The country that pays not to produce

Brazil does not have a capital scarcity problem. It has a destination problem: for every two reais put into building, one real remunerates those who do not produce.

Let me begin with the arithmetic almost nobody does out loud.

In 2025, Brazil as a whole — companies, families, and government combined — invested R$ 2.1 trillion in productive capacity. Factories, machines, construction, equipment, software, technology. Everything. It is what the IBGE calls gross fixed capital formation, and it corresponds to 16.8% of GDP.

In the same year, the public sector paid roughly R$ 1 trillion in interest on its own debt. It is the largest nominal figure ever recorded by the Central Bank, equivalent to 7.91% of GDP.

Read that again.

For every two reais this country put into building something, one real was transferred to remunerate those who already had money for doing nothing at all with it.

This is not an indignant newspaper-column metaphor. It is the arithmetic of the national accounts. It is published, audited, and available to anyone with patience and an internet connection.

Brazil does not have a problem of capital scarcity.

It has a problem of capital destination.

And before you hastily file me into some familiar political camp, let me take that possibility off the table: I will not ask for low interest rates by decree, I will not defend protectionism, I will not blame the financial market for existing, and I will not romanticize the Brazilian industrialist. I will do something more uncomfortable for every side.

I will look at the structure.

I. The diagnosis that is too easy

There is a ready-made, popular, and comforting explanation for Brazilian backwardness: the country became a service economy and therefore does not grow.

It is a bad explanation.

Rich countries are also service economies. Hospitals, engineering, logistics, software, research, insurance, consulting, education, design, maintenance. All of that is services. Many of these services raise industrial productivity, export knowledge, and sustain entire production chains. No mature economy escapes this destiny, and there is nothing degrading about it.

The Brazilian problem is of another nature, and it is harder to say.

We did not build a service economy.

We built, to a large extent, an economy of shelter.

We turned low-productivity services into the place where those expelled from production end up, and we turned the financial market into the preferred destination of the capital that should be financing production. They are two distinct movements with a common origin, and you must see both at once to understand anything.

The app driver is not to blame for this. He is working, taking risk, maintaining a car, paying for fuel, and supporting a family. He is, in almost every sense that matters, more of an entrepreneur than a good share of the people who call themselves entrepreneurs on event panels.

The problem begins when a country starts treating as modernization what is, very often, merely the last alternative available to a productive person.

Uber is not the cause.

It is the symptom.

And the symptom has a size. According to the PNAD Contínua survey, Brazil had about 1.7 million people working through digital platforms in the third quarter of 2024 — drivers, couriers, general service providers, and professionals. More than half were passenger transport drivers. But the absolute number matters less than the speed: in a study by the Central Bank itself, between 2015 and mid-2025 the country's employed population grew around 10%, while the contingent of app workers grew about 170%.

Keep those scissors in mind. They explain almost everything.

Because, at the same time, Brazil produced the best unemployment number of its recent history. Unemployment closed 2025 at 5.1%, the lowest in the IBGE series.

And productivity did not follow.

Surveys by FGV IBRE show that, in September 2025, the Brazilian worker produced only about 8% more than in 2012. Thirteen years. Eight percent. In the first quarter of 2026, productivity per hour worked fell again.

In other words: the country put more people to work without putting more capital behind each worker.

That has an ugly technical name and a simple consequence.

Employment rises. Income does not.

Unemployment can fall while the future disappears.

II. Size is not the problem. Composition is.

Let us go to the numbers, because numbers discipline the conversation.

In 2025, Brazilian GDP reached R$ 12.7 trillion, of which R$ 11.0 trillion correspond to value added at basic prices. Within that value added, services accounted for R$ 7.6 trillion, industry for R$ 2.6 trillion, and agriculture for R$ 775.3 billion. In proportion: roughly 69% services, 24% industry, 7% agriculture.

On manufacturing specifically, I need to issue a warning almost nobody issues — and which, to me, matters more than the number itself.

You will find it said that Brazilian manufacturing fell from 36% of GDP in 1985 to just over 10% today. That figure circulates in industry-association presentations, academic articles, and campaign speeches. It is, at best, exaggerated. The 1985 peak was calculated in the middle of hyperinflation, with the so-called financial dummy distorting the base of comparison; researchers at the very institutions that publish the figure have documented the overestimation. Other series, with different methodologies, put the peak somewhere between 21% and 27%.

And the current share varies between something close to 11% and something close to 14%, depending on whether you use current prices, constant prices, GDP, or value added.

I could have chosen the most dramatic number. It would have been more rhetorically efficient.

But whoever builds an argument on the most convenient statistic is building on sand, and sooner or later someone competent tears the whole thing down — including the true part.

And the true part survives any methodology: the productive base that should sustain this entire structure has become narrow, and keeps narrowing. In 2025, manufacturing value added shrank 0.2%. In the fourth quarter, the drop was 2.0% — the third consecutive negative result in that comparison.

The point, therefore, is not the size of services. It is their nature.

There is a categorical difference between a service that increases productivity and a service that merely disputes a slice of income that already exists.

An industrial software company can make a thousand factories more efficient. A trucking company can lower the cost of thousands of producers. A laboratory can create a molecule that did not exist. An insurer can allow a risky project to be executed — because insurance is not a cost; it is the technology that makes risk carryable.

These services do not compete with production.

They multiply production.

But an economy formed mainly of intermediation, consumption, informality, and low-productivity personal services depends on income that must have been created somewhere by someone.

Someone has to plant.

Someone has to build.

Someone has to manufacture.

Someone has to develop machines, molecules, materials, systems, and processes.

You cannot intermediate forever what has ceased to be produced.

III. The competitor nobody faces

Now we reach the point where I will lose some readers.

The Brazilian industrialist believes he competes against the competitor up the street, against the Chinese import, and against the tax burden.

He competes, above all, against the National Treasury.

Not as a declared enemy. As a rival buyer of the same money.

In July 2026, the Selic rate stands at 14.25% per year. Federal public debt reached R$ 9 trillion in May. Practically half of that stock is indexed to the Selic itself. And the holders are exactly who you imagine: financial institutions with about 31%, pension funds with about 23%, investment funds with about 21%.

Translating into a businessman's language: there exists in Brazil an asset with a sovereign guarantee, daily liquidity, known tax treatment, zero need for management, zero execution risk, zero labor litigation, zero inspections, zero delinquent clients — paying high double digits.

That asset is the floor of every allocation decision in this country.

It is what the textbooks call the risk-free rate.

Let me be blunt: there is no such thing as a risk-free rate at 14.25%. A rate of that magnitude is not the absence of risk. It is the presence of risk somewhere else — in the fiscal accounts, in the courts, in the currency, in the credibility of money, in the expectation of what comes next. The name "risk-free" is one of the most successful fictions in the financial vocabulary. It does not eliminate risk. It hides it and socializes it.

What it does, with brutal efficiency, is set the entry price of any productive project.

A factory does not merely need to be profitable. It needs to be profitable above a government bond that demands nothing from anyone. It must beat the financial cost, taxation, machine depreciation, regulatory insecurity, currency volatility, international competition — and still deliver a sufficient premium over an alternative that sleeps soundly.

In April 2026, the average rate on free-market credit to companies stood at 25.3% per year. The Cost of Credit Indicator, which measures the average cost of the entire national financial system's loan book, stood at 24.3%.

It is not enough to produce well.

One must produce miracles.

And here is the asymmetry that bothers me more than any other, because it is moral before it is economic:

The one who takes the risk is the businessman. The one who puts the family's assets into the operation is the businessman. The one who answers in court is the businessman. The one who loses sleep when the client pays late is the businessman.

And the one who is best remunerated, with the most predictability and the least exposure, is the one who took no risk at all.

A society can live with this inversion for a while. No society prospers with it permanently. When return systematically decouples from whoever carries the exposure, the system is not merely miscalibrated. It is training people to leave the game.

And they leave.

Not with a manifesto. With a spreadsheet.

IV. The industrial policy nobody called industrial policy

There is a sentence the Brazilian industrialist has repeated for thirty years, and it is wrong.

"Brazil has no industrial policy."

It does.

It is in the tax code, not in the inaugural speech. It was never voted under that name, never had a ministry, never appeared in a multi-year plan. And yet it directs more capital than any official development program this country has ever created.

It is called exemption.

Look at the list of instruments a Brazilian can buy today without paying income tax on the yield:

LCI. The Real Estate Credit Letter.

LCA. The Agribusiness Credit Letter.

CRI. The Real Estate Receivables Certificate.

CRA. The Agribusiness Receivables Certificate.

Incentivized debentures, under Law 12,431: infrastructure.

Real estate funds. Fiagro.

Now tell me where manufacturing is on that list.

It is not.

There is no industrial credit letter. There is no exemption for factory working capital. There is no tax-exempt paper for buying a machine tool, for an export line, for modernizing a plant, for applied research, for fine chemicals, for capital goods.

A Brazilian with R$ 500,000 can finance, without paying a cent of tax, a logistics warehouse, a soybean harvest, a shopping mall, a transmission line, and a residential development.

If he wants to finance a components factory, he pays tax.

And the size of this is not symbolic. At the end of 2025, according to B3, the stock of LCIs totaled R$ 508.8 billion, growing 29% in the year. LCAs totaled R$ 599.9 billion, up 16%. Those two instruments alone exceed R$ 1.1 trillion. Incentivized debentures issued in 2025 set a record: R$ 178 billion, an advance of more than 31% over the previous year.

Add it up. Compare it with the R$ 2.1 trillion the whole country invested in productive capacity that same year.

We are talking about more than half of the nation's annual investment parked — legally, rationally, with explicit state incentive — in instruments whose tax privilege points at three sectors.

And none of them is manufacturing.

I want to be fair here, because the lazy critique of this arrangement is as bad as its blind defense.

The exemption was not a conspiracy. It was a public policy decision with a defensible logic: real estate and agribusiness needed long-term funding, traditional bank credit did not deliver it, and the fiscal instrument created a market that today is deep, liquid, and reasonably efficient. To a large extent, it worked. Brazilian agribusiness is, by a wide margin, the most competitive sector of our economy, and it did not get there by accident.

The problem is not that the incentive existed.

The problem is that nobody ever asked what happens to those left out.

Because a tax exemption is not merely a benefit to the favored sector. It is a price signal to the entire country. When the State says, for twenty years, that the yield on a real estate paper is worth more after tax than the yield on an industrial paper of identical risk, it is not merely helping real estate. It is informing capital, every single day, in which direction to walk.

Capital listened. Capital always listens. It is the only thing it does well.

And there is a second, more perverse effect, which only now has begun to be said out loud. Representatives of the National Treasury itself have started attributing to the exempt papers part of the pressure on NTN-B rates — the inflation-linked government bonds. The logic is direct: if one instrument pays tax-free, all the others must pay more to compete with it. The exemption of the few makes money more expensive for all.

Join the two ends and you have the complete situation of the Brazilian industrialist.

He is outside the incentive.

And he pays, in his own funding cost, part of the bill for everyone else's incentive.

Twice.

In 2025, the government tried to fix this in the most direct and most clumsy way possible: Provisional Measure 1,303, which instituted a 5% tax on new issues of these papers. The measure lapsed in October without being converted into law. It will return to the agenda — everyone in the market knows it — but probably not before the elections, because the size of the stock is exactly what gives the resistance its political strength.

And here is the point that separates this letter from a complaint.

I am not asking for the removal of the agribusiness and real estate exemptions. Taking privilege away from those who hold it, in Brazil, is an exercise that consumes an entire legislature and usually ends in nothing.

I am pointing at something simpler and more embarrassing: for two decades, Brazil designed a sophisticated fiscal architecture to steer private savings toward chosen sectors, proved that the tool works, and never once pointed it at industry.

It was not for lack of instrument.

It was for lack of decision.

And, frankly, also for lack of organization by those who should have demanded it. Agribusiness organized and got its credit letter. Real estate organized and got its own. Infrastructure organized and got Law 12,431. Manufacturing spent the same period asking for payroll tax relief, favorable exchange rates, and tariff protection — that is, asking for relief on cost, not access to capital.

It asked for the wrong thing.

Cost relief extends a company's life by a few quarters.

Access to capital of compatible tenor builds a company for decades.

That is the whole difference between a sector that survives and a sector that expands — and it explains much of why Brazilian agribusiness competes in the world while our manufacturing shrinks at home.

V. Competing against another State

It is common to summarize the Brazilian disadvantage by saying the Chinese borrows at 3% while the Brazilian pays 20%.

The sentence works as provocation and fails as analysis.

Not every Chinese company gets cheap credit. Not every Brazilian industry pays the same rate. The real asymmetry is deeper and more interesting.

In June 2026, the OECD made public the MAGIC database, which tracks industrial subsidies received by 525 of the world's largest manufacturing groups, across 15 structural sectors, between 2005 and 2024. The numbers are consistent and unpleasant for those who like simplifications: Chinese companies received, on average, three to eight times more government support than those from OECD countries — in direct grants, tax breaks, and below-market loans. About 60% of Chinese companies' global market share gains in the period can be attributed to these subsidies, against a global average of 22%.

The Brazilian businessman believes he is competing against another company.

Frequently he is competing against another State.

Administrative efficiency does not solve a disadvantage of that magnitude. You can be twice as good as your competitor and still go broke before he does.

But — and this "but" is the reason I am writing this section instead of simply complaining about China — the same report carries a finding almost nobody cites, because it spoils the narrative on both sides.

The subsidies did not produce significant gains in productivity or profitability.

The OECD secretary-general used an image I wish I had used first: industrial subsidy works like doping in sport. It makes the less prepared athlete win the race. It does not make him a better athlete.

That completely changes the practical conclusion.

If subsidy created competence, the recipe would be obvious: subsidize more. Since subsidy buys market share without creating productivity, copying the Chinese model would give us exactly what it gives — large, dependent, mediocre companies, sustained by transfers that the Brazilian State, let it be said, cannot afford.

And there is an additional detail in the same report: the subsidies received by companies from economies like Brazil, India, and Indonesia are comparable, in relative terms, to those received by American companies.

We are not losing because we subsidize too little.

We are losing because we price our own capital as if the country were about to end.

That is a different disease, and the remedy is different.

VI. The venture capital mistake

Some conclude from all this that the solution would be to bring venture capital to industry.

In some cases, yes.

In most, no — and insisting on it has cost serious people a great deal of money and many years.

Venture capital is an instrument designed for a specific format of risk: fast-growth companies, high uncertainty, declining marginal cost, potentially extraordinary margins, and the possibility of multiplying capital in a few years. The model works because the portfolio accepts that most assets die, as long as one of them pays for all.

A food company neither dies nor multiplies by a hundred. A components factory neither dies nor multiplies by a hundred. A metalworks, a construction company, an equipment maker — none of them fits that distribution.

They have no tail. They have a cycle.

These companies do not need venture capital.

They need adequate capital.

They need long-term credit, intelligent receivables prepayment, machine financing with well-structured collateral, project finance, private credit funds, hybrid equity instruments, patient family capital, and structures compatible with the real economic cycle of each business.

A factory cannot be financed like an app.

A company that takes eight years to build productive capacity cannot depend on capital that demands an exit in four.

This sounds obvious written down. It is not what happens in practice, because in Brazil the instrument is rarely chosen by the nature of the asset. It is chosen by whatever happened to be on the table that month.

The Brazilian problem, therefore, is not just scarcity of money.

It is scarcity of architecture.

There is money looking for yield.

There are businessmen looking for capital.

There are machines that could be bought, companies that could grow, contracts that could be executed, technologies that could be developed.

And still the meeting between these parties does not happen.

The money prefers the collateral to the project.

It prefers the building to the entrepreneur.

It prefers the past recorded on the balance sheet to the future that still needs to be built.

That preference is rational for each institution in isolation. No credit officer is fired for declining a good deal; many are fired for approving a bad one. The asymmetry of incentives inside the institution reproduces the asymmetry of incentives outside it.

Rational at retail. Destructive at wholesale.

VII. The market that trades twice what it finances

Let me show this with the piece of data that, in my opinion, is the most revealing of all — and that went practically uncommented when it came out.

In 2025, the Brazilian capital market set a historic record: R$ 838.8 billion in offerings. Debentures alone totaled R$ 492.8 billion, the largest volume ever recorded. Commercial notes, receivables funds, securitization — everything grew. It was, by any conventional metric, an excellent year.

In the same year, the traded volume of debentures in the secondary market reached R$ 947.4 billion.

Almost double the primary.

In other words: the Brazilian market traded twice as much paper as it issued.

This is not a scandal. Liquidity is a virtue, and a deep secondary market is a condition for the primary to work. Whoever buys an eight-year paper needs to know he can sell it in three. I am not pointing at a crime.

I am pointing at a symptom, and it becomes sharper when you open up the destination of the money.

Of the funds raised via debentures in 2025, the largest slice went to infrastructure, and the second largest — 26.2% — went to debt repayment. In the first five months of 2026, the pattern repeats: infrastructure first, followed by ordinary management and debt repayment.

And the sectors? Electric power with R$ 119.8 billion, transport and logistics with R$ 88.3 billion, financial with R$ 79.5 billion, sanitation with R$ 44.5 billion.

Look closely at that list.

They are, almost all, sectors of regulated cash flow, long contracts, predictable revenue, and heavy fixed assets. They are exactly the assets a risk-averse market can price without needing to understand business.

What barely appears on that list is the mid-sized productive company — no concession, no rating, no thirty-year contract with a regulatory agency — which constitutes the backbone of any serious industrial economy.

Schumpeter wrote that the banker is the ephor of the capitalist economy: the magistrate who authorizes the entrepreneur to command resources that are not yet his, in the name of production that does not yet exist. It is the most creative function in the system. It is also the hardest, because it demands judgment about the future instead of accounting about the past.

A mature market is not the one that trades a lot.

It is the one that finances what does not yet exist.

VIII. A trade born beside the producer, not above him

It is worth remembering where all this comes from, because the etymology is more honest than most reports.

The word bank comes from the piece of furniture. From the wooden bench the Italian money-changer set up in the square, beside the merchant, the porter, the weaver, and the wool buyer. When the money-changer failed to honor his commitments, they broke his bench in front of everyone. Hence banca rotta. Bankruptcy.

Notice the moral architecture of that scene: the banker stood in the square. At the same ground level where the business happened. Exposed to the same crowd. Subject to the same shame.

The Medici did not grow rich managing abstractions. They grew rich building a network of branches in Florence, Rome, Venice, Bruges, and London to finance exactly what they knew how to evaluate: the wool and textile trade, with people they knew, along routes they understood. Their capital traveled glued to the merchandise. When it drifted from the merchandise and got too close to the papacy and the princes — that is, when it traded productive risk for sovereign risk — the Medici bank began to die.

The Fuggers, in Augsburg, financed silver and copper mines in Tyrol. They did not buy mining securities. They bought stakes in physical production, with technicians, engineers, extraction contracts, and operational knowledge. They too ruined themselves the day most of the balance sheet became credit to the Spanish Crown, which defaulted with the regularity of a clock.

The goldsmiths of London, in the seventeenth century, kept other people's gold and discovered they could lend part of it. There was born, in practice and not in theory, the mechanism that financed the Industrial Revolution: turning idle savings into productive capacity.

Walter Bagehot, editor of the Economist, described in the nineteenth century what made London different from every other financial center in the world. It was not the size of English savings. It was the speed with which those savings found the enterprise that needed them. The ordinary Englishman, without fortune and without name, could raise capital for an industrial idea that the French aristocrat, far richer, never could. That was the advantage. Not the stock. The routing.

None of this is nostalgia. It is a structural observation that crosses five centuries and admits no relevant exception:

The financial system produced wealth whenever it stood close to those who produce, and produced mere transfer whenever it drifted away.

And there is an even more uncomfortable pattern inside that pattern. In Florence, in Augsburg, in Amsterdam, in London — everywhere — the moment private capital migrates massively from financing production to financing the sovereign is the moment that financial center begins to lose importance in the world. Not because the sovereign does not pay. Frequently he pays very well. That is precisely the problem.

When the State becomes the financial system's best client, the financial system no longer needs to understand companies.

And when it no longer needs to understand companies, it unlearns.

That unlearning is slow, silent, and nearly irreversible, because knowledge about productive sectors is not in any manual. It is in people. In analysts who spent twenty years looking at metallurgy. In directors who have seen three full cycles of a sector. In relationships built on factory visits, not quarterly video calls.

When that generation retires without training a successor, the country loses something money cannot buy back.

It loses the capacity to judge.

IX. The financial market forgot its function

I need to be precise here, because this is the section where the careless will think I am preaching against banks, and the hasty will think I am defending banks.

I am doing neither.

The financial market was not created to enrich operators. That is a possible consequence of its activity, not its economic purpose.

Its function — the only one that justifies the legal, regulatory, and institutional privileges it receives everywhere in the world — is to collect society's dispersed savings, evaluate risks, and direct capital to projects capable of producing future wealth.

The bank should connect those who accumulated capital to those who know how to use it productively.

The investor should be paid for taking risk.

The analyst should separate good projects from expensive delusions.

Credit should anticipate a future production that cannot yet be financed with present cash.

When this mechanism works, the financial market is one of civilization's greatest inventions — comparable, in my reading, to bookkeeping and the enforceable contract, and probably more important than most technologies that win prizes.

When it stops working, it becomes a tollbooth.

The spread is not morally condemnable. Neither is bank profit. An institution that takes risk, maintains liquidity, fights fraud, meets regulation, provisions losses, and manages capital must be remunerated. Whoever thinks otherwise has never managed a loan book with double-digit delinquency.

The problem begins when the system becomes structurally better at extracting rent from the existing economy than at financing the economy that does not yet exist.

It is possible to make a great deal of money managing decadence.

For a while.

One can finance household consumption, roll over public debt, charge fees, structure collateral, trade assets, prepay receivables, and collect management fees on idle wealth. Fernand Braudel observed, studying four centuries of European capitalism, that the migration of capital from commerce and production into pure finance usually marks the autumn of a cycle, not its maturity. Genoa did it. Amsterdam did it. London did it. In every case, the financial years were the most profitable ones and the last ones.

Without growth in income, productivity, and production, the financial system's own base begins to rot.

Without healthy companies, there is no healthy corporate credit.

Without good wages, there is no sustainable mortgage lending.

Without production, there is no insurance, foreign exchange, investment, capital markets, or wealth management that stays prosperous for long.

The financial market does not live above the real economy.

It lives on top of it.

X. A country where few show up

Through July 9, 2026, the Federal Revenue Service had received 46.19 million individual income tax returns. Brazil's estimated population is approximately 213 million people.

That is, roughly, one return for every five Brazilians.

This figure is usually used the wrong way, and I will not use it that way.

It does not mean that only one in five people pays tax in Brazil. Brazilians of every income bracket pay taxes on consumption, energy, fuel, services, and goods — and, as a share of income, the poorest pay more. There are also dependents, children, retirees, and people legally exempt from filing. Whoever uses this number to say that "few sustain many" is doing demagoguery with statistics.

But the number reveals something else, harder to hide and harder to fix: the base of Brazilians with formal income and assets sufficient to appear fully in the State's records remains narrow.

