bear market. They have learned from every dip since 2020 that drawdowns are buying opportunities that resolve in weeks. An entire cohort has been trained by the most forgiving decade in market history. The dissonance between what the data shows and what their experience has taught them is one of the things that keeps me up at night, and it is why I refuse to stay quiet just because the timing is unknowable.

The Playbook

So what do you actually do? Not as a trader, but as a leader, an owner, a steward of a household or a team. A few disciplines matter more than any forecast.

• Stress test now, while it’s cheap. Run your business and your personal finances against $150 oil and a 30 to 40 percent equity drawdown. Not because either is certain, but because the exercise costs nothing today and everything later. If the numbers break, you want to know on a quiet Sunday, not in the middle of the storm.

• Prize liquidity over cleverness. In every crisis I have lived through, from the BodeTree days to the regional bank strains we watch at B:Side, the winners were rarely the smartest people in the room. They were the ones with cash and committed credit when others had neither. Boring balance sheets are what buy the bargains.

• Ignore the sirens on both sides. The permabears will tell you to sell everything; they have been wrong for fifteen years. The permabulls will tell you valuation never matters; they were saying that in March 2000 too. Position so that you are never forced to be a seller, and a forty percent decline becomes an event you endure and exploit rather than one that ends you.

• Lead visibly. If you run a team, your people will read this weekend’s headlines and look at you. Calm is contagious, and so is panic. Marcus Aurelius reminded himself that the mind adapts and converts every obstacle into fuel. Say plainly what you know, what you don’t, and what the plan is. That conversation, held early, is worth more than any hedge.

The Descent Is the Terrain

Here is where I land. The market is priced for a world without accidents, and we live in a world that is currently producing accidents on a weekly basis. Whether this weekend’s fire in the strait proves to be the catalyst or merely another rehearsal, the underlying arithmetic does not change: from a CAPE of 42, the next decade’s returns will be earned the hard way, and some portion of the descent, gentle or violent, is coming regardless.

The worst thing you can do is pretend the risk isn’t real. The second worst thing is to liquidate everything and hide, surrendering to a forecast that no one, bull or bear, can actually make. The people who come through periods like this intact are the ones who prepare while others party, who hold enough liquidity to act when others can’t, and who understand that surviving the angle of descent matters more than predicting it.

The forest is dry. The matches are lit and falling. You cannot control the weather in the Strait of Hormuz, but you can control whether your house is made of tinder.

That has always been enough. It still is.

P.S. If this piece resonated, it’s because moments like this one are not really about markets. They are about leading through a period of historic upheaval, what Neil Howe calls a Fourth Turning, when institutions buckle and the old rules stop working. I’ve been building a dedicated home for that conversation at thefourthturningleader.com, focused on the leadership disciplines this era demands. If you lead anything, a company, a team, a family, I’d encourage you to join me there.

If that scene sounds invented, ask around. Some version of it is playing out in businesses across the country right now, and the mechanics behind it guarantee it will get more common, not less.

You have probably noticed that the conversation around AI’s economics has changed. The industry’s financial critics spent two years being dismissed as cranks. They are not being dismissed anymore. Their argument, that the companies selling AI access are losing staggering amounts of money on every customer and have no credible path to stop, has migrated from contrarian corners of the internet to the mainstream financial press. The numbers stopped being disputable. The largest AI lab’s audited 2025 financials showed roughly thirteen billion dollars in revenue against thirty-four billion in costs, with net losses growing nearly eightfold in a single year. The Bank for International Settlements, an institution not known for drama, warned this June that data center spending is outrunning the cash flows meant to justify it. One infrastructure giant has taken on more than a hundred billion dollars in debt to build capacity for demand that exists mostly in projections, with its founder pledging personal shares as collateral.

I run a regulated SBA lender that was early to serious AI adoption, and I founded Main & Machine, an AI implementation firm, which means I get asked two questions almost weekly now. If the critics are right, is AI finished? And if I invest in this now, am I building on quicksand?

Here’s the truth: they are largely right about the money, and the technology is going to be fine. Both of those things are true at once, and the space between them is where every owner reading this should be doing their thinking.

So concede the critique. Assume the reset comes. What then?

The Bubble Is a Business Model, Not the Technology

Let me name what is actually happening in plain language, because the coverage keeps blurring two different things.

The thing inflating past reason is a specific business model: charging by the query for access to enormous general-purpose systems that cost more to run than customers will ever pay. The frontier labs priced their products below cost as a bet. Either running the models would get cheap fast enough to close the gap, or one winner would take the whole market and set prices at will. Neither outcome has arrived. Both look less likely with each earnings report.

Beneath the labs sits a second layer of exposure. The companies building the data centers have borrowed against the assumption that the demand curve holds forever. If the curve breaks, the debt breaks with it, and debt problems at that scale have a way of becoming everyone’s problem.

And then there is the third layer, the one that matters most to you: the customer. Owners have spent three years being trained to treat AI as a metered utility, priced per query, billed monthly, exposed to the balance sheet of a company they do not control. Every workflow built on that arrangement inherits its fragility. The shop owner from the opening scene learned this the expensive way. She thought she had bought a capability. She had actually rented one, at an introductory price, from a landlord who was losing money on the lease.