There is an entire country that works, consumes, and pays taxes, and accumulates almost nothing.

People who cross life without building assets, without access to productive credit, without the capacity to invest in companies, and without any share in the capital of the country where they were born.

Add that to the rest of the picture and the portrait is complete.

Brazil invests 16.8% of GDP. China and India invest above 30%. Chile, Peru, and Mexico run between 20% and 24%. In 2020, about 87% of the world's countries had an investment rate higher than Brazil's.

In research and development, Brazil spends around 0.5% of GDP, against a global average of roughly 2.2%.

In 1980, the country accounted for about 2.8% of world GDP. Today it accounts for something close to 2.1%.

None of these numbers is a scandal in isolation. Together, they are a diagnosis.

A society like this does not indefinitely sustain a sophisticated financial market.

It can sustain an expensive credit system.

It can sustain an enormous public debt market.

It can sustain a few very well-decorated islands of prosperity.

But it does not sustain a prosperous civilization.

XI. The false solutions

Since this letter has already irritated reasonably everyone, let me use the momentum to close the easy exits.

It is not closing the border. Protectionism without a deadline, without targets, and without accountability merely converts private inefficiency into public cost — and transfers the bill to the consumer, who is always the poorest link in the chain. Temporary protection with verifiable counterparts is industrial policy. Permanent protection without counterparts is corporate retirement paid for by the people.

It is not handing subsidized credit to the government's friends. Cheap capital without criteria creates dependent businessmen, artificial projects, and socialized losses. We have done this. It is documented. It cost dearly and left no industry behind.

It is not lowering interest rates by decree. The cost of money reflects inflation, fiscal risk, legal insecurity, delinquency, banking competition, collateral quality, and trust in institutions. Pretending these causes do not exist does not eliminate the price — it merely transfers the problem to the currency, where it reappears larger and crueler, charged to those who cannot protect themselves.

And it will not be a new generation of apps that replaces the lost industrial depth. Software is fundamental infrastructure of the modern economy and I have no smokestack nostalgia. But even the digital economy depends on energy, data centers, cables, equipment, construction, mining, semiconductors, and logistics.

The cloud also rests on concrete.

The challenge is not choosing between industry and technology, between services and production, between market and State.

It is making each institution return to its function.

The State must offer stability, infrastructure, legal certainty, and horizontal policies capable of lowering the cost of producing — and must stop being the most attractive buyer of its own country's money.

The businessman must take risk, innovate, build teams, and construct efficient organizations — and stop asking for protection as if it were a vested right.

The financial market must find, evaluate, and finance the businessmen capable of turning capital into productive capacity — and stop confusing wealth management with capital allocation.

Each one must occupy its own place.

XII. The rebuilding begins with capital

Brazil does not suffer from a lack of courageous entrepreneurs.

There may be too much courage.

There are men and women mortgaging homes, maxing out personal credit lines, prepaying contracts, and risking the family's assets to keep companies alive in an environment that systematically punishes whoever invests long. Much of that courage is admirable. Part of it is not courage — it is lack of alternative disguised as strategy, and the market has the obligation to tell one from the other.

What is missing is not audacity. It is a structure capable of recognizing courage, separating competence from imprudence, and delivering the correct financial instrument to each company.

Not every businessman should receive money.

Not every factory deserves to survive.

Not every national idea is strategic.

Financing production does not mean abolishing judgment.

It means elevating it.

Capital must be directed to the businessmen who know their markets, preserve margin, honor contracts, train people, and manage to convert scarce resources into something society values enough to pay for spontaneously.

The role of those who work with money is not to flatter the businessman.

It is to judge him correctly.

To deny capital to the incompetent.

To structure capital for the competent.

And to stay close to what was financed, because whoever finances and walks away was not financing — he was betting.

This requires more than a spreadsheet. It requires sector knowledge, trust built over years, judgment of character, real understanding of collateral, the ability to structure a deal, and the willingness to remain beside a company through its true cycle, not through the cycle of somebody's term in office.

In other words: it requires bankers.

Not salesmen of banking products.

The distinction is old and has nearly been lost. The banker is the one who answers for the decision. The salesman is the one who answers for the quota. One signs at the bottom. The other signs in the space reserved for the client.

And since this letter is read by businessmen, not economists, let me come down from the level of the country to the level of your desk.

If you own a company and have capital to allocate, there are three questions worth more than any Selic projection for 2027.

The first: what is the real tenor of my asset? Not the tenor of the contract. The tenor of the asset. A warehouse lasts thirty years. A production line lasts ten. Inventory lasts ninety days. If the tenor of your liability is shorter than the tenor of your asset, you do not have a business — you have a treasury operation with industrial risk embedded, and it will break at some point for a reason that has nothing to do with the quality of what you produce. Most of the bankruptcies I have seen up close were not caused by a bad product. They were caused by mismatch.

The second: does the person across the table answer for the decision or for the quota? It is the most revealing question and the least asked. It is answered by watching what happens when the deal goes wrong: who disappears, who calls, who renegotiates, who invites you to talk before the installment falls due. Everyone is excellent at the signing. The proof is at the first stumble.

The third: what am I really buying when I accept a guaranteed 14%? You are not buying yield. You are buying the option not to decide. There are moments in the life of a company when that option is worth a great deal — after a sale, before a succession, during a crossing. And there are moments when it is merely fear dressed up as prudence. The difference between the two is the difference between wealth that crosses generations and wealth that merely ages well.

I cannot answer these questions for you. Nobody can, at a distance.

But I can state one thing with some confidence: whoever answers these three questions honestly makes better decisions than whoever spends the whole year trying to guess the next Copom meeting.

One of them you control.

The other you do not.

XIII. The duty of money

For decades, we treated production as a vulgar activity and money as a superior one.

The industrialist dirtied his hands.

The financier managed abstractions.

One faced suppliers, employees, broken machines, inspections, idle inventory, and delinquent clients.

The other was paid to calculate the price of risk.

There is nothing wrong with calculating the price of risk. It is a rare, difficult, socially valuable skill.

But there is something deeply wrong when the one who calculates the risk becomes systematically more valued than the one who takes it.

A society that honors only its intermediaries ends up with nothing left to intermediate.

Brazil does not need to declare war on the financial market. It needs to return it to its mission — and, for that, it needs to stop offering it an alternative so comfortable that any other choice looks irrational.

We do not need less capital.

We need capital put in the right place.

Capital for machines.

Capital for technology.

Capital for construction.

Capital for exports.

Capital for the businessman who does not yet have a perfect balance sheet but has already demonstrated competence, character, and the capacity to execute — which is, by the way, exactly the kind of information no rating system captures and that exists only for those who take the trouble to be close.

Money should not replace work.

It should allow work to reach a scale it would never reach alone.

That is the difference between finance and rentism.

One finances the future.

The other charges for the past.

If we keep rewarding capital for staying sheltered and punishing whoever puts it in motion, Brazil will not run out of money.

Money will run out of country.

No country becomes rich when the best business is not producing.

Leo Bentier

The Bank Will Become Infrastructure

A letter on the quiet separation of balance sheet and trust — and on who will inherit the client once the bank becomes merely the place where money sleeps.

Money needs a vault; the client needs a face. For a century, the bank pretended to be both.

Observe what happens when a good banker changes banks.

He does not leave alone. He takes with him the late-afternoon phone calls, the Saturday messages, the requests for an opinion before a decision, the habit of being the first person consulted when unexpected money — or an urgent problem — appears. And, more often than not, he takes part of the book with him. Not because he stole anything. Because there was never anything to steal. The client was not following the bank. He was following the man.

That alone ought to settle the question of who owns the client. The contract belonged to the bank. The account, the file, the transaction history, the branch number — all of it was the institution's property, recorded in its systems, protected by its clauses. But the trust never belonged to it. Trust is a relationship between two people, and relationships do not appear on the balance sheet.

For a long time, that distinction had no practical consequence. The bank could pretend to own the client because the banker who owned the relationship was chained to the bank. There was nowhere to take the book. Changing institutions meant starting again from zero — no history, no pre-approved credit, none of the infrastructure that only a bank could offer. The professional's cage was, indirectly, the client's leash.

That architecture lasted decades. It is ending. And when it ends completely, the institutions will discover that they confused, for all that time, two things that merely looked alike: control of the infrastructure and ownership of the relationship. They are distinct things. They always were. Technology is only making that distinction visible — and expensive.

This letter is about that separation. It is the natural sequel to what I wrote earlier about the greatest asymmetry in Brazil: that the entrepreneur buys money alone, poorly represented, seated on the wrong side of the table. That letter diagnosed the problem. This one is about who will occupy the space that is opening — and why, in the end, he will be larger than the banks he represents.

How Trust Moved Into the Bank

It is worth remembering where the word comes from.

Bank comes from banca — the table, the wooden counter over which the money-changer of the Italian Renaissance squares counted and exchanged coins. In the beginning, a bank was not a building, nor an institution, nor a brand. It was a piece of furniture in the square and a man seated behind it. You did not trust the bank. You trusted that specific man, whose reputation you knew, whose family you knew, whose ruin would arrive at his own door before it arrived at yours. When a money-changer failed, the tradition holds, his table was literally broken in public: banca rotta, the broken bench — from which we inherit the word bankruptcy. The man's credit and the furniture in the square were one and the same thing.

The Medici, who made banking the foundation of a dynasty, did not build an institution in the modern sense. They built a network of trusted men scattered across Florence, Rome, Venice, London, Bruges — each branch run by a partner or a manager whose loyalty and judgment were the only asset holding it together. What circulated among those squares was not merely gold; it was correspondence, reputation, the word of men who answered personally for what they did. The Medici bank was, above all, a network of personal relationships that happened to move money.

The institution, in the sense we give the word today, came afterward — and it came slowly. It took three centuries. The goldsmiths of seventeenth-century London began issuing receipts for the gold they kept in their vaults, and those receipts began circulating as if they were the gold itself: there, almost by accident, the banknote was born, and with it the separation between the custody of the metal and the paper that represented it. Then came the joint-stock bank, limited liability, the branch network, the brand. Each of those inventions pushed the client a little further from the man and a little closer to the institution. In place of my banker — a person — one came to say my bank — a building, a logo, a balance sheet.

It was a remarkable migration, and it is important to understand that it was an anomaly. Throughout almost the entire history of credit, trust lived in individuals. Only in a relatively brief interval — that of the last hundred, hundred and fifty years — did it manage to move inside institutions, to the point where people trusted an acronym without ever having looked anyone in the eye. The building became the object of trust. The architectural solidity of the old branches, with their columns and their marble, was not decoration; it was rhetoric. It said: deposit your trust here, because this is going nowhere.

Why was trust able to make that move? There is a precise economic answer, and it was given in 1937 by a young economist named Ronald Coase, in a short essay on a question no one had thought to ask: why do firms exist? If the market is so efficient at coordinating production, why does so much activity happen inside firms, under hierarchical command, rather than through contracts between individuals? Coase's answer was that using the market has costs — the costs of discovering prices, of negotiating, of drafting and policing contracts, of trusting strangers. When those transaction costs are high, it is cheaper to internalize the activity inside a firm. The firm exists to spare the friction of doing everything through the market.

The modern bank is one of the most perfect illustrations of that law. It bundled, under a single roof, four activities that could, in principle, be performed by separate agents — and it bundled them precisely because, in the world in which it arose, coordinating them through the market would have been too costly. Information was scarce and dear. Trust in strangers was impossible. Settling an operation required paper, stamps, days. Gathering all of it inside one institution was the efficient solution. And because the institution held the infrastructure, it came to hold the relationship too — not because the relationship belonged to it by nature, but because, in that world, the relationship had no way of existing outside it.

Hold on to Coase's reasoning. It contains, inverted, the entire future. For if the firm exists to spare transaction costs, then there is a corollary Coase did not need to state but which is implied in every line: when transaction costs collapse, the boundary of the firm recedes. What was inside becomes able to happen outside again. And that is exactly what Brazilian financial technology has done over the last five years — it collapsed the transaction costs that kept the bank's four functions bound under the same roof.

The Four Functions

Before speaking of the separation, one must be precise about what is being separated. During the century in which trust lived inside institutions, the bank did four things at once, and the fact that it did them together was so natural that almost no one saw them as distinct things.

First, the bank manufactured the products. It created the fixed-income note, the fund, the insurance policy, the credit line, the structured operation. It was the factory.

Second, the bank controlled the infrastructure. It held the money, kept custody of the assets, took the credit risk, settled the operations, complied with the regulation, ran the systems that moved the money. It was the plumbing.

Third, the bank distributed the products. It carried the factory to the client through a network of branches and managers. It was the shelf.

Fourth, the bank held the relationship with the client. It knew the client, served the client, advised the client, was the one the client turned to. It was the table — in the old sense, the bench in the square, the man seated behind it.

The conceptual error that will cost the institutions dearly was confusing the last two with the first two. The bank believed that, because it controlled the infrastructure, it owned the relationship. It believed the shelf and the table were the same thing — that whoever distributes the product necessarily owns the client. For decades that confusion was harmless, because the four functions were in fact glued together and no technology could pry them apart. The banker represented one shelf, and the client, lacking any alternative, accepted the shelf the banker represented.

But the shelf was never the table. And the moment technology allows the same man — the same judgment, the same relationship — to consult several shelves rather than represent a single one, the confusion of a century comes undone at once. The banker represented one shelf. The independent operator can represent the client before all the shelves. That is the small sentence around which everything that follows turns.

What Technology Is Separating

There is no need to speculate about this future. It is already under way, in public figures, with dates and regulations.

Begin with the plumbing. Throughout history, moving money between institutions was slow, expensive, and therefore a competitive advantage for whoever controlled the pipes. Pix turned instant settlement into a public utility — free, universal, indifferent to which bank you use. The moment transferring money stopped costing time and money, the payments infrastructure stopped being a moat and became a sidewalk. Everyone walks on it. No one owns it.

Then came the portability of data. Brazilian Open Finance turned five in February 2026 as the largest open-finance ecosystem in the world: more than 100 million connected accounts, 154 million active consents, and a 143% growth in unique consents in a single year. What that means, translated into the language of relationships, is simple and profound: the client's history stopped being the bank's exclusive property. That which for a century held the client captive — the fact that his entire history, his credit, his financial reputation were locked inside a single institution — can now travel with him. The data that was a leash became a passport. And credit portability, in testing since late 2025 and slated for the payroll loans of federal public servants as early as 2026, promises to make the migration of debt between banks as automatic as changing your mobile carrier without changing your number.

Note, in passing, a figure that reveals the waste of the old architecture: among Brazilian small and mid-sized companies, 86% maintain a relationship with more than one financial institution, and a quarter of them operate with six to ten banks simultaneously. The entrepreneur already lives fragmented across counters. What he lacks is not access to more banks. It is someone who sits on his side of the table and operates all of them for him.

And then, in November 2025, the regulator itself wrote the thesis into law. Joint Resolution No. 16, issued by the Central Bank and the National Monetary Council, regulated Banking as a Service — the model in which a licensed institution offers its banking infrastructure, through programming interfaces, so that another company may use it. And in regulating it, the rule did something no manifesto could have done with the same authority: it formally separated, in legal text, two roles that tradition had kept fused. On one side, the providing institution — the one that holds the license, keeps the money, takes the risk, answers for custody, for compliance, for secrecy. On the other, the contracting entity — the one that relates to the client but which, the rule says expressly, may not transact in its own name. The infrastructure belongs to one. The relationship belongs to the other. The Central Bank wrote, in the language of a resolution, the sentence that governs this letter: the balance sheet on one side, trust on the other.

This is the point almost everyone reads backward. Faced with these movements, the easy commentary is to announce the death of the banks. Nothing could be more mistaken. While the relationship disperses, the balance sheet concentrates as never before. The four largest banks in the country hold 57.9% of all credit operations, 54.7% of all assets in the financial system, 57.1% of deposits. In some lines, such as housing credit, concentration exceeds 92%. And — the detail is delicious — the four largest banks' share of equity brokerage rose from 41.6% to 50.4% in a single year, because the giants have been buying back the independent brokerages that dared to exist. The balance sheet is not fleeing the banks. The balance sheet is entrenching itself in them.

What flees is another thing. What flees is the relationship. And to confuse the two — to imagine that the concentration of capital and the dispersion of trust are contradictory movements, when they are in fact the same movement seen from two angles — is the error that will keep the institutions from reacting in time. They will grow larger, more powerful, more solid in everything that requires scale. And, precisely for that reason, more distant from that which never required scale at all: knowing a client by name.

It Is Not Disintermediation. It Is Reintermediation.

There is an error that must be rejected firmly, because it is the error of the enthusiasts, and enthusiasts tend to be wrong exactly where they seem most advanced. It is the idea that the future will be without intermediaries.

It will not. The future will not have less intermediation; it will have a different intermediation. We are not witnessing a disintermediation — the fantasy of a world in which each individual, armed with an app, dispenses with any mediator and works the global financial system alone. We are witnessing a reintermediation: a new division of roles between whoever provides the capacity and whoever provides the judgment.

The distinction is not academic. It is the difference between understanding and not understanding what is coming. Whoever believes in disintermediation builds tools for the client to fend for himself — and discovers, years later, that the client does not want to fend for himself, that the wealthy client wants it even less, that no one with meaningful wealth and a busy life wishes to spend his afternoons comparing the yields of real-estate receivables. The client does not want a tool. He wants someone he can trust to use the tool for him. In a world of abundant options, scarcity is not the option. It is the judgment about the options.

So the new division will be this, and it is worth stating with the clarity of a principle:

The institution will supply capacity. The operator will supply judgment.

The bank will come to resemble ever more closely what electricity, plumbing, and the telecommunications network became: a necessary, powerful, invisible infrastructure. No one wakes up grateful to the power company; no one has a relationship with the water pipe. Infrastructure, when it works well, disappears. And the bank is heading toward that condition — indispensable and imperceptible, the place where the money sleeps, the engine no one sees beneath the hood.

The independent operator will be the human interface of that infrastructure. He will be the face, the voice, the judgment, the responsibility. He will be the person you call.

That pattern, incidentally, is neither new nor exclusive to finance. It has already repeated itself, with an almost monotonous regularity, in nearly every industry technology has touched. There was a time when the film studio owned the star: lifetime contracts, the face on the screen belonged to the logo before the film did. Technology freed the star, and the studio retreated to what required capital — to finance and distribute — while the talent, the relationship with the public, migrated to the individual. The record label owned the artist; today it holds the distribution infrastructure while the artist carries his own audience from platform to platform. The newspaper owned the journalist; today many journalists take their readers with them, and the paper discovers, too late, that the subscription followed the subscription to a name, not the brand on the masthead. In each case, the same design: the institution keeps the capital infrastructure, which requires scale; the individual keeps the relationship, which requires proximity. Finance is merely the last, and the most lucrative, of those industries to walk through the same door.

The Separation of Balance Sheet and Trust

I would give a name to this transformation, because things without a name are hard to see, and this one in particular deserves to be seen clearly: the separation of the balance sheet and trust.

They are two substances of opposite natures, and the fact that they coexisted inside the same institution for a century should not deceive us about how foreign they are to each other.

The balance sheet requires scale. It requires capital — a great deal of capital — and the capacity to absorb losses. It requires regulation, licenses, systems, compliance, teams of thousands, a machine that only makes sense above a certain size. The balance sheet is impersonal by vocation: the larger it is, the better it performs its function. No one wants the bank that keeps their money to be small and intimate. They want it to be large and solid.

Trust requires the opposite. It requires proximity, not scale. It requires reputation — which is always someone's, never something's. It requires judgment, which cannot be delegated to a system. And it requires individual responsibility: the possibility of looking into the eyes of whoever answers for the advice you were given. Trust is personal by vocation. It does not improve when it grows; it worsens. A bank with a hundred million clients is stronger than one with a thousand. A man with a hundred million relationships has no relationship at all.

For a long time, those two substances lived inside the same body, and the institution could present itself as though it were both — large and intimate, solid and close, the marble of the columns and the handshake of the manager. It was a sustainable fiction as long as technology could not separate them. Now it can. And, separated, each runs to its natural side. The balance sheet runs toward scale, toward the four largest, toward concentration. Trust runs toward the individual, toward the face, toward the name.

From that separation is born a professional category that does not yet have, in Brazil, the name or the definitive form it will have:

The independent financial operator.

He is not exactly the banking correspondent, nor the tied investment adviser, nor the product salesman, nor the manager who went independent. He is something that contains and surpasses all of those. He is the person who knows the client before he knows the product; who consults different institutions rather than representing only one; who compares rather than pushes; who organizes proposals rather than accepting the first; who reduces the client's questions rather than multiplying them; who records the decisions and follows the consequences; who preserves the relationship over time, across several institutions, because the relationship is his and not that of the institution of the moment.

He is, in short, the money-changer in the square again — the man behind the table, whose reputation is his only asset — except that now he has access to every bank in the world instead of to his own chest of coins. The future of money, curiously, looks very much like its most ancient past. We spent a century locking trust inside institutions. We are spending the coming decade releasing it back to individuals. It is less a revolution than a return.

The Unfinished Rehearsal

Someone will say that this figure already exists. That the independent investment adviser is exactly this, and that the transformation has already happened. The argument is worth examining, because it is half right — and it is in the wrong half that the opportunity lies.

The growth of investment advisers in Brazil is one of the most revealing phenomena of the last decade, and it is the first visible tremor of the separation I describe. In 2016 there were just over four thousand five hundred advisers in the country. By March 2026 there were 27,721 — the number multiplied sixfold. One brokerage, XP, concentrates roughly eighteen thousand of them. And the most eloquent figure lies not in the total but in the behavior: entire offices jump from one platform to another carrying the book with them. One of them migrated from XP to BTG bearing R$2.5 billion under custody and four thousand clients — who crossed the street behind the adviser, not behind the brokerage left behind. Proof, on an industrial scale, of what this letter has claimed from the first line: the relationship was never the institution's. It followed the man.

Up to here, the adviser confirms the thesis. But it is here too that the thesis reveals how incomplete the adviser still is — and why he is a rehearsal, not the work itself.

First, the adviser does not represent the client before several shelves. He represents one. He is tied to a platform — XP or BTG or Safra — and distributes that platform's products. He traded the bank's counter for the brokerage's counter, but it remains a counter. It remains the shelf, not the table. The client who trades the bank's manager for the brokerage's adviser did not become represented; he became represented by another house. Independence, here, is the house's independence from the banks — not the professional's independence from the house.

Second, and more serious: the market has already noticed the old problem being reborn inside the new structure. The advisers, paid by commission on the products they sell, went back to offering what pays them most, not what serves the client — reproducing, inside the supposedly independent model, exactly the conflict of interest once attributed to the bank manager. It became necessary for the securities regulator to issue, in 2023, a rule to force transparency of remuneration, and another to rechristen the trade itself, swapping autonomous agent for investment adviser — as if changing the name could resolve the ambiguity of the thing. It cannot. The first version of independence recreated dependence, in another suit of clothes.

Third — and this is the clearest sign that we stand before a rehearsal, not a destiny — the large banks are buying the independents back. One bought Órama, another bought Guide, and concentration in the distribution of investment products began rising again. The system is trying to re-institutionalize itself. The empire, feeling trust escape, goes shopping to bring it back inside the balance sheet. It is the natural reflex of a structure that has not yet found its definitive form.

What all of this indicates is not that the separation of balance sheet and trust is false. It indicates that it is still at its first and clumsy attempt. The investment adviser is the draft of the independent operator — the proof of concept that the book follows the man, accompanied by the demonstration of everything still missing for the man to be, in fact, independent. What is missing is representing the client before many institutions, not one. What is missing is remuneration that does not betray the advice. What is missing is the capacity to say do not buy. What is missing, in a word, is the infrastructure that would let independence be more than a marketing promise.

The true independent operator has not yet arrived. But the space for him has never stood so clearly empty.

The Moral Tension

One must face the conflict head-on, because an idea that ignores its own abysses does not deserve to be taken seriously, and this one has an obvious abyss. The word independent is easy to say and hard to earn. It can designate a man who serves the client — or a commissioning machine disguised as advice. The difference between the two is not in the rhetoric. It is in three demands, and whoever fails to meet them has no right to the name.

Independence without multiple options is merely marketing. A professional who calls himself independent but has only one shelf to offer is a salesman with a pretty adjective. Independence begins in the concrete capacity to compare — to lay several banks' proposals on the table and choose, with the client, the best. Without that, independent is a word in the service of a counter.

Independence without transparency of commission is merely disguised distribution. If the client does not know how much the operator earns, and from whom, then there is no way to tell advice from interest. The operator who hides his own remuneration asks to be trusted at exactly the point where he refuses to earn trust. Transparency about one's own earnings is not a compliance detail; it is the ethical foundation of the category. Whoever represents the client should be able to say, without embarrassment, how much he receives and why — and why, even so, that recommendation makes sense for the client and not merely for himself.

And independence without the capacity to say do not buy is merely a new way of selling. This is the ultimate test, the hardest of all. The bank manager could never tell the client that the best decision was to do nothing, because doing nothing generated no revenue. The operator earns the category of independent only when he is able to look at the client and say: right now, the best thing for you is to buy nothing at all. Whoever cannot say that does not represent the client. He represents the next close.

Those three demands share a common center, and the center is an old idea, almost forgotten in the contemporary financial vocabulary: that to represent someone is to answer for him personally. The independent operator I describe is not defined by the technology he uses nor by the institutions he accesses. He is defined by the order in which he does things. He begins with the client — with the real situation, the concrete problem, what serves that specific person — and only then searches, among the products the world offers, for the one that answers the problem. The salesman does the opposite: he begins with the product he needs to move and searches, among clients, for the one who can be convinced to buy it. The difference between the two is the whole difference. And it lies not in the speech. It lies in the direction of the arrow.

There is something here that can only be called aristocratic, in the proper and not the vulgar sense of the word — not that of inherited privilege, but that of assumed responsibility. The aristocrat, in the best acceptance, is the one who answers for his own name, who has something to lose in his own reputation, who would rather say no than say a yes that would shame him. The independent operator who earns that name has, in his relationship with the client, precisely that posture: his name is at stake, his reputation is his only capital, and it is exactly for that reason that he can be trusted in a way an institution — protected by its scale and its anonymity — never can. The institution does not blush. The man does. And it is the possibility of the blush — responsibility with a face — that makes trust possible.

The Table That Does Not Yet Exist

Why, then, has this figure not yet arrived in its full form, if all the pieces are on the board?

Because the pieces are on the board, but not on the same table. That is the whole answer, and it points to exactly where the missing thing lies.

There are the salesmen, who have clients. There are the experienced bankers, who have technical knowledge and know the workings of the institutions from within. There are the brokers and advisers, who have already won the trust of their books. There are the institutions, which have the products, the capital, the license, the balance sheet. Each of those pieces exists, and exists in abundance. What does not exist is the place where they meet — the common table on which an independent professional can set, at the same time, the client on one side and the institutions on the other, and operate between them.

The infrastructure is missing. What is missing is the equivalent, for the independent operator, of what the bank branch was for the manager: the set of systems, processes, access, and tools that would let him serve the client with the competence of a bank without belonging to any. What is missing is the place where there fit, in a single operation, several banks' proposals, the comparison among them, the documents, the decision, the record, the follow-up — everything the independent professional today must reassemble alone, for each client, out of pieces of tools that were not built for him.

Stated with the greatest precision prudence allows: what is missing is the infrastructure that would let an independent professional operate different institutions without losing his own identity, his own book, and his own judgment. What is missing is the table. Not a bank's shelf, nor a brokerage's platform, which merely transfer the dependence from one counter to another. What is missing is the neutral structure — one that does not manufacture products to push, and that exists to serve the operator who serves the client, in the same way electricity serves the house without having an opinion about how you light it.

When that infrastructure exists, one thing will change, and the change will be silent but total: leaving a bank will stop meaning abandoning the financial system. It will mean only ceasing to represent a single institution. The professional who today hesitates to go independent because he fears losing the structure only the bank offers will discover that the structure can exist outside the bank — and that, once outside, he no longer has to represent a shelf. He can represent the client. And the client, who today follows his manager from bank to bank in the hope of continuing to be well served, will discover that there exists someone whose job is precisely to belong to no bank — and who, by belonging to none, can at last be on his side.

Note that I have announced no name, no acquisition, and no date. It is not an omission. It is that the thing matters more than the announcement, and the thing is this: the table that does not yet exist will exist. And whoever builds it will not be entering a market. He will be creating a category.

Who Deserves the Client

It is worth ending where I began — on the wrong side of the table where the Brazilian entrepreneur still sits.

The fastest-growing segment in the country is precisely his. High-income investors — individuals with meaningful wealth, among whom the entrepreneur is a central figure — closed 2025 with R$3.13 trillion invested, up 21.2% in a single year, accounting for more than a third of everything Brazilians have invested. It is the segment that grows the most, that matters the most, that is most fought over. And it is, at the same time, the worst represented — because it continues to be served by the logic of the counter, by the shelf of one institution at a time, by a manager or an adviser who begins with the product he needs to sell and not with the problem he needs to solve.

The industry itself, faced with those numbers, asks the right question aloud. After observing the explosion of the high-income segment, the market's analysts ask: what must one deliver to deserve this client? It is the exact question. And the answer the old structure cannot give, because it runs against its nature, is this: to deserve this client, one must stop selling to him and begin representing him.

Think of the typical Brazilian entrepreneur — not the textbook one, the real one. He is rich in illiquid assets: a property, a company, a plot of land, a warehouse, an inventory. And he is, at the same time, poorly served by the financial system, because the system does not know what to do with wealth that will not fit into an investment product. When he needs money, he sits across the counter from a bank, alone, and receives a proposal he has nothing to compare it with. The bank treats him as a risk to be priced, not as a relationship to be won. And he accepts — not because the proposal is good, but because he has no one to contest it, no one to set other banks competing, no one to sit on his side of the table and turn an imposition into a competition.

That entrepreneur does not need one more bank. He already has six, as it happens. He needs someone whose interest is aligned with his — who does not manufacture the product he offers, who does not gain by pushing it in one direction, who can say do not buy, and who has, behind him, the infrastructure to set the institutions competing for him rather than leaving him to compete for them. He needs someone who does to the banks exactly what the banks have always done to him: forces them to compete. Turns the table.

This is the client that the separation of balance sheet and trust sets free. Throughout recent history, he was the institution's property — not because the institution deserved him, but because there was no alternative. The alternative is being born. And when it matures, the entrepreneur will discover that there exists a place outside the banks that represents him better than any bank ever represented him from within.

The End of a Union

For centuries, to own the balance sheet meant to own the client. The two seemed inseparable, and the institution that held one judged, with good practical reason, that it held the other. That union is ending — not by decree, not by rupture, but through the slow and irreversible erosion of the transaction costs that sustained it. Coase explained why it formed. Technology is explaining why it comes apart.

The institutions will go on existing, and they will go on being powerful. They will go on keeping the money, taking the risk, holding custody of the assets, settling the operations, complying with the regulation — everything that requires scale, capital, and a machine too large to be personal. The balance sheet will remain institutional, and more concentrated than ever. In that, the prophets of the death of the banks are flatly wrong.

But trust will migrate. It will migrate to whoever is willing to know the client by name, to compare alternatives rather than push products, and to answer personally for his own judgment. It will migrate from the marble buildings to the men who sit at the table. It will migrate from the logo to the face. The balance sheet will remain institutional. Trust will be, more and more, individual.

And here is the consequence almost no one has drawn yet, and which is worth stating plainly: the financial future will not be dominated by smaller banks. We are not witnessing the emergence of leaner institutions competing with the large ones. We are witnessing the emergence of operators who will be, in the relationship that matters most — the relationship with the client — larger than the banks they represent. The bank will supply the capital; the operator will hold the trust. And between the two, trust is the scarcest asset, the hardest to build, and the only one that cannot be bought with a balance sheet. Whoever holds it holds what matters. The empire keeps the pipes. The operator keeps the face at the door.

There are changes that begin with technology. Pix, Open Finance, the interfaces that separated the infrastructure from the relationship — all are necessary conditions, and none is sufficient. For there are other changes, deeper ones, that begin with no technology at all. They begin on the day an entire profession realizes it no longer needs to ask permission. That the structure which bound it to the institution can exist outside it. That the client was always his. That the bench in the square can be reassembled.

I have watched this change for years. I have described its contours, measured its figures, followed its false starts and its first clumsy attempts. I have come to the conclusion that it is inevitable, that the only missing element is the infrastructure, and that whoever builds it will not be competing in a market but founding a category.

I confess it no longer satisfies me merely to watch it. I am now more interested in building the infrastructure it requires.

The bank will become infrastructure. The operator will become the institution.

Leo Bentier

Eduardo Leite. The Statesman.

An open letter to those who still mistake noise for leadership — and to the Brazilian entrepreneur who complains about the country without observing the only operator who runs a state the way one runs a balance sheet.

I will start by irritating you. The truth almost always arrives dressed as an offense, and whoever takes offense easily is usually the one who most needs to listen.

Brazil has 513 federal deputies, 81 senators, 27 governors, a swarm of ministers, secretaries, and city councilors. And, among all of them, exactly one statesman.

One.

You will probably disagree by reflex, before finishing the sentence. Good. The reflex is the problem. It is the reflex, not the argument, that chose the leaders you have. A country does not get the rulers it deserves; it gets the rulers its attention span selects. And your attention span, today, is entirely devoted to the circus — the brawl, the meme, the ten-second clip, the outsourced indignation — while the only man who treats the public machine like a company that can go bankrupt passes unnoticed at the corner of the stage, doing the one thing nobody films: administering.

His name is Eduardo Leite. And this letter is about why I believe, with uncomfortable conviction, that he is the only living statesman in Brazilian politics — and why the entrepreneurs of this entire country, from Oiapoque to Chuí, should stop watching the foam and start watching that name.

I am not asking you to agree. I am asking you to think. They are different things, and the second is far rarer.

I. What a statesman is (and why almost nobody is one)

Before anything else, the word needs defining, because it has been worn out by misuse. In Brazil, "statesman" became a funeral compliment: it is applied to any dead politician or to any politician who gives pretty speeches about "the Brazil we want." In other words, it became an adjective. And the adjective is the cheapest currency in existence. Anyone can print it.

A statesman is not an adjective. He is an internalized structure of incentives.

Let me explain with a distinction I have carried for years, one that resolves almost all of Brazilian politics in a single cut.

The politician optimizes for the ensemble average of applause. He wakes up thinking about today's photo, tomorrow's headline, this week's poll. He lives in the spatial average: how many people like me right now, all added up at once. It is a game of instant popularity, and instant popularity is the easiest metric to defraud that humanity has ever invented — just promise what cannot be delivered and postpone the bill until after you have left office.

The statesman optimizes for the survival of the public thing across time. He thinks in the time average: will the state I govern be alive, solvent, and standing ten years from now, even if it costs me dearly today? It is a completely different game, played on a completely different clock. And here is the cruel detail that statistics teaches and politics ignores: the ensemble average and the time average are not the same thing. A system can have a lovely average popularity and still walk straight into ruin — because average popularity does not feel the fiscal hole opening beneath it. The politician who only watches the ensemble average is a man crossing a minefield by calculating the average depth of the craters. The average looks great. He is dead.

The statesman knows tail risk exists. He knows that the improbable, when it happens, does not ask permission — it arrives and collects everything at once. And that is why the statesman does the only thing that matters in a fragile republic: he builds robustness in calm times so the state survives the day the storm comes. He does not know when it will come. Nobody does. But he knows it will come, because it always does, and he prepares the structure to absorb the blow instead of trying to guess the date of the blow.

Repeat this with me, because it is the spine of everything that follows: the point is not to predict the black swan. The point is to be robust to it.

By that definition — and it is the only definition of statesman that survives five minutes of scrutiny — Brazil has hundreds of politicians and one statesman. And the statesman is a man from Pelotas, son of two university professors, who became the city's youngest mayor at 27 and the country's youngest governor at 33, and who spent the last several years doing exactly this: turning a state that was the portrait of Brazilian fiscal fragility into a state capable of absorbing two consecutive black swans without breaking.

Keep that sentence. We will return to it when it hurts more.

II. Skin in the game

There is one test, a single test, that separates those who lead from those who merely talk. The test is: does the man pay for his own mistakes? Does he have skin in the game? Or does he socialize the risk of his own decisions and privatize the glory, leaving the bill to others?

Brazil is an industrial machine of people with no skin in the game. It is the bureaucrat who was never fired for any catastrophe he helped create. It is the congressman who votes for the "goodies package," collects the applause, and leaves before the invoice arrives. It is the commentator who gets every prediction wrong and remains a commentator, because erring without consequence is the most profitable business model in the country. These people talk. They talk a lot. They talk because talking is free for those who do not pay for what they say.

Now observe what Eduardo Leite did, and observe how each thing cost him dearly — on purpose.

He arrived at the Piratini palace in 2019 inheriting a state that did not service its own debt except under court injunction, that paid civil servants' salaries in installments, that delayed transfers to hospitals, that did not invest. A state that, in the previous 50 years, had closed its accounts in the black in only seven fiscal years. Seven out of fifty. That is not a state in financial difficulty. That is a state whose natural condition is the red, like a chronic fever the population came to mistake for normal temperature.

And what he did was touch everything that is not to be touched.

He passed the administrative reform, merging departments and cutting positions. He passed the state pension reform, touching the retirement and benefits of the civil service — the third rail of politics, that high-voltage wire most governors look at and decide to walk around for the rest of their term. He touched the teachers' career plan. He approved a spending cap freezing expenditure for ten years. He struck down the constitutional requirement of a plebiscite to privatize state companies — the device that, in practice, made the inefficient state company eternal — and then privatized what he promised to privatize: CEEE Distribution, CEEE Transmission, CEEE Generation, Sulgás, and put Corsan under concession. He ended the border tax. He reduced the base ICMS rate in a state that needed every cent of revenue.

The reform package was nicknamed by the unions the "death package." Consider the weight of that expression. It is not a compliment. It is an accusation loaded with anger, and he carried it on his back, on his face, in the polls. He did it anyway. He did it because the alternative — the cowardly postponement, the "let's see next year" — was exactly what had taken the state to seven years in the black out of fifty.

And now comes the part that separates the man from the professional politician. In 2022, at the peak of his first term, with approval ratings placing him among the best-evaluated governors in the country, Eduardo Leite resigned his own office to compete for his party's presidential nomination. He bet everything. He left power on the table. And he lost the primary. Lost to João Dória. He stood, for a moment, empty-handed — no government, no candidacy, no consolation.

Stop and absorb that, because it is exceedingly rare. He had real downside. He risked his own political capital on a move that could zero it out, and for an instant it did. Then he came back, ran for Piratini again, and won — becoming, incidentally, the first reelected governor in the history of Rio Grande do Sul, in a state where, since reelection became possible in 1997, every governor had tried and none, not one, had succeeded. But the point is not the victory. The point is that he went to the board knowing he could lose everything, and went anyway.

And there is a detail of origin that stitches all of this together. He did not come from the political aristocracy. He came from Pelotas, son of two professors, with no inherited surname, no godfather opening doors. He built from zero: city councilor, council president, mayor, governor. Each step was contested at the ballot, in the arena, at risk — not handed on a platter. This matters because whoever builds step by step knows the real cost of every decision, unlike the one born at the top who treats power as a birthright. A man who climbed on his own risk has a relationship with downside that the heir will never have. He knows what it is to stand empty-handed, because he has — in the lost primary of 2022. And whoever has stood empty-handed and come back does not confuse power with vanity; he confuses it with responsibility, which is the only healthy confusion a ruler can make.

You, entrepreneur, should recognize this in your own skin. You know the difference between the one who opines on your business from outside, without a cent at risk, and the one who puts his own capital on the table and sleeps badly over payroll. The first category is infinite and worth zero. The second is rare and worth everything. A statesman is someone who put his own capital — political, reputational, personal — into decisions that could destroy him, and chose the correct decisions over the popular ones.

Do not tell me a politician's opinion. Show me what he risked. Leite risked. The bill came, he paid it, and the state remained standing. That is the definition of skin in the game, and in the wasteland of Brazilian politics it is nearly an endangered species.

III. The swan that came in May

Now we need to talk about the flood. And we need to talk about it with brutal honesty, because it is exactly here that the intellectual laziness of the whole country reveals itself — and exactly here that you learn to separate consequence from blame.

In May 2024, Rio Grande do Sul was hit by the greatest meteorological tragedy in its history. In some places it rained more than a thousand millimeters in an extremely short interval — volumes with no parallel in the state's records. More than 400 thousand connection points lost power. Nearly two hundred people died. Around 2.4 million were affected. Close to 900 thousand had to leave their homes; more than a hundred thousand passed through shelters. Entire municipalities submerged. An international airport underwater. A state of war with no enemy army — only the water.

And then the mob did what the mob always does in the face of catastrophe: it looked for a head. It needed a culprit with a taxpayer ID, because the idea that nature, in its fat tail, is indifferent to our political narratives is unbearable for those who need a villain by the weekend. So the bill was delivered to the governor. "Leite caused the flood." "Leite is responsible."

Let us use our heads here, because that is what they are for.

Nobody causes a thousand millimeters of rain. That is not an opinion, it is meteorology. Blaming a man for precipitation is the same as blaming the captain for the size of the wave. The wave is the wave. The tail event is the tail event — improbable, extreme, and when it arrives, it collects everything at once. Attributing the authorship of the rain to a governor is not analysis; it is the psychological need to turn the random into the intentional, because the intentional we hate and the random we fear. And hating is always easier than fearing.

Second point, and this one is subtler, so pay attention. Porto Alegre's protection system — the dikes, the floodgates, the pump houses — is not a work of the Leite government. It is a structure from the 1970s, raised after the floods of 1941 and 1967, designed to hold the water up to six meters. In May 2024, it failed with the level still at four and a half meters. Why did it fail? Because of the decay accumulated over decades, across successive municipal and state administrations, governments of every color. The capital's own city hall had been warned in 2018 — before, therefore — about the risk of pumping failure. And the maintenance of the local protection system is, by law, primarily the responsibility of the municipality, not the state.

In other words: what collapsed in 2024 was a half-century infrastructure debt, distributed across dozens of administrations, with legal responsibility concentrated at the municipal level. Hanging that on the neck of a single governor who took office in 2019 is intellectually dishonest. It is like blaming the last tenant for the crack the building has been accumulating since its foundation.

This is what I call separating consequence from blame. The rain was consequence — of climate, of the tail, of the improbable. The decay was consequence — of a country that, for generations, prefers inaugurating to maintaining, prefers the new photographable work to the old invisible dike. Neither of these things is Eduardo Leite's individual fault, and whoever pretends otherwise is doing militancy, not analysis.

But — and now comes the point that will make you understand why I call this man a statesman, and not a lucky man — what matters in a black swan was never the prediction. It was the robustness. And this is where the story turns inside out.

IV. The antifragile

What makes Eduardo Leite a statesman, and not merely a good manager, is this: the fiscal work he did before — without knowing a pandemic was coming, without knowing a flood was coming — was precisely what gave the state the capacity to respond to both.

He has said it himself, and the sentence is so exact it almost seems stolen from a treatise on fragility: everything we did beforehand, without knowing pandemic and flood were coming, gave the state a response capacity that was fundamental in those two critical moments. Until a year before the pandemic, the state was delaying payments to hospitals. Good thing that, in the pandemic, the accounts were already in order to face it. And again, in the flood, the state confronted the greatest climate disaster in its history with cash to react.

That is antifragility in its pure state, and it is the most important concept in existence for anyone living in a fat-tailed country like ours.

Let me be precise, because the difference is everything. Fragile is what breaks under the shock. Robust is what withstands the shock. Antifragile is what organizes itself in the absence of the shock so as to have response capacity when the shock arrives — it is the man who builds the reserve, the optionality, the slack, without knowing which specific disaster it will serve, knowing only that some disaster will come. Slack looks like waste until the day it is the only thing between you and the abyss.

Look at the numbers, because the numbers here are not ornament, they are the proof. The state that closed its accounts in the black in seven years out of fifty went on to accumulate five consecutive positive budget results — from 2021 to 2025 — with a surplus of R$ 2.67 billion in the last fiscal year. The privatizations added up to roughly R$ 8.5 billion, reinjected into the state. Public investment jumped from 2.3% of Net Current Revenue in 2019 to 10.7% in 2024. The state went back to paying court-ordered debts at record pace. Went back to being able to invest. Went back, in short, to having cash — the most underestimated and most decisive word in the public vocabulary.

And then, when the water rose, there existed Funrigs, the reconstruction fund, with R$ 14 billion mobilized across 227 projects. But — and this is the sentence that closes the argument — Funrigs only exists because the state rebuilt the capacity to pay its own debt. A broken state does not create a R$ 14 billion reconstruction fund. A broken state stands in line at the minister's office begging for alms, and prays. The fiscal robustness built in the calm of 2019 to 2023 was the ammunition of the response in the war of 2024. Without the first, the second would be a speech.

And it was not only the spreadsheet. The same robustness appeared in a dimension the entrepreneur feels every day in his operation: security. Between 2015 and 2024, vehicle theft in the state fell 87%. Cargo theft — the plague that bleeds the logistics of those who produce and transport — plummeted 90% since 2017. Street robbery receded 78% over the same interval. The security forces reached their largest headcount in years. You may debate the methodology, debate how much is national trend and how much is state management — debate it, it is healthy. But the pattern is unequivocal: a state that was institutionally adrift became a state that works, and "works" is the rarest and most expensive word in the Brazilian public sector. Fiscal robustness, logistical robustness, security robustness — they are faces of the same verb. It is a state being rebuilt the way one rebuilds a company: from the capital structure to the operation, from the cash to the delivery.

Notice that none of this makes an exciting headline. A 90% drop in cargo theft makes nobody cry on the chamber floor. A surplus has no soundtrack. That is why these things are invisible to the machine — and that is why, exactly why, they are the portrait of a statesman and not of a stage performer. What matters is rarely photogenic. What is photogenic rarely matters.