Now, the concession that keeps this piece honest. AI is real. The capability is real. I am confident enough in it that we are building our own lending operation around it right now, under the eyes of regulators and a board, and I will walk you through that decision shortly. The productivity potential inside well-designed workflows is as concrete as anything I have seen in twenty-five years of operating businesses.

What is fake is the pricing. AI is real. The bubble is a business model, not the technology. Hold those two sentences together and most of the panic, and most of the hype, resolves into something you can actually plan around.

Subscribe now

What the Dot-Com Collapse Actually Taught Us

We have run this experiment before, and recently enough that most Ampersand readers lived through it.

Between 2000 and 2002, the dot-com crash wiped out roughly five trillion dollars in market value. Pets.com died. Webvan died. Hundreds of companies whose entire thesis was “the internet, plus our logo” died. Serious people wrote serious columns declaring the internet a fad, a mania, a tulip craze with modems.

The internet did not die. What died was a set of business models built on the assumption that traffic itself was worth money, that growth could substitute for revenue indefinitely, and that being early was the same as being right. The technology kept improving straight through the wreckage. The crash killed the pricing and the ownership structure and left the capability standing.

Two things happened next, and both matter for the decade in front of us.

First, the companies that treated the internet as infrastructure rather than as a business model inherited everything. Amazon existed before the crash and survived it because it was a retailer first and a website second. It had inventory, margins, logistics, and customers who paid actual money for actual things. The internet made its operations better; the internet was never the product. When the froth burned off, the businesses built that way kept compounding while their flashier competitors liquidated office chairs.

Second, the overbuild itself became the foundation. The telecom companies of the late nineties laid staggering amounts of fiber optic cable at ruinous cost to the people who financed it. Most of those companies went bankrupt. The fiber stayed in the ground. That dark fiber, bought out of bankruptcy for pennies, became the physical substrate for everything the next two decades delivered: streaming, cloud computing, the modern web. The people who paid for the infrastructure and the people who profited from it were two different groups.

Think about that for a moment. The bubble’s investors funded the future and then handed it, at a discount, to whoever was still standing.

That is the pattern, and it is remarkably consistent across technology bubbles going back to the railways: the bubble destroys the pricing and the ownership, never the technology. The next generation of winners buys the wreckage cheap and builds on it. The question for an owner is simply which side of that handoff you want to be on.

The Six Things That Change When the Bubble Pops

If the pattern holds, and I believe it will, here is what the morning after looks like for business AI. Six predictions, each one already visible at the edges if you know where to look.

1. The token meter dies. Metered, per-query pricing exists because it lets vendors pass their unpredictable costs on to you. It survives only as long as customers tolerate it. After the reset, owners will pay for a workflow that works, priced as a fixed cost against a defined outcome, the way they pay for every other piece of operational infrastructure. Vendors who cannot commit to a fixed outcome will lose the customer segment that pays its bills on time, which is the only segment that matters in a downturn.

2. Small models eat most of the work. The industry spent three years chasing frontier scale because scale was the story investors funded. But sit inside an actual business and count the tasks. Ninety percent of the daily work in a lending office, a dental practice, or a fabrication shop involves reading documents, summarizing them, checking them for completeness, classifying them, and drafting routine language. None of that requires the largest system ever built. It requires a competent one that runs cheaply, predictably, on hardware someone you trust controls. The economics of the reset will force the industry toward what the work actually needs.

3. The workload moves in-house. Regulated industries figured this out first because they had to. When the system doing the work runs on a machine in your own building, three hard conversations get simple at once: the compliance conversation, the vendor conversation, and the pricing conversation. The direction of travel is toward local capability handling the bulk of the work, with selective use of larger outside systems for the small fraction of tasks that genuinely demand them. The frontier becomes a specialist you consult, and stops being a landlord you pay rent to.

4. Judgment stays human, on purpose. The bubble narrative sold “AI does the work.” The durable version draws a harder and more useful line. The machine drafts, summarizes, retrieves, and classifies. A person decides. Every business built around that boundary keeps running after the reset, because the reset touches the price of drafts and leaves the value of judgment exactly where it was. Businesses that handed decisions to the machine will spend the next few years discovering what that boundary was for.

5. Implementation beats subscription. The subscription era rewarded distribution: whoever could sign up the most users fastest won the funding round. The implementation era rewards delivery. The winning firms on the other side will be the ones who come inside your operation, learn your workflows, do the unglamorous integration work, and hand you a system that keeps working when the market convulses. Selling access was the old business. Building capability is the next one.

6. The most interesting companies after the reset will be mid-size operators, in every industry, who built AI into their operations quietly and now hold a durable cost advantage their larger competitors cannot match without ripping out the plumbing. The frontier labs will dominate the headlines through the correction, the way the telecom giants dominated the headlines through 2002. The compounding will happen somewhere much less glamorous: in the fabrication shop and the regional lender and the forty-person logistics firm that spent the bubble years building instead of subscribing.