That is why I am not impressed by the politician who promises to rebuild after the disaster. Promising to rebuild afterward is easy; it is the easiest thing in the world, because the audience is desperate and applauds any promise. What is hard, what is rare, what is statesmanlike, is having done the boring, unpopular, invisible work years before — the cut that earned the nickname "death package," the privatization that earned lawsuits, the reform that earned strikes — so that, when tragedy struck, the state had something to react with.

Robustness is boring until the day it is everything. And the statesman is the only one who does the boring work when nobody is watching, because he understands that the value of slack only appears in the tail, and the tail always comes.

V. The mistakes. Without cowardice.

Now I will do what the hagiographer does not do and what the militant is incapable of doing: I will point out the mistakes. Because the strongest defense of a man is the one that concedes the real flaws and demolishes the false ones. Whoever only praises is selling. Whoever praises and criticizes is thinking. And, ironically, it is the honest criticism that gives the praise its weight — the rest is advertising.

I will state the mistakes as a friend states them, not as an enemy shouts them.

First: the Environmental Code. In 2019, in his first year, Leite altered the state's Environmental Code — the code conceived from the work of José Lutzenberger, a world reference in ecology, whose formulation took nearly a decade of debate. The reform touched hundreds of provisions. Technical staff at Fepam and researchers at UFRGS's hydraulic research institute accused the project of weakening licensing, above all through more automatic licenses in cases assessed as low-risk. Leite answers that there was no flexibilization, but rather an update to align the state norm with the federal one — and there is technical merit in that argument; part of the criticism confuses licensing with a direct cause of the flood, which does not hold. But here is the criticism that does hold: even if the alignment was technically defensible, loosening the perception of environmental rigor in a state walking toward a climate catastrophe was an error in reading tail risk. The statesman I described in the previous sections — the man who understands that the improbable collects everything at once — should have been, on this subject, the most conservative of governors, and he was less. A statesman owns the tail even when the causality is debatable, because the asymmetry commands it: the cost of having been too rigorous was small; the cost of having been lenient at the wrong time is enormous. He bet on the wrong side of that asymmetry. It is a mistake. I say it in full.

Second: the tone. During the crisis, there was the pump-house episode — the moment he asked, with irritation, whether people really thought the flood's problem had been two pump houses, in a state where millions of cubic meters had rained down. From an engineering standpoint, he was right: two pump houses do not hold back a biblical deluge. He was technically correct. And he was politically wrong, because the statesman does not win arguments, he holds the room. In a moment of collective pain, being right and cold is a way of being wrong. The population did not need an argument; it needed a commander who would absorb the anguish instead of rebutting it with technical precision. Winning the debate and losing the temperature of the room is a mistake a man of his intelligence should not repeat. It is an error of temperature, not of reasoning — but it is an error.

Third: the asymmetry of his own discipline. The same surgical rigor he applied to the balance sheet — the same horror of waste, the same obsession with long-term sustainability — took too long to arrive, with the same intensity, at the state's physical resilience. Prevention, the dikes, the warning systems: that invisible infrastructure received, for years and across many administrations including his, a fraction of the attention the spreadsheet received. It is not exclusively his fault — we have established that the decay is a half-century debt and above all a municipal responsibility. But the statesman who masters the art of investing in invisible fiscal robustness should have transported that same art, earlier, to invisible physical robustness. The two are the same discipline applied to different matters. He got one right before getting the other right.

Did you notice something? None of these three is the mistake of a crook. They are the mistakes of a builder — of someone who was, the whole time, trying to do something difficult and got most of it right. The thief's mistake tells you the man's character. The builder's mistake tells you only where he can still improve. And there is a sign of learning in real time: today the state runs the Plano Rio Grande with a scientific committee on climate adaptation, runs preparedness programs for extreme events, faces El Niño with structure instead of improvisation. The man is correcting course while flying, which is the only honest way to correct course. A statesman is not the one who does not err. He is the one who errs, pays, learns, and rebuilds more robust than he was.

VI. The antagonist is not him

And here, my dear reader, is where I need you to feel a little anger. Good anger, anger that thinks. Because by now you must have noticed the obvious question hovering over all of this:

If the man is all that, why is he not president? Why did he not even come close? Why, in 2026, was the only name at the center with a service record, with a balance sheet in the black, with two black swans absorbed and survived, simply not in the race?

The answer is the most revealing thing in existence about Brazil, and the antagonist of this story — pay close attention — was never Eduardo Leite. The antagonist is the machine.

Let me tell you how the machine works, because it is a lesson.

Leite positioned himself as the third way — the only one, by his own account, who embraced neither Lula nor Bolsonaro in 2022. He left the PSDB, went to the PSD, where the party boss gathered three well-evaluated governors on the same shelf. And when the time came to choose, the machine chose — by the reading of its own analysts — not the most competent, but the one who pulled the most proportional votes. The logic was not "who would govern Brazil best." The logic was the drag: a ticket head who drags deputies, who fattens the caucus, who enlarges the party fund, who guarantees bargaining power with whoever wins. The object of the game, in the end, was never the presidential palace. It was the size of the caucus.

Absorb what that means. The Brazilian political machine looked at a statesman — using the word in the rigorous sense I built in this letter — and discarded him not because he governed badly, but because governing well does not pull votes. Competence has no electoral drag. Fiscal prudence does not become a meme. A surplus does not climb the algorithm. And so the third way was buried in 2026, not for lack of merit, but for an excess of merit in the wrong kind of currency.

And there is another layer, because the machine is not only the party boss. It is also the class that measures a man by the wrong metrics. It is the analyst who evaluates Eduardo Leite by his rejection rate, by the poll column, by the percentage of people who "do not know him" — and never, ever, by the balance sheet of the state he administers. I have an affectionate name for this class: they are the intellectuals-yet-idiots. People with sophisticated vocabulary and zero skin in the game, who can recite the voting-intention ranking by heart and would be incapable of closing a single quarter in the black of anything whatsoever. This class looks at a man who turned around 50 years of deficit and asks: "but does he blow up on social media?". It is the question of someone who has never administered anything beyond their own opinion.

And — I will be direct with you now, because I have been polite this far — the machine is also you.

Yes, you. The entrepreneur who complains about Brazil at dinner, who curses the tax burden, who swears "this country is hopeless," who forwards the outraged voice note to the family group chat. You, who understand perfectly what antifragility in a balance sheet is, who know in your flesh the difference between cash and promise, who would never confuse the man who risks his own capital with the one who only opines from outside — you looked at the only political operator who applies those exact principles to a state, and did not look. You were distracted. You were watching the circus. You were, like everyone, cheering for one side of the penalty kick while the only man who knows how to assemble the whole team passed unnoticed.

Maybe — and here is the provocation I promised, served without anesthesia — maybe Brazil does not deserve a statesman. Maybe you do not deserve one. A country gets the leaders its attention selects, and our attention, collectively, selected the noise. We elected the algorithm. We rewarded the drag. We retired competence for lack of engagement. And then we sat down to complain about the rulers we ourselves, with our screen time, helped produce.

The antagonist of this letter is not a man. It is a mirror.

VII. Why the entrepreneur must watch this name

So let me close by telling you, with the frankness this whole letter has earned, why you — entrepreneur, capital allocator, builder of things in a fat-tailed country — need to watch the name Eduardo Leite more closely. And why this has almost nothing to do with voting.

I am not asking you to vote for him. Your vote is your business, and the office he will or will not hold in 2026, in 2030, in 2034, is a variable not even he controls. What I am asking of you is pattern recognition — the only skill that actually compounds wealth over the long run.

Because what Eduardo Leite represents is not a candidacy. It is a case study in the only thing that survives in a fragile country: robustness built in the calm to be spent in the storm. He took a chronically ill financial entity — a state that only knew how to lose money — and made it capable of absorbing two consecutive tail events without breaking. He did it paying the political price of unpopularity, with all his skin in the game, risking his own capital to the point of zeroing it for a moment. He erred where he underestimated the environmental tail and where he let the temperature of the room slip, and he is correcting it while flying. And he was discarded by the machine not for what he did wrong, but for what he did right in the currency the electoral market does not know how to price.

If you understand antifragility in a balance sheet, you have the intellectual obligation to study the only politician who applied it to a state. Because the qualities that make him a statesman are not fashionable — they are Lindy. They are old, tested, and will keep mattering long after 2026, regardless of whatever office he holds or does not hold. Fiscal prudence is Lindy. Backing the right decision against popularity is Lindy. Building slack before needing it is Lindy. These things cross time because they work, and they will work in any country, in any decade, under any party label.

There is one last asymmetry worth keeping. Watching this name costs you nothing — no capital, no vote, no commitment. Ignoring it may cost you the only thing that does not come back: the lucidity of having recognized, early, the pattern the rest of the country only understands late. Cost of watching: zero. Cost of not watching: possibly everything. When the asymmetry is that one, the rational man does not hesitate. He looks.

Betting on a statesman is betting on time. It is the bet contrary to the entire country's — which bets on the instant, the clip, the drag — and precisely for that reason it is the bet that, historically, pays. The serious entrepreneur never bet on the noise. He bet on the structure beneath the noise. And, in Brazilian politics today, there is exactly one man who is structure, not noise.

Watch this name. Not because I told you to. Because you, of all people, are the one with the repertoire to recognize what you are seeing. You have seen this before — in a founder who cuts into his own flesh when it is unpopular, in an allocator who holds cash while everyone levers up, in an operator who builds robustness that looks like waste until the day it is the only thing left standing. You already know how this story ends. You just had not noticed that, this time, it is being written at the Piratini palace.

Brazil has 513 deputies, 81 senators, 27 governors, and one statesman.

I have already told you who he is. The rest is your attention.

If this letter irritated you, good. Irritation is the exact distance between what you believe by reflex and what you believe after thinking. The whole country fits inside that distance.

Leo Bentier

Administered poverty and the gravy train in a suit

Brazil distributes dependency downward and privilege upward. The same ruler must measure the poor man's card and the powerful man's legal opinion.

I published a provocation:

A benefit with no exit door becomes a leash.

The reaction came fast. I was told I should also talk about the tax exemptions granted to the powerful. That the "gravy train" only bothers people when it is used to fight hunger.

It is a good objection.

Not because it invalidates the criticism of Bolsa Família. Because it widens the field of the accusation.

Brazil distributes dependency downward and privilege upward. It hands benefits to the poor, tax expenditures to the organized sector, favored credit to the well-connected group, and tenure to the caste that administers it all. Each social class receives its own private version of the protective State. The name changes according to the recipient.

For the poor, assistance.

For the connected businessman, incentive.

For the public-sector corporation, acquired rights.

For the politician, governability.

The bill arrives, without euphemism, at the taxpayer left standing outside.

Unemployment insurance is no substitute for Bolsa Família

Let us start by clearing a bad claim off the table.

Unemployment insurance was not created to do the job of Bolsa Família. It temporarily protects formal workers dismissed without cause and depends on prior requirements of employment history and earnings. A poor family may never have held a formal job. It may live off intermittent work, informality, odd jobs, or rural income. It may be made up of children, the elderly, and adults whose income never came close to generating a right to the insurance.

Bolsa Família operates on a different criterion: household income per person. In 2026, the entry door remains a monthly income of up to R$ 218 per family member. Having a formal job, being a registered microentrepreneur, or receiving other income does not automatically eliminate eligibility. (Portal Gov.br)

So saying that unemployment insurance would suffice is confusing the temporary loss of a formal job with structural poverty.

This needs to be said bluntly, because a criticism built on a false premise makes the lazy defense of the entire program easier. Knock down one exaggerated sentence and you can pretend every incentive has been absolved.

They have not.

The "laziness effect" is too comfortable an explanation

Nor is there solid ground for claiming that Bolsa Família, in general, makes its beneficiaries abandon work.

Studies by Ipea found no relevant effect on the probability of men and women working; later research also disputes the simple narrative that the program produces a widespread withdrawal from the labor market. (Ipea)

This does not mean perverse incentives never exist.

It only means that millions of people cannot be reduced to the caricature of loafers who traded prosperity for a few hundred reais. The benefit is far too small to finance a comfortable life. For many people, it tops up a precarious income, reduces hunger, keeps a child in school, or stops a household emergency from becoming a catastrophe.

Real poverty rarely looks like the poverty imagined by those who observe it from a distance.

It is fragmented. One week there is work; the next, there is not. A child gets sick. A day's wage disappears. The rent comes due with Swiss precision, while income arrives with Latin American punctuality.

The mistake is concluding that, because the aid is necessary, its design has become immune to criticism.

A bridge can save someone from a river and still have been badly built.

The leash is not in the benefit. It is in the fear of the exit

The government knows the problem.

The so-called Protection Rule exists precisely because cutting the benefit the moment income rises would create a cliff. A family could accept a job and, in doing so, immediately lose a secure income in exchange for one still fragile.

Since June 2025, families who cross the entry threshold may, under certain conditions, keep receiving 50% of the benefit for a transition period. The Ministry of Social Development itself explains that overcoming poverty does not happen automatically with the signing of an employment contract. (Ministry of Social Development)

The existence of this rule is an important admission: benefits alter decisions.

Every system does.

Taxes alter decisions. Subsidies alter decisions. Public guarantees alter decisions. Favored credit alters decisions. Job tenure alters decisions. Even the expectation of a future amnesty alters decisions.

The adult debate begins when we stop asking whether there is an incentive and start asking which behavior is being incentivized.

A well-designed social policy must make ascent less dangerous. The beneficiary needs to know that accepting work, formalizing income, opening a small business, or growing revenue will not produce instant punishment.

The exit door cannot look like a trapdoor.

When a person fears losing the benefit more than they desire to increase their income, the program has failed somewhere — even if it was indispensable until then.

The government does not need to manufacture loafers

The most serious criticism of welfarism is not that the government gathers people in a room and teaches them to give up.

Governments are rarely that competent.

The risk is more banal: administering poverty can be politically more convenient than overcoming it.

Whoever leaves dependency stops needing the intermediary. They start negotiating wages, opening businesses, building wealth, and making choices that no longer depend on the continuity of a program. Autonomy reduces the power of whoever distributes.

No politician needs to say this out loud.

The incentives speak for him.

The program grows. The registration machine grows. The advertising grows. The budget grows. Every improvement is announced as a government achievement; every threat of change is presented as a personal risk to the beneficiary.

Gratitude, when exploited electorally, stops being gratitude. It becomes fear.

There is no need to buy the vote directly. It is enough to convince millions of people that a political defeat will put food, income, or medicine at risk.

The modern corral requires no fence.

It requires insecurity.

And the exemptions of the powerful?

Now we arrive at the question left in the comment.

Yes, we should talk about them.

The federal government forgoes hundreds of billions of reais a year through tax, financial, and credit benefits. The TCU, Brazil's federal audit court, treats the transparency and effectiveness of these expenditures as a high-risk area and records amounts reaching hundreds of billions, equivalent to a relevant share of public revenue and of GDP. (TCU)

Some tax expenditures make economic sense.

A benefit can correct distortions, stimulate investment, offset regional disadvantages, avoid cascading taxation, or produce a social return greater than its fiscal cost.

Others survive because the beneficiary has an office in Brasília.

Here is the detail that almost always disappears from the debate: the poor must prove poverty. The organized group must prove influence.

One faces registration, queues, reviews, and the risk of having the benefit suspended.

The other hires an economist, a lawyer, a lobbyist, and produces a study explaining why its particular advantage is indispensable to national development.

When the benefit costs R$ 600, they call it dependency.

When it costs billions, they call it industrial policy.

The difference is not always in the economic quality. Sometimes it is only in the vocabulary of the recipient.

There is no automatic virtue in paying less tax

It would also be intellectually cheap to treat every exemption as corruption.

Taxation is not a natural right of the government over everything that exists. When the State stops collecting a badly designed tax, it may not be "giving" money to anyone. It may simply be ceasing to take.

The term "forgone revenue" already carries a small linguistic fraud. It suggests that all income existed first as the moral property of the Treasury and was generously returned to the citizen.

It did not.

The money was produced outside Brasília.

But selective benefits create another problem. If two competitors operate under different rules because one obtained special tax treatment, the State has not reduced its weight. It has picked a winner.

A low, general tax is freedom.

A negotiated exception is power.

The same reasoning applies to social assistance. A transparent minimum income, with predictable rules and a transition toward autonomy, can be civil defense against misery. A system used to cultivate gratitude, fear, and permanence is clientelism.

The instrument carries no built-in morality.

The design reveals the intention.

The hypocrisy of both sides

The right frequently denounces the poor man's benefit while staying silent before the ally's subsidy.

The left denounces corporate tax breaks while treating any social program as a sacrament.

Both protect their clientele.

One calls its own privilege investment.

The other calls its own justice.

The taxpayer pays for both and still has to choose which speech will insult him less.

I have no interest in building a theory in which every poor person is a parasite and every businessman is a hero. There are incompetent businessmen living off state protection, market reserves, and public contracts. There are beneficiaries working every day under conditions many critics would not endure for a week.

The moral divide is not between rich and poor.

It is between autonomy and capture.

Between policies that help someone stand up and systems that prosper by keeping him on his knees.

The question that should end the discussion

Every public benefit should answer four questions:

How much does it cost?

What result does it deliver?

Who loses if it ends?

And, above all: how does the beneficiary stop needing it?

The last question almost never receives enough attention.

A social program that celebrates only entries measures its own expansion, not its own success.

An industrial policy that tallies billions granted, but not additional productivity, exports, wages, or innovation, also measures administrative effort instead of results.

The Brazilian State is fascinated by volume.

Number of beneficiaries.

Amounts disbursed.

Companies served.

Credit granted.

Exemptions approved.

What rarely appears with the same pomp is the emancipation rate.

How many families left the program because they increased their income sustainably?

How many companies stopped needing the incentive because they became competitive?

How many protected sectors finally learned to compete without protection?

Brazilian public policy loves to inaugurate entry doors.

Exit doors threaten the size of the machine.

The poor should not be abandoned. They should be set free.

The answer to poverty is not to abruptly withdraw the benefit and wish people good luck.

Nor is it to turn the transfer into an administrative inheritance.

Aid should reduce hunger today without charging tomorrow's autonomy as the price.

That requires gradual transition, training that corresponds to real jobs, daycare, transportation, security, a serious treatment of informality, and rules that do not immediately punish an increase in income.

It also requires economic growth.

No motivational course beats an economy without openings. No conditionality fixes a city where formal work simply does not exist. No one is emancipated by decree.

At the same time, fighting poverty without discussing productivity, basic education, the business environment, capital, and employment is just handing out painkillers in an infirmary that keeps receiving the wounded.

Bolsa Família can keep someone from dying along the way.

It should not be confused with the destination.

The same ruler

So, to the comment that asked me to talk about the exemptions granted to the powerful, my answer is simple:

I agree.

Let us talk about all of them.

The poor man's benefit and the rich man's benefit.

The income transfer and the tax expenditure.

The social card and the favored credit.

The family that receives a few hundred reais and the company that obtains a few billion in protection.

But we will use the same ruler.

I will not accept the suffering of the poor serving as a moral shield for a badly designed program.

Nor will I accept the word "investment" sanctifying privileges purchased in Brasília.

Public help must have a reason, a deadline, a metric, and an exit.

Without that, only the dependent's outfit changes.

The poor man carries a card.

The powerful man carries a legal opinion.

And both learn to look to the State before looking to their own legs.

A benefit without an exit can become a leash.

An exemption without results can become legalized graft.

Brazil does not need to choose which dependency to finance.

It needs to learn to distinguish rescue from capture.

Leo Bentier

Brazil's Greatest Asymmetry May Not Be in the Stock Market

It may be on the balance sheet of the entrepreneur who built wealth, but still buys money as if he had built nothing.

I have spent much of my life looking for things that were mispriced.

Stocks the market treated as corpses, but that were still breathing.

Companies that looked expensive because present earnings concealed future economic power.

Risks everyone feared, although few had stopped to measure them.

And, in the opposite direction, assets celebrated as inevitable whose prices already required the future to behave with a perfection the future has never shown it possesses.

Investing is, to a large extent, the search for the distance between a thing and the opinion the market has formed about it.

When that distance is large, there is asymmetry.

When it is small, there is only activity.

But some time ago I began to notice a strange distortion.

The entrepreneur spends decades building assets, but when he needs money, he is priced as if none of them counted.

He has real estate.

Receivables.

Inventory.

Equipment.

Contracts.

Relationships.

History.

Reputation.

A company that survived crises, governments, currencies, competitors, taxes, and bad decisions.

Yet he walks into a branch, hands over the company registration number, and receives a rate produced by a table.

Everything he built over twenty years is compressed into a classification generated in a few seconds.

They call this credit analysis.

Sometimes it is only the inability to analyze anything that does not fit the conveyor belt.

The bank is not necessarily wrong.

This point matters.

Banks were not built to understand each entrepreneur deeply. They were built to process large volumes of transactions with enough consistency not to die.

Scale requires averages.

Averages require standardization.

Standardization requires ignoring precisely what makes each company different.

The system works.

The problem is that it works better for the system than for the client.

A credit product must be replicable. It must fit an internal policy, a screen, an approval level, and a committee meeting that cannot last three days for each transaction.

That is why the product begins with a convenient question:

How much can we lend to this company registration number within what we already sell?

But that is not the question the entrepreneur should be asking.

The correct question would be:

Given everything I have built, what is the smartest way to transform wealth, cash flow, and credibility into capital?

The two questions sound similar.

They are not.

One begins on the bank's shelf.

The other begins on the entrepreneur's balance sheet.

One looks for an available product.

The other designs a possible structure.

One prices the client according to one institution's policy.

The other presents the transaction in a way that allows different institutions to price it.

That changes everything.

Money is also a product.

That statement is uncomfortable because banks spent centuries wrapping money in an almost priestly aura. Marble branches, private rooms, managers, acronyms, committees, contracts, and technical vocabulary help maintain the impression that the bank grants some exclusive substance.

It does not.

The real from one institution is worth the same as the real from another.

What changes is the source, the term, the collateral, the priority, the structure, the perceived risk, and the price.

Money is a commodity.

The structure is not.

An entrepreneur can have a good company and a bad debt.

He can have wealth and no liquidity.

He can own a fully paid property while paying obscene interest on working capital.

He can discount receivables every month because no one organized his cash flow for a longer-term transaction.

He can offer too much collateral on a small debt because he negotiated under urgency.

He can accept a high rate not because he represents high risk, but because he arrived at the table alone.

This may be the least discussed part of business credit.

The entrepreneur is almost always alone.

On the other side sits an institution that lends money every day.

It has models, lawyers, analysts, contracts, policies, committees, and the accumulated memory of thousands of transactions.

On the entrepreneur's side there is a man who may make one relevant financial decision every three or four years.

One side sells money professionally.

The other buys money occasionally.

Even so, we pretend the negotiation is balanced.

It is not.

The institution knows which clauses matter.

It knows which guarantees it could accept.

It knows how far it can lower the rate.

It knows what the committee fears.

It knows what must appear on paper for that fear to be reduced.

The entrepreneur only knows that he needs the money.

Urgency is an expensive form of ignorance.

The more urgent the need, the less time there is to organize the story of the transaction, compare sources, redesign guarantees, or refuse bad terms.

That is why the most expensive credit is not necessarily granted to the worst entrepreneur.

It is often granted to the least represented entrepreneur.

The largest companies understood this long ago.

They do not walk into a branch and ask which line is available.

They have a chief financial officer, treasury, advisers, lawyers, banks competing for the mandate, and professionals responsible for presenting each debt as what it really is: a negotiable structure.

They issue.

They securitize.

They link receivables.

They separate risks.

They combine maturities.

They negotiate guarantees.

They exchange short debt for long debt.

They make the creditor compete not only on rate, but on the entire architecture of the transaction.

We call this capital markets when it happens with a huge company.

When it happens with a medium-sized company, we still treat it as an exception.

But the difference between them is not only size.

It is representation.

One has people sitting beside it.

The other has a manager sitting across from it.

For a long time, the financial market convinced entrepreneurs that they suffered from a lack of credit.

Perhaps the diagnosis is incomplete.

Brazil has money.

Banks have money.

Funds have money.

Cooperatives have money.

Securitization companies have money.

Investors look for credit assets.

Institutions look for good transactions.

The problem is not only lack of capital.

It is the distance between available capital and the company that does not know how to turn its economic history into a transaction intelligible to whoever lends.

Capital exists.

What is missing is translation.

The institution sees documents.

The entrepreneur sees his own story.

Between the two there is an abyss.

On one side, the man knows he paid every debt for the last fifteen years, that his property is worth more than it shows in the accounting, that customers keep buying even during crises, and that a specific opportunity would justify taking money now.

On the other, the analyst sees revenue, leverage, score, collateral, sector, and concentration.

Both may be right.

But they are speaking different languages.

This is the space where money becomes expensive.

Interest does not compensate only real risk.

It also compensates the risk that was not explained, the collateral that was not organized, the cash flow that was not demonstrated, and the transaction that arrived incomplete.

Every unanswered question becomes spread.

That is why reducing the cost of capital does not necessarily begin by looking for a lower rate.

It begins by reducing the questions the creditor needs to price.

What is the source of repayment?

What happens if revenue falls?

Which asset supports the debt?

What is the real liquidity of that collateral?

Does the debt term match the investment term?

Who gets paid first?

How will the money be used?

What prevents a temporary difficulty from becoming a permanent loss?

A well-built transaction does not eliminate risk.

It makes risk legible.

And legible risk costs less than confused risk.

Modern finance automated credit execution.

Forms became digital.

Documents can be sent by phone.

Scores are calculated in seconds.

Offers appear in apps.

Contracts receive electronic signatures.

The money arrives faster.

All of this is useful.

But speed does not correct direction.

Putting a bad product in a good app only allows the client to make the mistake without leaving home.

The great transformation in business credit will not come only from approving faster.

It will come from deciding better.

Deciding whether the company should take credit.

How much it should take.

For what term.

With which collateral.

In which entity.

For what purpose.

In what order its debts should be reorganized.

And, perhaps most importantly, when the entrepreneur should hear something no credit seller has an incentive to tell him:

Do not take this money.

That will be the distinction between distribution and advice.

The distributor is paid when money moves.

Whoever stands on the entrepreneur's side must be able to earn trust also when he recommends that nothing move.

It is an economically difficult position.

Perhaps that is why it is so rare.

Twelve years ago I wrote that the bank of the future would be a person.

Not because technology would fail.

On the contrary.

The more payments, investments, analyses, and contracts were automated, the cheaper it would become to execute a financial transaction.

And the cheaper execution became, the more expensive judgment would become.

I still believe that.

But today I would add one thing.

The banker of the future will not merely be someone who knows the client.

He will be someone capable of economically representing everything the client has built before institutions designed to see only parts.

Knowing without being able to structure is friendship.

Structuring without knowing is distribution.

What the entrepreneur lacks lives in the space between the two.

Perhaps Brazil's greatest asymmetry is not in a forgotten stock, a new currency, or a technology company.

Perhaps it is on the balance sheets of millions of entrepreneurs who own valuable assets, real businesses, and histories of survival, but still buy money as anonymous consumers.

A mispriced asset can enrich the investor.

A mispriced debt impoverishes the entrepreneur every month.

I spent years looking for the first kind of distortion.

I have been thinking more and more about the second.

Some ideas ask for a letter.

Others, when they mature, demand an institution.

Leo Bentier

The largest companies in the country do not take credit. They design it.

The price of money is not defined by your risk. It is defined by the structure you accepted.

Two companies. Same sector, same size, same revenue, similar margins.

The first pays 4% per month for the capital that runs the operation.

The second pays 1.2%.

The instinctive reading looks for the difference in the balance sheet. One must be more solid, older, better managed.

It is not.

The difference is not in the companies' risk. It is in the design of the debt. The first bought a product. The second built a structure.

And that distinction — product versus structure — is the line that separates the two floors of the Brazilian financial system.

What the branch sells has product names: working capital, corporate card, receivables prepayment, overdraft.

A product is manufactured in series. A conveyor belt, a score, a rate table. The price embeds the average of everyone who passes through it — the good payer subsidizes the bad one, and the bank's margin is calibrated for the worst case in the batch.

When you accept a shelf product, you accept being priced as the average.

It does not matter what you built. It does not matter the paid-off warehouse, the inventory, the predictable receivables of ten years of operation. The conveyor belt was not designed to read any of that.

It was designed to sell fast.

Now look at the upper floor.

The large companies of this country do not buy credit products. None of them. What they do has another verb: they structure.

Debentures designed to measure. Certificates backed by receivables. Funds that buy the portfolio they originate. Debt secured by real assets, with tenor drawn over the cash flow, with cost defined clause by clause.

None of this is financial magic.

It is a specific technology with a specific function: removing fear.

I have written here that credit, reduced to its essence, is a single question: what happens if you do not pay?

Structure is the art of answering that question before it is asked.

A registered lien answers. A receivable locked into the operation answers. A cash flow routed through an escrow account answers. Each element of structure eliminates one of the creditor's fears.

And each eliminated fear eliminates a piece of the price.

The interest you pay is not the bank's opinion of your character. It is the price of the fear your debt left unanswered.

The company paying 1.2% is not more trustworthy than the one paying 4%.

It simply left no question open.

Here comes the detail that should cause more discomfort than it does.

The tools of the upper floor are not private. The fiduciary lien has been law since 1997. The assignment of receivables, the escrow account, the real collateral — all public, all tested, all available to any company with assets and flow.

What is private is access to the design.

Because designing takes work. It requires actually reading the balance sheet, mapping the asset, understanding the flow, negotiating clauses. The branch does not do that — the branch is paid for speed and volume, and the product conveyor delivers both.

And there is a less comfortable reason: structure transfers power. A well-designed debt gives the debtor predictability, a compatible tenor, and a compressed rate. A shelf debt gives the creditor margin, implicit guarantees, and a client who rolls over for lack of alternatives.

The system does not hide structured credit from the average entrepreneur.

It simply has no incentive to offer it.

The entrepreneur with the paid-off warehouse — the same one from this series of letters — does not need a better product.

He needs a better question.

The branch's question is: "which product would you like to contract, sir?"

The structure's question is: "what have you built — and how does it answer, clause by clause, the fears of the one who lends?"

They are two different conversations. The first ends in a rate table. The second ends in a design where your assets, your flow, and your tenor work in your favor — and the price falls in exact proportion to what the structure answers.

Same company. Same taxpayer ID. Same risk.

Half the cost.

The distance between the two floors was never legal. The laws are the same for everyone.

The distance is one of representation: on the upper floor, there is always someone sitting on the debtor's side, designing. At the branch, there is someone sitting on the product's side, selling.

As long as that remains true, the entrepreneur will keep paying shelf price for a risk he never had.

The credit you take says less about your company than about who designed the debt.

And so far, nobody has designed yours.

Leo Bentier

Switzerland or Paraguay: when crypto discovers that wealth has an address

Financial freedom without a fiscal exit is not freedom; it is a pretty balance waiting for the State to wake up — and the State has just woken up.

Brazil's Federal Revenue Service published a number this month that almost nobody read properly. Between August 2019 and December 2025, Brazilians declared R$ 1.58 trillion in crypto-asset transactions. Of that total, R$ 1.13 trillion was stablecoins. The news read adoption. I read dollarization. Eighty percent of Brazilian crypto is not a bet on technology. It is currency flight with a technical layover.

And every successful escape attracts cartographers.

Since the first of July, crypto-asset transactions must be reported to the Revenue Service through DeCripto, the new declaration created by Normative Instruction 2,291, which retires the 2019 system. The name sounds like bureaucracy. Regime changes usually sound like bureaucracy. DeCripto creates no tax at all, and that is exactly why it deserves attention. It creates something worth more than a tax: the map.

States do not tax what they cannot see. First comes the census. Then comes the collector.

DeCripto was designed to the standard of the OECD's Crypto-Asset Reporting Framework, the agreement that makes crypto-asset information circulate automatically between tax authorities worldwide. Standardized layout, file generated by the declarant himself, everything ready for cross-referencing. The man who treats his cold wallet as a bunker forgets that he himself signs the record at the endpoints: when he buys, when he sells, when he converts, when he spends. The blockchain does not forget, and now neither does the taxman.

The asset may be decentralized. The owner is not.

The government, by the way, has already shown the price it intends to charge. In 2025 it tried to tax all crypto gains at a flat 17.5%, with no exemption, by provisional decree. Congress let the measure lapse and the market celebrated it as a victory. I read it as a postponed budget. A State that breaks revenue records and still runs a deficit does not give up on a tax base; it waits for the base to become visible. The 2025 attempt failed for lack of a map, not for lack of appetite. Now the map exists.

The exemption for sales of up to R$ 35 thousand per month survives. For now. An exemption on visible wealth is not a right; it is a courtesy. And fiscal courtesy is the kind of thing that gets revoked by provisional decree on a Tuesday night. The Central Bank, meanwhile, is already discussing a financial-transactions tax on stablecoins. The siege is not one law. It is a sequence.

That is why I say: the crypto market will take a lot of people to one of two places by next year. Switzerland or Paraguay.

Switzerland for those who made real money and understood that wealth needs jurisdiction. Paraguay for those who discovered too late that profit on a screen does not pay for a tax lawyer.

I am not talking about countries. I am talking about methods.

Switzerland is a method. It is the man who accepted that large wealth does not live anywhere by accident, paid the lawyer before needing him, declared everything, structured his exit in broad daylight, paid the farewell toll where there was a toll, and sleeps today. The method is expensive, boring, and early. The three characteristics of almost everything that works.

Paraguay is a symptom. It is the man who crosses the border after the taxable event, believing geography erases the past. It does not. The tax does not chase the body; it chases the record. And the record, as of this month, travels in a standardized format, straight to the taxman's computer — and, through CARF, to the computers of dozens of other taxmen. Changing your address after the taxable event is not planning. It is a confession with a new zip code.

Brazil is teaching an entire generation, in the most expensive way possible, that financial freedom without a fiscal exit is just a pretty balance waiting for the State to wake up. And the State woke up the way competent States wake up: in silence. First the reporting obligation. Then the data cross-referencing. Then the assessment. The new tax rate comes last, when resisting is already too late.

What would I do with this? First, what I would not do.

I would not sell crypto because of DeCripto. The thesis of this letter is not against the asset; it is against the illusion that the asset has no address. Whoever bought bitcoin as insurance against local monetary incompetence still holds a valid policy. What expired was the anonymity, not the insurance.

I also would not romanticize self-custody as a fiscal shield. Self-custody protects against the exchange going under, not against data cross-referencing. The problem was never storing. It was entering and exiting. And nobody lives on the blockchain; people live on rent, groceries, and school, all charged in state money.

I would treat jurisdiction as an asset class. It behaves like insurance: you pay before, it looks like waste in the meantime, and it is worth everything on the day of panic. Whoever holds meaningful wealth in crypto and has never spent an hour with a tax lawyer is running leveraged on a risk that does not show up on the screen. The price of bitcoin updates every second. The price of fiscal disorganization updates once, with interest, penalties, and your name on the defendant's side.

As an investor, I would look at whoever sells the inevitable. Heavy regulation is a subsidy for those who are already big. DeCripto imposes a compliance cost the informal exchange cannot pay and the institutional platform has already paid. The volume that runs through gray channels today will not disappear; it will migrate to whoever holds a license. I would be long whoever sells compliance and suspicious of any business whose margin depends on tolerated invisibility. Tolerated invisibility is the most perishable raw material in finance.

The risk of the thesis is familiar. The taxman may trip over his own complexity. Congress may protect the exemption for years. CARF may take time to produce a meaningful assessment. States botch execution as often as they nail intention. But notice that none of these risks restores the invisibility. They only postpone its price.

The old question was: how much is your portfolio worth? The new question is: in which jurisdiction does it exist — and who else knows?

Whoever answers the second question before the State chooses between Switzerland and Paraguay calmly, in broad daylight, within the law. Whoever answers it afterwards does not choose. He is chosen.

I would not bet against Brazilian crypto.

I would bet against the idea that it can keep pretending it has no address.

Leo Bentier

Monopoly: an entire childhood looking at the wrong side of the board

The winner is not whoever owns the most houses; it is whoever finances everyone else's houses — and the Brazilian bank charges those who have wealth the price of those who have nothing.

You spent your childhood playing Monopoly. You bought avenues, built little green houses, collected rent from your younger brother, and thought you were learning how to get rich. Nobody told you two things. First: the game exists in real life. Second: the winner is not whoever owns the most houses. It is whoever finances everyone else's houses.

Look at the bank.

On the board, everyone fights over the avenues and nobody looks at it. The bank never lands on the wrong square. It never pays rent. It never goes to jail. It sells the deeds, finances the mortgages, collects the fines, and gets paid on both sides of every turn. The players bankrupt each other; the bank outlives them all. Not by luck. By position.

The game was invented in 1904 by Lizzie Magie, an American woman who wanted to denounce exactly this mechanism: the owner of capital defeats the owner of effort, and the extracted rent impoverishes the whole table. The denunciation became a toy. The mechanism became childhood. And the main lesson crossed a hundred years without being read: all the tokens lose. The only permanent chair is the bank's.

Now step off the board and look at Brazil this month.

The Central Bank shows corporate credit rates at their highest level since 2017 — and the average is still misleading, because it mixes worlds. With the base rate at 15%, in unsecured lines the branch offers working capital to the small and mid-sized entrepreneur at three, four percent per month. So far, nothing new: Brazil has always been expensive. The news is in the detail almost nobody prices. The bank charges this from everyone. From those who have nothing — and from those who have.

This is the villain of this letter, and he deserves a precise description. The traditional bank accepts your property as collateral, registers the fiduciary lien, immobilizes your wealth at the registry office — and hands you the same working capital, at the same price, it hands to someone who owns no property at all. It charges those who have wealth the price of those who have nothing. The collateral goes into the vault. The discount never comes out of it.

Why? Because the branch bank does not price your collateral. It prices its own process. The manager has no authority, the model was calibrated for the average, the committee sits three floors and two weeks away, and the shelf carries only three products. In a country where five counters concentrate nearly four fifths of all credit, the interest rate does not measure the client's risk. It measures his lack of alternatives.

And the entrepreneur accepts. He accepts because payroll is due Friday, the supplier on Monday, and he is in a hurry all month long. Hurry is the raw material of the spread.

Abecip has just celebrated a record: home equity lending reached R$ 31.66 billion in the first quarter, the largest volume in its historical series, growing 25% in one year. The press read triumph. I read the number backwards. Brazilians' real estate is worth, by the most conservative accounts, more than ten trillion reais. The all-time record of credit secured by it is thirty-one billion. Three tenths of one percent. The largest idle collateral on the planet sleeps in a Brazilian registry office.

In the United States, using the house as collateral is a middle-class routine. In Brazil, the entrepreneur owns a warehouse, his own headquarters, a farm, receivables — and an overdraft line at four percent per month. He is not a bad risk. He is a mispriced client.

The most ironic part is that the legal plumbing already exists. The 2023 Collateral Framework Law allows the same property to secure more than one operation and unlocked extrajudicial foreclosure. The law is ready. The incumbent is not, and for a reason that is not technical. Whoever profits from working capital at four percent per month is in no hurry to sell the same money at one and a half. No incumbent rushes to cannibalize his own margin. The board-game bank never offered a discount either.

I know both chairs at this table. I have spent my life looking at companies from the side of capital: who has it, who needs it, and how much the bridge between the two costs. And the credit I know begins with a question the branch never asks. The branch asks how much you invoice. The right question is: what do you own? The difference between those two questions is worth two points per month. And two points per month, compounded, is the difference between a company that grows and a company that works for the bank.

The entrepreneur usually discovers this late. He spent ten years building the asset with decade-long patience and accepted the liability with end-of-month haste. He assembled wealth like a saver and financed it like a desperate man.

If I had wealth and needed working capital, I would never again accept the price of someone who has none. Real collateral is bargaining power. Whoever does not use it, donates it. Donates to the bank, every month, the difference between the price of the process and the price of the risk.

And as an investor, I would sit on the bank's side of the board. The most asymmetric business in Brazilian private credit is financing those who own real assets, against the real asset, with a margin of safety: low loan-to-value, fiduciary lien, honest tenor. Lending at one and a half percent per month against a property worth twice the debt is not a bet. It is renting out capital with a deed attached. The risk for whoever does this properly is not default; it is the temptation to grow too fast and forget the margin.

The remaining risks are familiar. Foreclosure in Brazil tests any creditor's patience. Optimistic appraisal turns collateral into fiction. An illiquid property is a painting on the wall: worth a lot until the day you sell. And the base rate may fall and compress the premium. But notice that none of these risks resurrects the branch model. They merely select who will replace it.

The childhood board had a manufacturing defect: the bank was not for sale. You chose between the dog token and the top hat, and the chair that never loses kept the rules. In real life, the defect does not exist. The bank's chair sits vacant everywhere the incumbent charges process prices for wealth-backed risk.

The old question was: how many houses do you own on the board? The new question is: which side of the financing are they on?

I would not bet against Brazil's big banks. They are the most resilient profit machines in the country; they survived inflation, confiscation, economic plans, and every president.

I would bet against the idea that only they can be the bank.

Leo Bentier

The Tariff Is Not the Crisis. It Is the Invoice.

Brazil was not surprised by Washington. It was billed — with two decades of interest and penalties.

This Tuesday in Washington, on the second day of public hearings over the 25% tariff on Brazilian goods, Brazil was defended — but not by Brazil. By Tesla. By Coca-Cola, Nestlé, eBay — American companies explaining to their own government that Florida's oranges are gone, that supply chains do not survive the measure, that tariffing Brazil is too expensive for the United States.

Keep that image; it is worth more than any country-risk report: at the table where the commercial future of a nation of two hundred million people is being discussed, the Brazilian interest is represented by the procurement departments of foreign corporations. Our best defense is other people's convenience. Not for lack of diplomats — Itamaraty remains one of the most sophisticated bureaucracies on earth. For lack, over twenty years, of the raw material diplomacy works with: a decision.

A former trade secretary summarized the episode in the dialect of Brasília: Brazil "fell asleep at the bus stop" by not negotiating earlier. Good line. Generous diagnosis. He who sleeps at the stop missed a bus — a lapse, bad luck, a nap. Brazil did not miss a bus. Brazil spent two decades commissioning working groups to discuss whether there should be buses, hiring consultants to estimate the bus's impact on sensitive sectors, and postponing the tender for the stop. This was not sleep. It was method.

Do the archaeology of postponement and you find the strata. The European Union agreement, negotiated for over twenty years, shelved and unshelved with the political seasons. Trade liberalization, forever scheduled for after the next election — and the next election, like the horizon, forever walking alongside us. The founding question — what does Brazil want to sell the world besides ore and protein? — never answered, because answering it would require displeasing someone today in exchange for a better country later, and that is precisely the one transaction the Brazilian political system does not execute. A postponed decision is not neutral; it is a decision made by whoever moves first.

And someone always moves first.

In the same edition of the newspapers reporting the tariff, a second story, laid out as if it belonged to another subject: Brazil has become the world's top destination for Chinese investment. Ports, energy, minerals, digital infrastructure. It is not another subject. It is the same subject, seen from the other side of the counter. While Washington sets the entry price of our products, Beijing sets the purchase price of our assets. One charges at the exit; the other buys the gate. The two capitals are doing to Brazil exactly what Brazil refuses to do to itself: choosing.

Do not read this as moral complaint. The United States and China act as they act because they can, and they can because they decided to be able to — each built, in its own way, the capacity to turn commerce into an instrument. Moral complaint is the luxury of those without a strategy. What interests me is the structure: the map of trade is redrawn by whoever controls the route, not by whoever makes the product. We make the product. Rather well, in fact — the soybean is guilty of nothing. But the route — the tariff, the port, the submarine cable, the agreement, the settlement currency — is being drawn in two languages, and neither of them is Portuguese.

There was a time when this outsourcing looked cheap. The globalization of 1990–2018 was one long holiday from geopolitics: produce at low cost and the system — WTO rules, the American navy, Chinese patience — would handle the rest. The holiday is over. Globalization did not end; it acquired a political price that no spreadsheet had priced in. The countries that understood this early — the ones that signed agreements while there were agreements to sign, that diversified customers while diversification was still a choice and not a rescue — turned commerce into foreign policy and foreign policy into a balance sheet. The ones that did not, send senators to Washington to watch, from the gallery, hearings about their own fate.

The hurried reader will want to know which side I am on: negotiate with Trump or embrace Beijing? Wrong question, and wrong questions are expensive. Choosing between dependencies is not strategy; it is an auction. The right question comes earlier: why does a continental country — with the cleanest energy matrix among the large economies, self-sufficient in food, sitting on the reserves the world's energy transition requires — arrive at the table holding no cards, while owning every card in the deck? The answer is not in Washington, nor in Beijing. It is in the incentive structure of Brasília, where postponing is always individually rational: he who decides risks his job; he who requests one more study keeps it. Multiply this little theorem by forty thousand government offices and you obtain Brazilian trade policy — or the absence that answers to that name.

I do not write to predict the outcome of the tariff. It may be suspended next week by the stroke of a pen, and nothing that matters will have changed — the event is noise; the structure is the signal. I write to register, with a date on it, what this episode reveals: a country that does not decide its own place in the world does not remain neutral. It becomes available. Available for one power's tariff, for the other's capital, for the next hearing in which our fortune is debated in a foreign language, defended by lawyers we never hired.

The final question, as always, is not "what will Washington do to us." It is: who, in Brazil, wins by keeping the decision postponed? Find that answer and you will understand why it remains postponed. And when the next invoice arrives — and it will, with a different sender and the same addressee — do not call it a crisis. Crisis is the name the debtor gives to the collection call.

The new world order will be tariffed, routed, and processed through payment systems

Modern sovereignty lives in the chip, the port, the data, the court, and the payment button.

On July 7, 2026, the event looked like news. A serious letter would have treated it as a signal. News tells you what happened; a signal reveals what may repeat later with another name, another face, and another flag. Politics is a machine for recycling old passions with new vocabulary. Whoever reads only the event becomes hostage to the calendar. Whoever reads the incentive begins to see the future while it is still disguised as exception.