Run all six predictions through a single filter and they say one thing. The bubble rewarded scale. The reset will reward control.

How We Are Building for the Morning After at B:Side

I want to show you what control looks like in practice, and I can do it from the inside, because we are in the middle of building exactly this kind of system right now.

B:Side Capital, the company I lead, is a nonprofit SBA lender operating across Colorado, Arizona, New Mexico, and Utah. We are regulated, examiner-audited, and board-governed. Every borrower file we touch contains sensitive financial data, and every credit decision we make is a decision a specific person has to own, in writing, in front of a regulator if it comes to that. When AI got hot in 2024, the temptation was the same one every executive team felt: buy the flashy vendor tools everyone else was buying and announce an AI strategy.

Our board approved a different path. We are building an internal system, named MARCUS, with an executive owner and a set of constraints that came before a single line of anything else. No borrower data leaves the building. No machine ever approves or denies a loan. No vendor sits between us and our regulator. The system runs on hardware we own, inside our own walls, using small models that fit on that hardware, with a deterministic rules engine making every eligibility and pricing determination the same way every time. The models draft. A person decides.

Here is what the system is being built to do. It reads incoming borrower files. It summarizes them. It runs a completeness check against what the file should contain. It drafts the first pass of internal memos, and it traces every claim it makes back to a source document, so the human reviewing the draft can verify rather than trust.

And here is what humans kept, deliberately and permanently. Every credit decision. Every hardship conversation with a borrower going through the worst stretch of their business life. Every submission to a regulator. The operating rule underneath all of it is simple enough to fit on an index card: the machine can hold knowledge, and it cannot hold responsibility.

Notice what building this way does to the economics. The cost of the system is knowable before it runs: the hardware, the electricity, the people who maintain it. Every one of those numbers sits on our side of the ledger, which means the cost does not move when someone else’s board decides to reprice. That is the part that matters for this article. The shop owner from the opening absorbed a tripled rate on ninety days’ notice. Our exposure to that kind of Tuesday is the electric bill.

I will offer the honest coda, because a case without mistakes is a brochure, and we are far enough into the build to have already collected a few. When we announced the project to our staff, I led with what the technology could do. I should have led with what would not change. People hear “AI initiative” and privately calculate their own odds, and I let that anxiety run longer than better leadership would have. We also learned that the operating rhythm around a system like this, the weekly review of what it drafts and where it fails, belongs on day one rather than on the list of things to formalize later. Whatever you build, at whatever scale, start there.

Three Questions Every Owner Should Ask a Vendor This Quarter

Now, I know what some of you are thinking. You run a twelve-person company, you are never going to build an internal system with a name and an executive owner, and a nonprofit lender’s architecture is not your Monday morning problem. Fair enough. You do not need our system. The design principles underneath it are fully portable, and they compress into three questions you can ask any AI vendor this quarter. Three, and only three.

One. Where does the work run, and who controls that hardware? If the answer is a vendor’s cloud, priced per query, you are exposed to the reset and to every repricing decision made between now and then. If the answer is hardware you or your implementation partner controls, at a fixed and knowable cost, the reset becomes a headline you read about rather than a bill you receive.

Two. What does the machine decide, and what does a human decide? If the pitch involves the machine making a decision your business will be held responsible for, a credit call, a medical judgment, a safety determination, walk away. The machine drafts. A person decides. Every time, in every workflow, with no exceptions carved out for convenience.

Three. What happens to your workflows if this vendor triples the price, gets acquired, or shuts down? If the honest answer is “we start over,” then what you have bought is a dependency wearing a workflow’s clothes, and dependencies fail on someone else’s schedule, usually at the worst possible moment.

That is the whole test. A vendor who answers all three cleanly is selling you capability. A vendor who cannot is selling you subscription-era plumbing at implementation-era prices, and the clock on that arrangement is already running.

Build So the Next Morning Belongs to You

Come back to the shop owner one last time. Eighteen thousand dollars, three tools, one tripled rate, and a workflow held hostage on a Tuesday. What she needed was a workflow she owned, running on infrastructure she controlled, priced against a result she could measure. A bigger subscription solves none of that. A better-designed operation solves all of it, and that option is available to her right now, at her scale, for owners willing to treat AI as an operating decision rather than a shopping decision.

The correction, when it comes, will be loud. Write-downs, bankruptcies, columns declaring the whole thing a mania. Remember the fiber in the ground. The technology will keep working straight through the noise, and it will get cheaper on the other side, and the owners who spent this era building workflows they control will wake up with a cost structure their competitors cannot touch.

AI is real. The bubble is a business model, not the technology. The bubble rewarded scale, and the reset will reward control, and control is built the same way it has always been built in a small business: deliberately, unglamorously, one workflow at a time.

The bubble will do what bubbles do. The businesses built on top of it, thoughtfully and quietly, will keep running the next morning. Build so the next morning belongs to you.

Share

To learn more about MARCUS or to see what human-centric AI adoption could look like in your business, check out Main & Machine.