Geopolitics is no longer only tanks and treaties; it is API, card, QR code, sanctions, cloud, semiconductors, tariffs, and environmental reputation.

The argument would not be born from party preference, but from an older question: what kind of regime, elite, and people does this event reveal? Small politics asks who won today. Good politics asks which vice was rewarded, which virtue was punished, and which cost was pushed onto the next generation. This is where almost everyone fails. The commentator looks for convenient culprits; the statesman looks for mechanisms. The cheap moralist points the finger; the serious moralist asks why that finger found so many accomplices.

Brazil and the modern world suffer from the same disease with different accents: the confusion between form and substance. We have elections, but not always trust. Courts, but not always perceived justice. Markets, but not always responsibility. State, but not always common good. Communication, but not always truth. Data, but not always wisdom. Modernity multiplied instruments and impoverished judgment. The result is a civilization technically equipped to err at speed.

What could be seen on this date was less an isolated fact than a movement of plates. The surface would still allow calm speeches. It always does. Before ruptures, ministers guarantee normality, analysts calculate small probabilities, newspapers choose moderate adjectives, and investors pretend political risk fits in a spreadsheet. But history does not move only when there is an explosion. It moves when important men begin to lie differently. A change in the lie often precedes a change in the regime.

The letter would avoid the vice of the cheap prophet. The cheap prophet wants to look like owner of the future. The serious observer knows only that certain arrangements are fragile, certain incentives rot institutions, and certain accumulated humiliations look for a date. It is not necessary to predict the exact lightning strike when one knows the forest is dry. The vanity of guessing the day is inferior to the discipline of recognizing flammable wood.

There would also be a warning to the Brazilian reader. Brazil likes importing foreign explanations without importing the structures that make them true. It speaks of liberalism without impersonality, strong state without efficiency, sovereignty without strategy, democracy without civilized losers, market without competition, conservatism without self-control, progressivism without fiscal responsibility. That is why the country spends decades debating the vocabulary of great politics while preserving the habits of the small court.

Cicero would have seen in this episode a question about the republic: does law serve the city or the factions? Augustine would ask whether power, separated from justice, is truly government or only a gang with protocol. Aquinas would remind us that the common good is not the sum of private appetites, but a moral order that allows the city to seek higher ends. Marcus Aurelius would distrust applause, panic, and indignation, for all are poor emotional states for government. Churchill would look for will in the event, not excuse. Plato would ask what kind of collective soul chose this moment.

The forecast, then, would be simple and hard. The event recorded here would not end in itself. It would create habit, precedent, resentment, architecture, or language. The first error of observers would be to treat it as a period. The second would be to judge it only by the immediate winner. The third would be to believe the institution touched by the event would emerge unchanged. Institutions are not buildings; they are memories of behavior. When behavior changes, the building may remain standing while the republic has already changed skin.

Brazil will keep discussing twentieth-century ideology while its strategic infrastructure becomes external bargaining currency if it fails to understand this.

In retrospect, the essential point would be this: almost every major political event first appears as anomaly, then as repetition, then as doctrine. The mistake is laughing at the anomaly. The opportunity is asking why it appeared now, why it found an audience, and who gains from its normalization.

There is a difference between good politics and cheering. Cheering needs innocents and monsters. Good politics needs causes, limits, and consequences. Cheering wants to absolve its own and condemn the others. Good politics asks which sin is common to both sides. Cheering calls prudence cowardice when it is winning and persecution when it is losing. Good politics knows that every faction, when it believes itself redemptive, begins rehearsing small tyrannies.

This is why the letter would not ask the reader for comfort. It would ask for maturity. The world is not governed by declared intentions, but by tolerated incentives. An elite can speak of democracy while despising the people. A people can speak of liberty while desiring revenge. A market can speak of efficiency while asking for public rescue. A state can speak of justice while expanding its own appetite. A power can speak of order while surrounding rivals. A rival can speak of security while preparing aggression.

The reader who wants to understand the future should abandon the superstition of proper names. Names matter, but less than mechanisms. The man passes; the incentive remains. The party changes; the method persists. The crisis ends; the learned fragility moves elsewhere. The modern world is a school of disguises. Whoever clings to the disguise always arrives late.

A good letter does not end by asking for agreement. It ends by leaving useful discomfort.

Modern sovereignty lives in the chip, the port, the data, the court, and the payment button.

Leo Bentier

Bolsonaro under house arrest is the ghost sitting in the room

An imprisoned leader can still govern the imagination of those who have not accepted political mourning.

On July 3, 2026, the event looked like news. A serious letter would have treated it as a signal. News tells you what happened; a signal reveals what may repeat later with another name, another face, and another flag. Politics is a machine for recycling old passions with new vocabulary. Whoever reads only the event becomes hostage to the calendar. Whoever reads the incentive begins to see the future while it is still disguised as exception.

The Brazilian right would have to choose between managing the chief's ghost and politically burying him in order to win.

The argument would not be born from party preference, but from an older question: what kind of regime, elite, and people does this event reveal? Small politics asks who won today. Good politics asks which vice was rewarded, which virtue was punished, and which cost was pushed onto the next generation. This is where almost everyone fails. The commentator looks for convenient culprits; the statesman looks for mechanisms. The cheap moralist points the finger; the serious moralist asks why that finger found so many accomplices.

Brazil and the modern world suffer from the same disease with different accents: the confusion between form and substance. We have elections, but not always trust. Courts, but not always perceived justice. Markets, but not always responsibility. State, but not always common good. Communication, but not always truth. Data, but not always wisdom. Modernity multiplied instruments and impoverished judgment. The result is a civilization technically equipped to err at speed.

What could be seen on this date was less an isolated fact than a movement of plates. The surface would still allow calm speeches. It always does. Before ruptures, ministers guarantee normality, analysts calculate small probabilities, newspapers choose moderate adjectives, and investors pretend political risk fits in a spreadsheet. But history does not move only when there is an explosion. It moves when important men begin to lie differently. A change in the lie often precedes a change in the regime.

The letter would avoid the vice of the cheap prophet. The cheap prophet wants to look like owner of the future. The serious observer knows only that certain arrangements are fragile, certain incentives rot institutions, and certain accumulated humiliations look for a date. It is not necessary to predict the exact lightning strike when one knows the forest is dry. The vanity of guessing the day is inferior to the discipline of recognizing flammable wood.

There would also be a warning to the Brazilian reader. Brazil likes importing foreign explanations without importing the structures that make them true. It speaks of liberalism without impersonality, strong state without efficiency, sovereignty without strategy, democracy without civilized losers, market without competition, conservatism without self-control, progressivism without fiscal responsibility. That is why the country spends decades debating the vocabulary of great politics while preserving the habits of the small court.

Cicero would have seen in this episode a question about the republic: does law serve the city or the factions? Augustine would ask whether power, separated from justice, is truly government or only a gang with protocol. Aquinas would remind us that the common good is not the sum of private appetites, but a moral order that allows the city to seek higher ends. Marcus Aurelius would distrust applause, panic, and indignation, for all are poor emotional states for government. Churchill would look for will in the event, not excuse. Plato would ask what kind of collective soul chose this moment.

The forecast, then, would be simple and hard. The event recorded here would not end in itself. It would create habit, precedent, resentment, architecture, or language. The first error of observers would be to treat it as a period. The second would be to judge it only by the immediate winner. The third would be to believe the institution touched by the event would emerge unchanged. Institutions are not buildings; they are memories of behavior. When behavior changes, the building may remain standing while the republic has already changed skin.

The tension among Flavio, Michelle, governors, Congress, and the militant base shows that political inheritance can also become contested probate.

In retrospect, the essential point would be this: almost every major political event first appears as anomaly, then as repetition, then as doctrine. The mistake is laughing at the anomaly. The opportunity is asking why it appeared now, why it found an audience, and who gains from its normalization.

There is a difference between good politics and cheering. Cheering needs innocents and monsters. Good politics needs causes, limits, and consequences. Cheering wants to absolve its own and condemn the others. Good politics asks which sin is common to both sides. Cheering calls prudence cowardice when it is winning and persecution when it is losing. Good politics knows that every faction, when it believes itself redemptive, begins rehearsing small tyrannies.

This is why the letter would not ask the reader for comfort. It would ask for maturity. The world is not governed by declared intentions, but by tolerated incentives. An elite can speak of democracy while despising the people. A people can speak of liberty while desiring revenge. A market can speak of efficiency while asking for public rescue. A state can speak of justice while expanding its own appetite. A power can speak of order while surrounding rivals. A rival can speak of security while preparing aggression.

The reader who wants to understand the future should abandon the superstition of proper names. Names matter, but less than mechanisms. The man passes; the incentive remains. The party changes; the method persists. The crisis ends; the learned fragility moves elsewhere. The modern world is a school of disguises. Whoever clings to the disguise always arrives late.

A good letter does not end by asking for agreement. It ends by leaving useful discomfort.

An imprisoned leader can still govern the imagination of those who have not accepted political mourning.

Leo Bentier

Brazil entered the election before the election

The campaign begins when factions decide no fact will be neutral until October.

On July 1, 2026, the event looked like news. A serious letter would have treated it as a signal. News tells you what happened; a signal reveals what may repeat later with another name, another face, and another flag. Politics is a machine for recycling old passions with new vocabulary. Whoever reads only the event becomes hostage to the calendar. Whoever reads the incentive begins to see the future while it is still disguised as exception.

The 2026 campaign would be less a contest between candidates than a plebiscite over sovereignty, memory, and fear.

The argument would not be born from party preference, but from an older question: what kind of regime, elite, and people does this event reveal? Small politics asks who won today. Good politics asks which vice was rewarded, which virtue was punished, and which cost was pushed onto the next generation. This is where almost everyone fails. The commentator looks for convenient culprits; the statesman looks for mechanisms. The cheap moralist points the finger; the serious moralist asks why that finger found so many accomplices.

Brazil and the modern world suffer from the same disease with different accents: the confusion between form and substance. We have elections, but not always trust. Courts, but not always perceived justice. Markets, but not always responsibility. State, but not always common good. Communication, but not always truth. Data, but not always wisdom. Modernity multiplied instruments and impoverished judgment. The result is a civilization technically equipped to err at speed.

What could be seen on this date was less an isolated fact than a movement of plates. The surface would still allow calm speeches. It always does. Before ruptures, ministers guarantee normality, analysts calculate small probabilities, newspapers choose moderate adjectives, and investors pretend political risk fits in a spreadsheet. But history does not move only when there is an explosion. It moves when important men begin to lie differently. A change in the lie often precedes a change in the regime.

The letter would avoid the vice of the cheap prophet. The cheap prophet wants to look like owner of the future. The serious observer knows only that certain arrangements are fragile, certain incentives rot institutions, and certain accumulated humiliations look for a date. It is not necessary to predict the exact lightning strike when one knows the forest is dry. The vanity of guessing the day is inferior to the discipline of recognizing flammable wood.

There would also be a warning to the Brazilian reader. Brazil likes importing foreign explanations without importing the structures that make them true. It speaks of liberalism without impersonality, strong state without efficiency, sovereignty without strategy, democracy without civilized losers, market without competition, conservatism without self-control, progressivism without fiscal responsibility. That is why the country spends decades debating the vocabulary of great politics while preserving the habits of the small court.

Cicero would have seen in this episode a question about the republic: does law serve the city or the factions? Augustine would ask whether power, separated from justice, is truly government or only a gang with protocol. Aquinas would remind us that the common good is not the sum of private appetites, but a moral order that allows the city to seek higher ends. Marcus Aurelius would distrust applause, panic, and indignation, for all are poor emotional states for government. Churchill would look for will in the event, not excuse. Plato would ask what kind of collective soul chose this moment.

The forecast, then, would be simple and hard. The event recorded here would not end in itself. It would create habit, precedent, resentment, architecture, or language. The first error of observers would be to treat it as a period. The second would be to judge it only by the immediate winner. The third would be to believe the institution touched by the event would emerge unchanged. Institutions are not buildings; they are memories of behavior. When behavior changes, the building may remain standing while the republic has already changed skin.

By July, tariffs, Pix, Washington, the Supreme Court, Bolsonaro imprisoned, and family succession already indicated that the election would be fought inside and outside Brazil.

In retrospect, the essential point would be this: almost every major political event first appears as anomaly, then as repetition, then as doctrine. The mistake is laughing at the anomaly. The opportunity is asking why it appeared now, why it found an audience, and who gains from its normalization.

There is a difference between good politics and cheering. Cheering needs innocents and monsters. Good politics needs causes, limits, and consequences. Cheering wants to absolve its own and condemn the others. Good politics asks which sin is common to both sides. Cheering calls prudence cowardice when it is winning and persecution when it is losing. Good politics knows that every faction, when it believes itself redemptive, begins rehearsing small tyrannies.

This is why the letter would not ask the reader for comfort. It would ask for maturity. The world is not governed by declared intentions, but by tolerated incentives. An elite can speak of democracy while despising the people. A people can speak of liberty while desiring revenge. A market can speak of efficiency while asking for public rescue. A state can speak of justice while expanding its own appetite. A power can speak of order while surrounding rivals. A rival can speak of security while preparing aggression.

The reader who wants to understand the future should abandon the superstition of proper names. Names matter, but less than mechanisms. The man passes; the incentive remains. The party changes; the method persists. The crisis ends; the learned fragility moves elsewhere. The modern world is a school of disguises. Whoever clings to the disguise always arrives late.

A good letter does not end by asking for agreement. It ends by leaving useful discomfort.

The campaign begins when factions decide no fact will be neutral until October.

Leo Bentier

Open USD: when the tollbooth discovers it can also issue the road

The stablecoin is not destroying the traditional financial system; it is teaching the traditional financial system to steal back the float it let escape.

Visa and Mastercard spent decades pretending to be at total war over a few basis points on every transaction on the planet. The ordinary investor saw two rival networks. I saw two private customs houses. They do not produce the product. They do not carry the merchandise. They do not necessarily lend the money. They simply stand in the right path, at the right moment, charging a fraction small enough to look inevitable and large enough to become a compounding machine.

Now they appear on the same side of the table.

Alongside BlackRock, Google, Coinbase, Stripe, banks, fintechs, processors, wallets, networks, and more than a hundred participants, they sign the birth certificate of Open USD. The name is bad. That does not matter. Good names often hide bad businesses. Generic names sometimes hide regime changes.

The superficial thesis is crypto. The real thesis is float.

A stablecoin is not a new currency. It is a digital receipt on top of an old currency. The customer delivers real dollars. He receives a token that promises to be worth one dollar. The issuer keeps the real dollars and buys liquid assets, usually short-term U.S. Treasuries. The user receives stability. The issuer receives interest. When rates are high, the stablecoin becomes a profit factory. When the team is small, the factory looks obscene.

Tether showed the size of the obscenity.

The public calls this innovation. I would call it a bank with no branch, no smiling manager, no carpet, no bank advertising, and no interest paid to the depositor. The beauty is in the absence. Absence of physical cost. Absence of legacy network. Absence of compensation to the end user. Absence of shame.

But every profitable absence attracts predators.

The mistake of the crypto-natives was believing the traditional financial system would look at a pile of profit floating in T-bills and say: "congratulations, keep it." Banks, payment networks, and asset managers are not romantics. They do not hate crypto on principle. They hate it only when the profit sits on the wrong side of the table.

Open USD is the most logical institutional answer: if stablecoin profit comes from float, and if float depends on distribution, liquidity, trust, integration, compliance, merchants, banks, wallets, processors, and channels, why should the issuer keep everything?

That is the question threatening the entire model of pure issuers.

The proposal is simple. Partners can issue and redeem OUSD without meaningful cost, without volume limits, and receive almost all the yield on reserves after a small management fee. In other words, Open USD tries to turn the stablecoin from a proprietary product into shared infrastructure. The money stops remunerating only the owner of the printer. It starts remunerating whoever brings circulation.

It sounds fair. The market does not pay for fairness. It pays for power.

And the power here is the channel. Tether has liquidity and habit. Circle has compliance and institutional relationships. But Visa, Mastercard, Stripe, Coinbase, banks, and large platforms have distribution. The scarcest asset in finance was never technology. It was trust packaged into repeated behavior. The user does not want to study reserves, attestations, smart contracts, custody, solvency, regulatory risk, and secondary liquidity. The user wants it to work.

Whoever controls "it works" controls the spread.

The most interesting part is not Visa entering stablecoins. It is Visa admitting that stablecoins may be inevitable enough to deserve a seat at the table. The intelligent incumbent does not try to stop the tide. He buys a dock, regulates the port, and begins charging for disembarkation.

Mastercard knows the same thing.

The card networks are not committing suicide. They are trying to avoid letting someone else choose where they will be stabbed. Stablecoins threaten part of the settlement, cross-border, treasury movement, and B2B payments model. But they also create new surfaces of monetization: identity, compliance, merchant acceptance, fraud prevention, token orchestration, gateways, alternative chargeback, bank integration, programmable settlement, and business accounts.

The public thinks payment is moving money. Payment is moving trust.

Money without trust is metal, paper, or data. Trust without distribution is a monastery. Distribution without economics is charity. Open USD tries to combine the three: dollar, network, and incentive.

But there is poison inside the elegance itself.

If almost all the reserve yield goes back to partners, where is the operator's structural margin? If all the large participants govern, who decides when the decision needs to be unpopular? If the board is broad, who acts when there is a redemption run? If the stablecoin is open, who absorbs the damage when something fails? Shared governance sounds noble during conferences. During panic, shared governance usually becomes a line of lawyers.

A stablecoin is truly tested only in three moments: when interest rates fall, when the regulator changes mood, and when the market demands redemption while everyone else pretends to be calm.

We have not yet seen Open USD in those three moments. We have seen an announcement. Announcements do not break. Products break.

That is why I would not buy the euphoria as the primary thesis.

I also would not buy the token. The token is not equity. A tokenized dollar is an operational promise, not a share of the profit. The investor who buys a stablecoin thinking he is buying the company is confusing the receipt with the factory.

The trade, if it existed at the right price, would be in stocks and derivatives, not in the token.

I would look first at the pure stablecoin issuers that were, or will be, valued as if the yield on reserves were permanent property. Nothing in finance that depends on a fat spread remains without competition. High profit is an invitation. Obscene margin is a subpoena.

If Circle is priced as the owner of an exclusive bridge, I would ask: exclusive for how long? If Coinbase depends economically on revenue-sharing agreements with USDC, I would ask: how much of that revenue is partnership and how much is transition until a stablecoin more aligned with the distributor appears? If Tether were public, the question would be even more brutal: how much of the premium comes from opacity tolerated because the profits are too large to ignore?

Tether is not dead. That would be childish thinking. Tether has liquidity, global use, emerging markets, trading pairs, and a form of parallel trust that traditional institutions do not fully understand. In many places, Tether does not compete with an American bank. It competes with bad local currency, capital controls, and state incompetence. That is a different fortress.

But Tether's profit revealed the mine.

And revealed mines receive railroads, taxes, guards, and thieves.

My way to position would be small, asymmetric, and suspicious. I would not buy Visa or Mastercard merely because they are in the consortium. They are excellent companies, but size already charges an expensive entry ticket. I also would not go around shorting the card networks as if stablecoins will replace them tomorrow. That is the kind of mistake impatient men make when they confuse direction with duration.

I would seek a relative structure.

Long distribution infrastructure that benefits from adaptation, short or protected against monoline issuers whose valuation depends on concentrated float. In plain terms: selective long in rails that can incorporate stablecoins without losing their soul; long-dated puts or put spreads on public issuers the market has priced as if interest on reserves were an eternal private property. If the option is expensive, there is no trade. The thesis can be right and the operation still stupid.

That distinction saves capital.

I would rather wait for a relief rally in the names hit by the announcement. The market first panics, then rationalizes, then exaggerates in the opposite direction. The best short is rarely born on the first day of bad news. It is born when the company explains that there is "no material impact," analysts repeat the phrase, and the stock recovers enough to sell insurance to those who forgot where the fire started.

The risk of the thesis is clear. Open USD may fail from too much committee. It may fail to gain liquidity. It may become one more institutional token with no real use. It may be captured by banks. It may face regulatory restrictions. It may discover that splitting the economics with everyone creates an incentive for nobody to carry the piano. It may even prove that the incumbent understood the threat too late.

But the structural direction seems hard to deny.

The profit of float is being repriced. Before, the question was: who issues the stablecoin? Now, the question becomes: who deserves the yield of the stablecoin? The issuer? The distributor? The bank? The merchant? The wallet? The network? The user? The regulator?

When a question changes, the multiple changes before the balance sheet.

Most people will still debate whether "crypto won" or "banks won." That debate is for amateurs. What is happening is dirtier and more interesting. The traditional financial system is not surrendering. It is absorbing the mechanism that threatened it, removing part of the pioneers' margin, and redistributing the prize to whoever controls flow, license, and trust.

The stablecoin was sold as a rebellion against intermediaries.

Open USD is the revenge of the intermediaries.

And perhaps that is why it deserves attention.

Not because it will necessarily be the winner. But because it shows the game has changed. When Visa and Mastercard stop fighting over a transaction and join forces to redesign the economics of the tokenized dollar, the investor should not ask whether this is "crypto." He should ask who keeps the interest.

The old answer was: the issuer.

The new answer may be: whoever moves the money.

Markets do not punish those who are poetically wrong. They punish those who are economically wrong. If Open USD works, value will not be destroyed. It will be redistributed. And value redistribution is the polite name for multiple compression in whoever was receiving too much.

I would not bet against stablecoins.

I would bet against the idea that their profit will keep living at the same address.

Leo Bentier

How to Become a Real Investor

Without Falling for the Smooth Talk of Market Charlatans

There is a convenient lie in financial markets.

It says that investing is about discovering the next thing that will go up.

The next stock.

The next coin.

The next company.

The next thesis.

The next opportunity that, through some moral accident, will be revealed to you for free by a man in a well-pressed shirt, holding an expensive microphone, speaking in casino vocabulary.

This lie works because it offers the beginner an infantile promise: that there is a shortcut between ignorance and fortune.

There is not.

What exists is something else, less sexy, less sellable, and infinitely more powerful.

A way of thinking.

That is what separates the real investor from the consumer of financial narratives.

The real investor does not begin by asking what to buy.

That is the amateur’s question.

The question of the man who still believes the asset contains the answer.

The man looking for salvation in tickers, reports, charts, videos, calls, recommended portfolios, and sentences dressed up as depth.

The correct question is different.

Is the market pricing this risk correctly?

It sounds like a small distinction.

It is not.

It is the whole distinction.

One question tries to guess the future.

The other tries to measure the present.

The first produces anxiety.

The second produces discipline.

The first feeds charlatans.

The second makes them useless.

The market charlatan needs you to keep believing in prediction. He lives by turning uncertainty into spectacle. He takes the natural volatility of the world, puts a suit on it, calls it a thesis, and sells it as if he had access to a secret room where the future is decided.

But the future is not decided in secret rooms.

The future is merely the place where today’s mistakes become visible.

That is why the real investor does not buy assets.

He buys distortions.

More precisely, he looks for situations where the distance between perceived risk and real risk has become too large.

That distance has a name.

Asymmetry.

An asymmetry is not an opportunity.

Opportunities are abundant. Every broker has a list of them. Every influencer has ten per week. Every course salesman has a formula. The world is full of opportunities because the word has lost its cost.

Asymmetries are rare.

They appear when the crowd has believed the same story for too long.

When complexity repels the lazy.

When an asset is too small for the large players and too complicated for the small ones.

When liquidity disappears.

When bureaucracy creates an intellectual tollbooth.

When the market confuses noise with risk.

When everyone looks at price and almost no one looks at the mechanism.

The real investor learns to look in those places.

Not because he is smarter.

But because he accepts paying the price the charlatan hates: studying the mechanism before buying the narrative.

The market is not a machine for rewarding opinions.

It is a machine for punishing fragility.

This distinction should be taught before the first stock purchase, before the first Bitcoin, before the first real estate fund, before the first option, before the first “monthly contribution.”

But it is not.

The market prefers to teach vocabulary.

Duration.

Convexity.

Alpha.

Beta.

Hedge.

Drawdown.

Valuation.

Momentum.

Useful words, in some contexts.

But also dangerous, because they give the beginner the feeling of sophistication before giving him a decision system.

Nothing is more dangerous than a man who has learned the terminology before learning the risk.

He becomes easy prey.

Not because he is stupid.

Because he is vain.

Financial vanity has a particular shape. It makes a man believe he understands an asset because he can explain the beautiful story behind it.

The company will dominate the sector.

The government will print money.

Bitcoin will replace the financial system.

Interest rates will fall.

Inflation will come back.

Artificial intelligence will change everything.

Perhaps.

Perhaps not.

But the real investor’s question is not whether the story is beautiful.

The question is how much of that story is already in the price.

And what happens if the story is wrong.

Here begins the first concept anyone must learn to become a real investor: price is not truth. Price is momentary consensus.

Sometimes consensus is right.

Often it is roughly right.

But at certain moments, it goes mad with elegance.

And when consensus goes mad with elegance, the market often looks safest precisely when it is most dangerous.

The charlatan sells certainty.

The investor looks for a margin of error.

That is the second lesson.

You do not become a real investor when you find a wonderful thesis. You become a real investor when you learn how to survive a wrong thesis.

The amateur asks: how much can I make?

The investor asks: how does this kill me?

The amateur thinks about return.

The investor thinks about ruin.

The amateur diversifies out of fear.

The investor concentrates only when he understands the asymmetry.

The amateur copies portfolios.

The investor builds criteria.

The amateur outsources judgment.

The investor outsources tasks, but never responsibility.

This point is unpleasant, which is why it is true.

Most people do not want to invest.

They want an authority to think for them.

They want an analyst, an influencer, a bank, a newsletter, a community, a manager, a teacher, a guru, a sufficiently confident name to remove the weight of decision.

But investing is precisely the act of carrying that weight.

You may listen to opinions.

You may read reports.

You may speak with people more intelligent than you.

You may use tools.

You may study models.

But in the end there is a moral line no one should cross: if the money is yours, the decision is yours too.

The person who does not accept this does not want to be an investor.

He wants to be a passenger.

And passengers should not complain about the destination chosen by the driver.

The good news is that anyone can become a real investor.

Not everyone can become rich.

That is another salesman’s lie.

But anyone can stop being a financial idiot.

Anyone can learn to distrust easy promises.

Anyone can learn to separate asset from narrative.

Anyone can learn to ask where the fragility is.

Anyone can learn to measure risk before imagining return.

Anyone can build a decision system.

That system begins with a simple rule.

Before asking whether something will go up, ask why it is mispriced.

If you cannot answer, you do not have a thesis.

You have hope with a spreadsheet.

The second rule is harder.

Before asking how much you can make, ask what must happen for you to lose permanently.

A temporary loss is discomfort.

A permanent loss is mutilation.

The market forgives discomfort.

It does not forgive repeated mutilation.

The third rule: never confuse complexity with intelligence.

Some of the worst errors arrive dressed as sophistication. Structured products, fashionable funds, strategies with foreign names, hidden leverage, apparent protections, fragile correlations. Financial markets are very good at turning poison into premium packaging.

When an explanation needs too many layers in order to look good, distrust it.

The fourth rule: do not buy what you can only defend by repeating someone else’s sentence.

If the thesis cannot survive outside the guru’s mouth, it is not yours.

It is intellectual borrowing.

And intellectual borrowing charges interest.

The fifth rule: look for mechanisms, not headlines.

The headline says a company grew.

The mechanism shows whether it grew with cash, debt, subsidy, luck, monopoly, accounting fraud, or macroeconomic wind.

The headline says an asset fell.

The mechanism shows whether it fell because of real deterioration, temporary panic, forced liquidity, or a simple change in mood.

The headline says everyone is buying.

The mechanism asks who will be forced to sell.

That is where asymmetry lives.

The sixth rule: turn opinion into criteria.

Opinion without criteria is mental decoration.

“I like this company” means nothing.

“This asset is cheap” means nothing.

“This thesis makes sense” means nothing.

Cheap against what?

Makes sense under which assumptions?

What is the tolerable error?

What is the loss limit?

What event invalidates the thesis?

What signal tells you that you are wrong?

If you do not know what would make you change your mind, you do not have a thesis.

You have a religion.

The market is full of religions with tickers.

Some speak in technological language.

Others speak in macroeconomic language.

Others speak in libertarian language.

Others speak in institutional language.

Others speak in passive-income language.

All have their priests.

All have their faithful.

All have their tithes.

The real investor may visit narrative temples, but he does not kneel.

He observes.

He measures.

He compares perceived risk and real risk.

He asks who is taking risk, who is transferring risk, and who is being paid to pretend risk does not exist.

That is the great protection against charlatans.

It is not cynicism.

Cynicism is laziness with an intellectual pose.

The real protection is structure.

Charlatans prosper where structure is absent.

They win when the investor has no criteria of his own. When the investor wants to belong. When he confuses confidence with competence. When he treats recent gains as proof of skill. When he forgets that luck also wears a suit.

A real investor does not need to hate the market.

He only needs to lose his innocence.

He needs to understand that the market is an arena of incentives.

The person selling a report has incentives.

The person selling a course has incentives.

The person managing other people’s money has incentives.

The person appearing on television has incentives.

The person posting a thesis at the height of euphoria has incentives.

The person recommending without suffering the loss has incentives.

This does not make everyone dishonest.

It makes everyone interested.

And interested people must be read carefully.

The question is not, “Is this person lying?”

The better question is:

What does this person lose if I follow his advice and it goes wrong?

Most of the time, nothing.

You lose money.

He loses reach for a few days.

Then he changes the thesis.

Rearranges the vocabulary.

Says the scenario changed.

Publishes another conviction.

And the game begins again.

That is why skin in the game is not a moral detail.

It is an epistemological filter.

Those who do not bleed when they are wrong should speak more quietly.

And you should listen with less reverence.

The real investor does not look for prophets.

He looks for asymmetries.

He does not look for certainty.

He looks for structure.

He does not look for enthusiasm.

He looks for a margin of safety.

He does not look for fashionable assets.

He looks for pricing errors.

He does not ask where the noise is.

He asks where the market is looking incorrectly.

All of this may sound abstract.

It is not.

It is the opposite.

It is too practical to be sold as fantasy.

You look at an asset.

You identify the dominant narrative.

You ask what is embedded in the price.

You look for the risk the market is exaggerating.

You look for the risk the market is ignoring.

You define what would invalidate your thesis.

You define how much you can lose without compromising your survival.

You define what would make you increase the position.

You define what would make you exit.

Then you act.

Measure.

Correct.

Without romance.

This is the difference between investing and consuming content about investing.

The real investor does not need to be right all the time.

He needs to be structured when he is wrong.

He does not need to predict the world.

He needs to avoid being destroyed by it.

He does not need to find every opportunity.

He needs to recognize a few real asymmetries when they appear.

The financial market has built an entire industry to convince you that investing is too complicated to be done without intermediaries and simple enough to be learned in a weekend course.

Both are false.

Investing is not simple.

But it is not mystical either.

It is a discipline of perception, risk, behavior, and decision.

You learn to see narratives.

You learn to measure fragility.

You learn to distrust elegant consensus.

You learn to look for asymmetries.

You learn to survive.

And, with time, perhaps you learn something even rarer.

You learn to do nothing.

Because the real investor does not need to be positioned all the time.

The person who must always be positioned is the salesman.

The salesman needs a topic.

The investor needs a price.

The salesman needs urgency.

The investor needs patience.

The salesman needs novelty.

The investor needs criteria.

The greatest financial freedom is not buying the right asset.

It is not being bought by the wrong narrative.

It is ceasing to be the emotional customer of those who monetize your anxiety.

It is looking at the market’s next shining promise and asking the question that ruins almost every beautiful speech:

Where is the asymmetry?

If there is no answer, there is no investment.

There is only smooth talk.

And smooth talk, in financial markets, is usually the most expensive asset a beginner buys.

Leo Bentier

I Don't Buy Companies.

Or, at least, that's not what I do.

There seems to be a small misunderstanding about me, and it may very well be my fault.

Over the past few weeks, some people have approached me trying to sell a company. Others wanted to buy one. Some wanted to discuss taxes. Others wanted to talk about artificial intelligence.

They all seem to believe I've developed an obsession with small Brazilian businesses.

I haven't.

The business was simply the first clue.

I spent a good part of the last few years building a tool to analyze businesses. The original idea was straightforward. Brazil is far too complex to be understood through financial statements alone. Beneath the surface lies an entire underground economy of poorly chosen tax structures, forgotten credits, ignored incentives, and decades of administrative improvisation.

The plan was to sell that intelligence.

I built the product.

I designed the commercial model.

Then something curious happened.

I realized I was looking in the wrong direction.

The tool was finding interesting companies.

But the companies were not the discovery.

The discovery was something else.

What it was actually finding were pricing errors.

A small manufacturer changing hands without competition.

A medical practice handicapped by mediocre accounting.

An owner convinced his business was worth very little.

A buyer convinced it was worth even less.

The companies were merely the place where the asymmetry revealed itself.

That was the moment I decided not to sell the tool.

Markets obey a simple rule.

The moment you start selling gold detectors, you increase the competition for gold.

There was no economic logic in manufacturing competitors.

Many people interpreted that decision as a bet on small businesses.

It wasn't.

It was a bet on asymmetries.

That word probably deserves a better definition.

An asymmetry is not an opportunity.

Opportunities are abundant.

Asymmetries are rare.

An asymmetry emerges when the gap between perceived risk and actual risk becomes too wide.

When everyone believes the same story.

When complexity drives buyers away.

When an asset is too small for large players and too large for small ones.

When liquidity disappears.

When bureaucracy creates an intellectual toll booth.

When most people simply decide to look somewhere else.

The small Brazilian business has these characteristics.

But it is hardly unique.

I began noticing the same phenomenon elsewhere.

Different markets.

Different people.

Different mechanisms.

The same underlying logic.

Investors have a curious habit.

They believe they make a living buying assets.

They buy real estate.

They buy stocks.

They buy businesses.

They buy Bitcoin.

They buy bonds.

I've never liked that description.

It feels incomplete.

Perhaps nobody buys assets.

Perhaps everyone buys narratives.

An expensive house is a narrative.

A billion-dollar startup is a narrative.

A forgotten family business is a narrative.

A currency is a narrative.

Government debt is a narrative.

The price almost never represents the asset itself.

It represents the story people tell about it.

That realization explained something that had always puzzled me.

I've always found it strange that investors spend so much time trying to figure out whether an asset will go up or down.

The question sounds important.

In practice, it rarely is.

The more interesting question is usually different.

Is the market pricing the risk of this asset correctly?

Those are very different questions.

One attempts to predict the future.

The other attempts to measure the present.

The first demands intelligence.

The second demands discipline.

I spent years believing I was building a tool to analyze companies.

Today, I think I was building a tool to analyze behavior.

Perhaps that explains a shift in perspective.

I have no particular interest in small businesses.

Or in real estate.

Or in commodities.

Or in Bitcoin.

Or in any specific asset class.

What interests me are places where collective perception and objective reality drift too far apart.

That gap interests me.

The small Brazilian manufacturer was merely the first laboratory.

There may be others.

There is an interesting consequence to all of this.

People often ask me what I'm investing in.

The question seems reasonable.

But it may be the wrong one.

The asset is usually the least interesting part of the story.

A good asymmetry can appear in a business.

It can appear in a financial structure.

It can appear in a contract.

It can appear in a piece of land.

It can appear in an entire market.

The wrapper changes.

The pattern remains.

Perhaps that was the real purpose of the tool I built.

It was never a system for finding companies.

It was never a tax platform.

It was never software.

It was simply a scanner.

A scanner for asymmetries.

And perhaps I realized too late that companies were merely one of the things it could see.

Leo Bentier

The Winner Will Be Whoever Turns AI Into Auditable Decision

AI's next frontier is not enchantment; it is operational trust, governance, and responsibility.

In June 2026, AI no longer needs to prove that it impresses. That has become cheap. Everyone has seen models write, summarize, program, draw, speak, research, be wrong with confidence, and be right with speed. The initial enchantment is over. The question now is less romantic: who can turn AI into auditable, safe, integrated, and economically defensible decision?

The winner will not simply be whoever "has AI". That phrase has already lost value. The winner will be whoever can place AI in processes where consequence exists: credit, collections, logistics, defense, energy, health care, legal, industry, insurance, procurement, audit, planning, compliance, maintenance, service, risk. In environments like these, an answer is not enough. One must trace origin, permission, explanation, decision, action, and responsibility.

Palantir, Nvidia, and AMD are obvious names. Palantir because it tries to occupy the operational layer of decision. Nvidia and AMD because they provide compute capacity for models, inference, and accelerated workloads. But Constellation Energy, Vertiv, Eaton, and Cadence reveal another dimension of the thesis: reliable AI depends on energy, electrical infrastructure, and silicon design. Auditable decision needs a reliable physical world underneath.

Constellation Energy enters because firm energy can become a strategic bottleneck. Data centers do not run on climate slogans or promises of availability. They need abundant, reliable electricity and, increasingly, electricity aligned with carbon commitments. If AI increases energy demand, firm generation assets and long-term contracts gain importance. Nuclear returns to the conversation not out of nostalgia, but because intermittency and compute density will have friction.

Vertiv and Eaton capture the infrastructure that allows density. Energy needs to be distributed, protected, converted, cooled, monitored. The more critical the application, the less tolerance for failure. AI in entertainment can go down and return. AI in critical operations cannot become an unstable toy. This increases the value of electrical and thermal infrastructure.

Cadence appears because the race for custom chips, efficiency, and advanced design will require design tools. If hyperscalers, startups, and suppliers seek specific silicon for inference, networking, memory, or acceleration, design complexity grows. EDA tools are an intellectual toll on the ambition to manufacture better machines.

Perhaps in 2027 the market begins punishing performative AI. Products that only answer beautifully can be compressed. Companies that can audit, govern, and integrate decisions can receive a premium. The difference will be clear: a demo answers; a system records. A demo enchants; a system justifies. A demo speaks; a system executes with control.

The reader's profit will be in seeking companies that sell operational trust. Palantir fits if it proves real use and expansion. Nvidia fits if it keeps supplying essential infrastructure for inference and training. AMD fits if it gains space through the economics of alternatives. Constellation fits if firm energy becomes a bottleneck. Vertiv and Eaton fit if data centers keep becoming denser. Cadence fits if custom silicon becomes a structural race.

The counter-thesis is that the market can exaggerate. Nuclear energy can face regulation, delays, and costs. Vertiv and Eaton can be repriced too much. Cadence can be excellent and expensive. Palantir can promise auditable decision and deliver service dependence. Nvidia can face capex digestion. AMD can fail to capture as much as expected. The thesis can be correct and the price wrong.

But the direction is hard: AI without trust does not enter the center of the company. It can stay at the edge, in marketing, in generic support, in individual productivity. To enter the heart, it needs to be governed. The next frontier is not only intelligence. It is responsibility.

The market was enchanted by machines that speak.

The next wealth will be in machines that can explain why they spoke, who authorized it, which data they used, what action they took, and who answers if it goes wrong.

This is not less revolutionary.

It is harder to copy.

Leo Bentier

Artificial Intelligence May Be Right. The Price Is Not.

I have no problem admitting that artificial intelligence may be right.

That is precisely the point the lazy investor fails to understand. He thinks every ounce of caution in front of an absurd rally is denial of the technology. It is not. The internet was right. Electricity was right. Railroads were right. Personal computing was right. The problem is almost never the invention. The problem is the price men pay when they mistake invention for salvation.

The market loves turning a truth into a delirium.

First comes a real technology. Then a real company. Then real growth. Then a narrative real enough to justify the first excess. Then come the late managers, the journalists, the analysts with elastic multiples, the chart influencers, and the crowd that does not buy businesses. It buys belonging. At that point, price stops being a measure. It becomes a religious confession.

I look at the market today and see less of a technological revolution than an old disease with a new vocabulary.

Before, it was “the internet.” Now it is “artificial intelligence.” Before, it was “the new economy.” Now it is “computational infrastructure.” Before, it was portals, fiber optics, and companies putting “.com” in their names. Now it is GPUs, data centers, semiconductors, language models, electric power, private credit, and mature companies being priced as if they had just discovered fire.

Perhaps AI will change the world. It probably will.

But the market should not ask only whether something will change the world. The market should ask how much of that is already in the price. And that is the question that disappears at the end of cycles. When the question of price disappears, the bubble no longer needs to introduce itself. It is already sitting at the table, drinking with the guests, being called a “secular thesis.”

I do not write this because I know the day of the collapse.

Those who ask for the exact date still do not understand the game. The amateur investor wants a calendar. The professional wants asymmetry. The amateur asks, “When does it fall?” The professional asks, “How much do I lose if I am wrong, how much do I preserve if I am right, and what will I be able to buy when others are forced to sell?”

The market does not send invitations before it breaks.

In 2000, the internet bubble did not need one clean, elegant catalyst acceptable for television. The market simply got tired of paying eternity for promises. In 1929, the story also refused to fit inside one explanation. In 2007, the signs were already there before consensus found the courage to call them a crisis. Most of the time, the catalyst is just the alibi that appears after the fragility has already been built.

The error is to look for the match and ignore the warehouse full of straw.

Today, I see that warehouse.

I see indexes at record highs while fewer companies carry the rally. I see concentration. I see technology stocks pulling the entire market imagination with them. I see semiconductors being treated as if every chip were a sacred coin. I see old, heavy, cyclical, capital-intensive companies rising like newly discovered startups. I see investors repeating “this time is different” with the same statistical innocence of every cycle in which, in the end, it was not.

The phrase “this time is different” almost always means: “I need history not to collect from me.”

But history collects.

It does not collect when it seems fair. It collects when excess has domesticated prudence. It collects when the cautious man has been mocked enough. It collects when cash looks stupid, when hedging looks cowardly, when valuation looks like an old man’s obsession, and when every chart tells the same upward story. The market does not punish euphoria at the beginning. First, it rewards it, so it can increase the number of victims later.

That is the moral machinery of bubbles: they enrich the reckless for long enough to persuade the prudent to abandon prudence.

I am not saying AI is a fraud.

That would be the easy argument, and easy arguments usually attract easy minds. AI may become one of the largest economic transformations of our generation. It may compress costs, replace processes, expand productivity, alter software, customer service, analysis, programming, diagnostics, research, defense, education, and almost everything that depends on language, inference, and decision. None of this obliges me to pay any price for any company that has managed to get close to the narrative.

The company can be excellent and the stock can be expensive.

The technology can be inevitable and the shareholder return can be mediocre.

The thesis can be right and the buyer can still be condemned.

That is the distinction separating investment from cheering.

The vulgar investor wants to be right about the world. The serious investor wants to make money after adjusting for price, risk, time, and survival. Being right about a technology is not enough. Many were right about the internet in 1999. Many were right about fiber optics. Many were right about e-commerce. Many were right about computers. What destroyed them was not the direction of history. It was the price paid to participate in it.

There is an elegant way to lose money: get the future right too early, too expensively, and with too much leverage.

The current market has that smell.

Not everywhere. Not in every company. Not in every asset. But in specific sectors, the rally has stopped looking like value discovery and started looking like an escape into the same story. When everyone needs to own the same thesis in order not to look like an idiot, the thesis has stopped being an opportunity. It has become a social tax.

The investor buys because the benchmark owns it. The manager buys because the client asks about it. The client asks because he saw it go up. The analyst raises the price target because the price went up. The journalist writes because the topic engages. The algorithm pushes because the theme retains attention. And suddenly, price becomes proof of its own truth.

That circle looks like collective intelligence. Very often, it is just contagion.

Today, contagion uses sophisticated words: infrastructure, compute, inference, capex, chips, hyperscalers, foundation models, structural demand. But the old question remains untouched: who pays the bill, over what time frame, at what margin, against what competition, with what cost of capital, and how much of this has already been anticipated in the price?

If these questions bother people, all the better. They are the right questions.

The market dislikes those who reduce a grand narrative to a spreadsheet. But every narrative, sooner or later, must pass through the customs office of cash flow. There, adjectives do not enter. Only margins, cost of capital, return on investment, depreciation, real demand, saturation, and time.

Artificial intelligence requires infrastructure. Infrastructure requires capital. Capital requires return. Return requires price. And price, that word the market tries to hide at the end of parties, is exactly where the investor should begin.

I am not worried that there is enthusiasm.

I am worried by the lack of shame in the enthusiasm.

Parabolic rises have a cruel feature: they look inevitable until the instant they start looking absurd. And that change in perception does not happen slowly, politely, academically. It happens all at once. The stock nobody wanted to sell at any price becomes the stock nobody wants to buy without a discount. Liquidity, which seemed infinite on the way up, disappears on the way down. The investor discovers that the price on the screen was a promise, not a guarantee.

That is why betting directly against it is a trap for most people.

Shorting is not an opinion. It is a risk structure. You can be right and go broke before you are proven right. You can identify the bubble and be crushed by its final gasp. You can buy protection too expensively and bleed while the market rises another 10%, 15%, 20%. You can turn a prudent thesis into a vain bet.

I do not want vanity.

I want survival.

The intelligent answer, for most investors, is not to seek fame by betting against the asylum. It is to reduce exposure. Cut excess. Decrease concentration. Sell what you would not buy today. Raise cash. Stop mistaking unrealized gains for intelligence. Stop believing a position is good merely because it has gone up.

The correct question is brutal:

Would I buy this today, at this price, with new money?

If the answer is no, the position may not be conviction. It may be attachment with a profit.

Cash is humiliated in periods like these. They call it idle money. They call it fear. They call it opportunity cost. Cash does not defend itself in a bull market. It stays quiet, looking stupid, while everyone around it shines. Then, when the market remembers that gravity has not been repealed, cash stops being stupid and becomes oxygen.

Serious money is not born from the obligation to participate in every rally.

It is born from the ability to be alive when others need to sell.

That is the part few accept. Investment is not a continuous proof of intelligence before others. It is a private discipline of survival. In certain periods, survival means looking less brilliant. It means accepting that you may leave money on the table. It means not fighting for the last glass of wine before the music stops.

I would rather look too early than be trapped too late.

Perhaps the party continues. Perhaps for weeks. Perhaps for months. Perhaps for another year. That does not invalidate the argument. On the contrary, it may strengthen it. The longer price drifts away from prudence, the larger the bill tends to be when prudence returns. Bubbles do not die because skeptics win the debate. They die because marginal buyers run out.

And when they run out, the narrative discovers it had no legs. It had flows.

The catalyst may come from oil. It may come from a geopolitical conflict. It may come from private credit. It may come from data centers financed on expectations that were too aggressive. It may come from a surprise in the U.S. Treasury market. It may come from inflation, interest rates, liquidity, balance sheets, guidance, capex revisions, margin compression, or simple exhaustion.

Or it may come from nowhere.

That is the part that frightens those who need clean explanations. Sometimes the market falls because it has risen too much. Sometimes the proximate cause is merely an elegant excuse for the real cause: price. A high price is a cause. Concentration is a cause. Internal fragility is a cause. Complacency is a cause. A lack of new buyers is a cause.

The market prefers blaming events because events appear external. Blaming price would require admitting responsibility.

I do not know whether we are exactly in 1999, 2000, 1929, or 2007. That literal comparison matters less than it seems. History does not return wearing the same clothes. It returns with the same human behavior. Greed, fear of missing out, faith in new narratives, contempt for risk, extrapolation of recent returns, ridicule of caution, concentration in obvious winners, and the final belief that the cycle has been domesticated.

The name changes. The instinct does not.

AI may become the engine of a new economy. But even extraordinary engines can be installed in cars bought at idiotic prices. The investor who does not understand this is condemned to confuse technological progress with financial return.

The question is not whether AI will be big.

The question is who will capture value, how much value they will capture, for how long, at what margin, and how much the market has already charged in advance for that capture.

That is the point.

When everyone buys the future, the present loses its margin of safety. And without a margin of safety, the investor becomes dependent on only one thing: that someone else will accept paying even more tomorrow. That is not investment. It is a transfer of anxiety.

I look at this market and I do not see only opportunities. I see traps dressed as inevitability. I see investors repeating correct phrases at wrong prices. I see technology being used as a safe-conduct pass to suspend arithmetic. I see intelligent people committing old mistakes with modern vocabulary.

There is no need to abandon the entire market. No need to become a prophet of disaster. No need to sell everything, buy cans of sardines, and wait for the end of the world. That too is theater. What is necessary is calibration. Reduce where excess is obvious. Preserve where asymmetry has disappeared. Reassess concentration. Treat parabolic rises as risk, not as proof of safety.

The greatest mistake investors make in moments like this is to think prudence means pessimism.

Prudence is not pessimism. It is respect for the possibility of being wrong. It is understanding that the investor’s job is not to win every month. It is to remain in the game when the bad months arrive. It is knowing that a portfolio should not be built to impress during euphoria, but to survive when euphoria becomes an embarrassing memory.

I have no interest in convincing those who need the crowd in order to think.

The market will always have priests of the present. They will explain why the multiples make sense, why the technology is different, why demand is structural, why the cycle has changed, why the old models no longer apply, why the skeptics do not understand, why the price is still reasonable if you look far enough out. The future, conveniently, always allows excess to be hidden on the horizon.

But investment is not theology.

It is price, risk, time, and survival.

Artificial intelligence may be right. The price is not.

That sentence should sit on every investor’s desk in this cycle. Not to deny the technology. To prevent the technology from becoming anesthesia. The market does not destroy investors only with lies. It destroys them, above all, with truths bought too expensively.

And perhaps this is the most dangerous truth of the moment: AI is real.

Because it is real, it persuades better. Because it persuades better, it allows worse prices. Because it allows worse prices, it attracts too much capital. Because it attracts too much capital, it reduces future returns. Because it reduces future returns, it prepares disappointment. Not despite being a great technology, but precisely because it is great enough to seduce everyone at the same time.

Risk rarely looks like risk while it is working.

It looks like intelligence. It looks like courage. It looks like vision. It looks like “being on the right side of history.” But the right side of history protects no one from buying badly. History belongs to civilization. Price belongs to the investor.

And the investor pays for the error alone.

I do not write to predict the end of the world. I write to remember that the world does not need to end for a stock to fall 50%. A company does not need to go bankrupt to destroy returns. A thesis does not need to be false to become a terrible investment. It is enough for the price to have demanded perfection and for reality to deliver only excellence.

Excellence is not enough when the price demands a miracle.

That is the trap.

Therefore, I prefer the less seductive posture: less exposure where there is euphoria, more cash where there is complacency, more patience where there is hurry, more skepticism where there is consensus. Not because I know the day. Because I do not need to know the day to recognize a crowded room, a small door, and music that is playing too loudly.

The wolf may not arrive tomorrow.

But when everyone starts laughing at the alarm, I start looking for the exit.

Leo Bentier

The Cold Read of the Musk-Altman Conflict: When OpenAI's Founding Mission Became a $150 Billion Betrayal

A cold reading of the Musk-Altman trial in Oakland: The founding mission was not abandoned by accident. It was exchanged for valuation.

The trial that began this week in Oakland is not just another Silicon Valley billionaire feud. It is the public autopsy of a promise made in 2015: that artificial intelligence would be developed for the benefit of all humanity, not to enrich a few men in a closed room. Elon Musk is suing Sam Altman, Greg Brockman, and OpenAI for breach of charitable trust, fraud, and unjust enrichment. He is asking the court to remove Altman and Brockman, unwind the for-profit structure, and return tens of billions to the original nonprofit mission. This is not drama for the sake of drama. It is the cleanest case study yet of how the AI industry's language of safety and humanity collides with the physical and economic realities of building the most powerful technology in history.

The Original Pact

OpenAI was born as a nonprofit. Its charter was explicit: build AGI for the benefit of humanity, keep the technology open, and avoid capture by any single corporation. Musk was the largest early donor precisely because he believed the alternative, closed-source AGI controlled by profit-maximizing entities, represented an existential risk. Altman and Brockman signed the same documents. Musk left the board in 2018, citing conflicts with Tesla's AI work. But the founding bargain remained. What followed was the classic turn. OpenAI created a capped-profit subsidiary, took billions from Microsoft, closed the models, and became what Musk's lawsuit calls a de facto subsidiary of the largest technology company in the world. Its valuation exploded past $150 billion. The mission, according to the complaint, was quietly placed in the drawer.

The Physical and Strategic Reality

This is where the cold read becomes uncomfortable for the hype machine. AI is not just software in the cloud. It is chips, energy, water, data centers, and geopolitical power. The essays on this site have shown again and again that scalable intelligence requires physical infrastructure no company, and no nation, can monopolize without consequences. Yet OpenAI's path has been one of concentration: proprietary models, exclusive partnerships, and governance that prioritizes investor return over the original nonprofit charter. Musk's position, by contrast, has been consistent: competition and transparency are the only realistic guardians. xAI was created explicitly to accelerate the understanding of the universe, not to sell chatbots or trap users inside an ecosystem. Critics call this rivalry. The record shows Musk warning for years about exactly the kind of concentrated power OpenAI now represents. He is not asking for a monopoly. He is demanding that the original promise be honored.

The “Jealousy” Narrative Does Not Survive Contact with Facts

OpenAI and its defenders try to frame Musk's lawsuit as the resentment of a competitor. Convenient. It ignores the fact that Musk has repeatedly offered to return any damages to OpenAI's nonprofit arm if he wins. It also ignores the timeline: Musk's public criticism of closed-source AGI predates xAI by several years. This lawsuit is not about market share. It is about whether a charitable mission can be converted into private wealth without consequence. The trial will expose emails, messages, and board minutes showing how quickly the language of benefiting humanity was subordinated to fundraising and valuation. Expect testimony from Altman, Brockman, and even Microsoft's Satya Nadella. The jury will not merely decide legal liability. It will decide something larger: whether Silicon Valley can still be trusted when it wraps profit in the language of existential salvation.

Why This Matters Beyond the Courtroom

This case is a stress test for the entire AI ecosystem. If a nonprofit charter can be discarded the moment the money arrives, then every sermon about safety, alignment, and democratization is provisional, subject to the next funding round. Sovereign nations are already waking up to this. AI infrastructure is becoming a matter of national security, not corporate convenience. Regulation is coming not because governments hate innovation, but because the alternative is unregulated private power over the most strategic technology of the century. Musk's side carries more weight in this equation. He bet early, funded the mission, and left when the direction changed. He continues to defend multiple competing efforts precisely because the concentration of AGI inside a single closed entity is the scenario he always feared. Altman's achievement in scaling ChatGPT is undeniable. But scaling is not fidelity. The physical chain of AI, energy, chips, geopolitics, does not care about press releases. It demands realism. The Musk-Altman trial is forcing that realism into public view. Whatever the jury decides, the public record will show that the founding mission was not abandoned by accident. It was exchanged for valuation. And in the long game of artificial intelligence, history rarely forgives those who treat existential infrastructure as just another startup exit.

Leo Bentier

Sequoia Is Right. And Still Incomplete.

The Error Is Not in Execution. It Is in the Decision.

On March 5, 2026, Sequoia Capital published an essay that quickly became a reference point in the debate over the future of software: "Services: The New Software." Written by Julien Bek, the piece makes a claim that, at first glance, sounds merely provocative. It is not. It is structural.

"The next $1T company will be a software company masquerading as a services firm."

The idea is simple enough to be ignored by those who do not pay attention, and deep enough to redefine capital allocation.

For decades, software sold tools. The customer bought capacity. Potential. Access. The work, however, was still done by people. Excel does not deliver the report. The CRM does not close the deal. Accounting software does not close the books. It merely organizes human effort.

Sequoia is saying this model is over.

The new software does not sell the tool. It sells the work done.

And by doing so, it shifts the center of value capture. Because the relevant budget was never software. It was always labor. Sequoia illustrates this with an asymmetry that, once seen, cannot be unseen: companies spend a fraction on software and multiples of that on the services required to operate that software. The expensive part is not the system. It is the person using it.

The consequence is unavoidable. Once software begins to execute, it stops competing with other software and starts competing with people. And people are the largest cost inside an organization.

So far, the thesis is correct. More than correct. It is inevitable.

But this is where the argument begins to fail.

Sequoia describes a transformation at the level of execution. What it does not describe, and perhaps has not fully perceived yet, is that execution was never the main problem inside companies.

Companies do not fail because they cannot execute tasks. They fail because they execute the wrong tasks. Because they prioritize badly. Because they interpret signals inconsistently. Because they make decisions under noise, not under system.

Automating execution does not solve this. It amplifies it.

If the decision is right, automatic execution is a multiplier of efficiency. If the decision is wrong, automatic execution is a multiplier of error.

This distinction is uncomfortable because it moves the problem to a less visible place. Execution is tangible. It can be measured, optimized, automated. Decision cannot. Decision is diffuse, contextual, dependent on memory, feedback, and interdependence between variables. It does not fit neatly into a simple interface.

But precisely for that reason, it concentrates value.

There is a fundamental difference between markets and companies that is rarely examined with the seriousness it deserves.

Markets are systems.

Companies are organized improvisations.

In financial markets, principles are tested, refined, transformed into rules. Rules are encoded into systems. Systems execute. The human interacts with the system, but the human is not the primary decision point. The system is.

Inside companies, the process remains incomplete. Principles exist: strategy, culture, directives. But they do not become consistent operational rules. They do not become systems. The result is predictable: decisions fall back onto individuals. And individuals do not scale.

This creates a pattern that repeats with almost statistical regularity. The same types of companies, with the same leadership profiles, under similar conditions, produce similar results. And still, they continue to operate as if every decision were unique, as if every problem were new, as if isolated human judgment were sufficient.

It is not.

Sequoia's thesis misses this point because it starts from an implicit assumption: that the work to be executed has already been correctly defined.

In practice, it has not.

Most of the invisible cost inside a company is not in execution. It is in defining what should be executed. It is in decisions made too late, too early, or simply wrong. It is in opportunities not prioritized, risks not detected, actions not initiated.

This does not appear in the budget. It is not classified as a cost line. But it determines the result.

And, more importantly, it is not solved by task automation.

There is a structural limit to the "services-as-software" model that few are willing to admit.

Execution can be compressed. It can be standardized. It can be replicated. As models advance, every operational workflow tends toward automation. This is not a hypothesis. It is a direction.

But what happens when everyone automates execution?

Execution loses differentiation.

If everyone can close the books automatically, the value is no longer in closing the books. If everyone can generate outreach, screen candidates, answer tickets, the value is no longer in those tasks. It is in deciding which books matter, which candidates to hire, which customers to prioritize, which actions to take.

Execution converges.

Decision does not.

Decision depends on accumulated context, operational memory, relationships between events, and the interpretation of weak signals. It is not merely calculation. It is structure.

And structure is not easily copied.

The error in the thesis is not what it says. It is what it omits.

It describes the industrialization of execution.

But ignores the systematization of decision.

And that is the level where the game is actually decided.

Today, most corporate systems operate through a broken flow. Data is collected. Interfaces organize the data. Humans interpret. Humans decide. Systems execute. Feedback is fragmented and rarely incorporated in a consistent way.

Sequoia's implicit proposal removes part of this flow. It removes the human from execution. But it keeps the human in the most critical point: the decision.

That is partial optimization.

The complete system is something else.

A complete system does not separate data, decision, and execution. It integrates them. It connects context, preserves memory, structures decisions, and closes the loop through execution and continuous feedback. It does not depend on episodic intervention. It does not depend on isolated interpretation. It operates as a cycle.

Detect. Interpret. Decide. Execute. Measure. Adjust.

Without rupture.

The consequence of this shift is less technological than strategic.

Whoever controls execution gains efficiency.

Whoever controls decision gains direction.

And direction precedes efficiency.

It is possible to execute something irrelevant perfectly. It is impossible to decide correctly without context.

The real control of an organization is not in who does the work. It is in who defines the work.

The trajectory of software makes this clear.

Traditional SaaS increased productivity. The new generation, driven by AI, begins to replace labor.

The next stage will not merely replace labor.

It will replace judgment.

Not in the philosophical sense. In the operational sense.

Systems capable of deciding, consistently, what should be done next, based on memory and context, not merely on instruction.

This does not eliminate the human. It repositions him.

The human stops being the executor. He stops being even the primary decision-maker. He becomes the supervisor of the system.

There is a risk attached to this transition, and it is rarely discussed with the seriousness it deserves.

Automating execution without systematizing decision does not merely fail to solve the problem. It amplifies it.

If a company already makes inconsistent decisions, giving it a system that executes faster does not improve the result. It accelerates the error. It reduces the time between bad decision and consequence.

Efficiency, in this context, is dangerous.

Without decision coherence, it is only speed.

Sequoia Capital is right to say that software will stop selling tools.

But that is not what defines the next generation of dominant companies.

What defines it is not the ability to execute work.

It is the ability to decide work.

Services-as-software solves the problem of execution. But it does not solve the problem of direction.

And without direction, execution is irrelevant.

The final distinction is simple, but not trivial.

Software that executes wins workflows.

Software that decides controls companies.

Leo Bentier

The Physical Chain of AI Is the Blind Spot of the Hype

A cold reading of pressure on energy, chips, water, and data centers: Without the physical world, scalable artificial intelligence does not exist.

Most executives would read this signal as news. That is the first mistake. News is what arrives after the market has found a comfortable word for the change; a signal is what appears before that, crooked, incomplete, and badly priced. In 2026-04, pressure on energy, chips, water, and data centers already pointed to a structural shift, not an isolated episode. The point was not to guess the next headline. The point was to see that the system was beginning to punish companies without cash, operational memory, decision discipline, or an honest relationship with the cost of their own growth.

The correct reading was less theatrical and more severe: without the physical world, scalable artificial intelligence does not exist Anyone who understood this did not need to pose as a prophet. He only needed to reject the managerial superstition that good outcomes prove good processes. Many companies grow because the wind helps them, not because they know how to sail. When the wind turns, you discover who had a system and who had only busy people, clean spreadsheets, long meetings, and a private museum of opinions sold internally as strategy.

That is the kind of reading I would have recorded without asking consensus for permission. Businesses do not break on the day the market notices; they break when the organization loses the ability to turn context into decision, decision into execution, and execution into correction. The event of 2026-04 would have been used as a lens, not as a historical fetish. The lesson was simple, therefore almost always ignored: the company that does not build a system for deciding will be decided by the environment.

Leo Bentier

The Global Economy Still Grows, but With Less Innocence

A cold reading of the global environment of fragile growth: AI, tariffs, and geopolitics pull in opposite directions.

2026-03 deserved a cold essay because the global environment of fragile growth was not a calendar detail. It was a test of corporate sanity. Whenever a shock looks too external to be absorbed by management, it reveals an internal weakness that was already being ignored. The common executive turns surprise into an excuse; the rare executive turns surprise into architecture. The difference between the two is not raw intelligence. It is the willingness to place the company against reality before reality places the company against the wall.

AI, tariffs, and geopolitics pull in opposite directions That sentence should sit at the center of the article, because it forces the reader to abandon the comfort of simple causality. Companies are not defeated only by competitors, technologies, crises, or governments. They are defeated by slow decisions, scattered memory, disconnected metrics, crooked incentives, and the inability to convert small signals into coordinated action. The world changes first in narrow margins; then it appears in quarterly reports, when reaction has become expensive.

The intellectual signature here is not to confuse pessimism with precision. The point was not to root against companies, markets, or institutions. The point was to recognize that optimism without a mechanism is just emotional marketing. If I had been writing this archive in that month, I would have closed with the same demand: less opinion, more system; less narrative, more execution; less worship of growth, more respect for the invisible cost of operating badly for long enough.

Leo Bentier

AI Will Not Be Borderless. It Will Be Sovereign, Auditable, and Expensive

A cold reading of the dispute over AI infrastructure and regulation: Location of data and models would become strategy.

The comfortable error would have been to call the dispute over AI infrastructure and regulation an exception. Exception is a word used by people who do not want to revise their model of the world. In 2026-02, the important signal was not the public spectacle but the hidden mechanism: bad incentives, borrowed confidence, dependence on outsiders, and weak correction loops. The average manager hunts for culprits after impact; the serious operator asks which rules allowed the impact to remain invisible for so long.

The thesis that mattered was brutal: location of data and models would become strategy That changes the whole conversation. If a company depends on friendly macro conditions, cheap capital, disciplined suppliers, patient customers, or heroic employees, it does not have a robust operation. It has a temporarily financed fiction. Most management fails because it tries to look sophisticated before it becomes true, and truth almost always begins with an unpleasant question about fragility, not with a beautiful deck.

I would have written it as a warning, not as mystical prediction. Prediction is vanity when it does not produce a rule of action. What matters is deciding before the crowd discovers the vocabulary. In every cycle, the same kinds of people, under the same conditions, produce the same outcomes because their internal systems do not change. The opportunity was to build a discipline: observe the signal, reduce ambiguity, choose the action, measure the deviation, and correct without romance.

Leo Bentier

The Turbulent Decade Reached Its Midpoint

A cold reading of the beginning of 2026 under geopolitical and technological risk: Risk stopped being an event and became an environment.

Most executives would read this signal as news. That is the first mistake. News is what arrives after the market has found a comfortable word for the change; a signal is what appears before that, crooked, incomplete, and badly priced. In 2026-01, the beginning of 2026 under geopolitical and technological risk already pointed to a structural shift, not an isolated episode. The point was not to guess the next headline. The point was to see that the system was beginning to punish companies without cash, operational memory, decision discipline, or an honest relationship with the cost of their own growth.

The correct reading was less theatrical and more severe: risk stopped being an event and became an environment Anyone who understood this did not need to pose as a prophet. He only needed to reject the managerial superstition that good outcomes prove good processes. Many companies grow because the wind helps them, not because they know how to sail. When the wind turns, you discover who had a system and who had only busy people, clean spreadsheets, long meetings, and a private museum of opinions sold internally as strategy.

That is the kind of reading I would have recorded without asking consensus for permission. Businesses do not break on the day the market notices; they break when the organization loses the ability to turn context into decision, decision into execution, and execution into correction. The event of 2026-01 would have been used as a lens, not as a historical fetish. The lesson was simple, therefore almost always ignored: the company that does not build a system for deciding will be decided by the environment.

Leo Bentier

The State wants the base, the bank wants the margin, the entrepreneur pays both

After nearly twenty years, the final thesis is not that banks are evil or that credit should be cheap; it is that price is a consequence of where the fear lives — and the job of whoever designs is to step back one stage and reorganize the risk.

The debate on corporate credit in 2025 is contaminated by a tempting number: 49% per year. It circulates as if it were the average rate of all credit to companies. It is not. Central Bank aggregates sit far below that, while certain unsecured, revolving, or short-term lines can approach or exceed that level. The correction does not weaken the thesis. It strengthens it. The entrepreneur does not pay a single Brazilian rate; he pays the product that was pushed on him when he arrived without structure.

The State wants the base, the bank wants the margin, the entrepreneur pays both. Taxes, reserve requirements, cost of capital, regulation, defaults, expenses, and concentration enter the price. But the individual contract adds another layer: the debtor's inability to present a risk financeable through different channels. He arrives with urgency, incomplete documents, mixed assets, and collateral already committed. The bank does not need to design. It offers the most profitable shelf compatible with the limit.

An average of corporate credit hides worlds. Large companies issue debentures, raise money abroad, negotiate secured lines, and access the capital market. Small companies use overdrafts, guaranteed accounts, receivables discounting, cards, and working capital. Putting them all in one number is statistically correct and economically useless. The average entrepreneur does not live at the average rate; he lives in the tail of the product he manages to contract.

That is why repeating "corporate rates at 49%" as a general truth would be an error of authority. The attentive reader will find the correct series and conclude that the whole argument was built on exaggeration. We do not need it. It is enough to observe the difference between lines, the persistence of high spreads, and the cost of unsecured credit. Indignation gains reach with big numbers; the thesis gains duration with exact ones.

The relevant question is why a company accepts credit at forty, fifty, or more percent per year. Sometimes because the margin supports a short bridge. Sometimes because losing inventory, payroll, or a supplier costs more. Often because the borrower computes the installment, not the rate, and believes future cash will solve it. The short-term product monetizes urgency. The bank does not sell capital; it sells time in small, expensive packages.

The State's cost also appears indirectly. Taxes on intermediation, regulatory requirements, and the financing of public debt raise the alternative return. When the government pays liquidity well at sovereign risk, private credit must offer more. The entrepreneur competes with the Treasury for domestic capital. Then he is told he should invest to raise productivity. The country charges a premium to finance production and calls the result a lack of entrepreneurial ambition.

The bank, in turn, exercises real power. Concentration, proprietary data, the current account, and collateral create dependence. An institution knows the client's history, but that knowledge is not fully portable. The entrepreneur changes banks and becomes unknown again. The information that should reduce risk becomes a competitive barrier. The good payer has no financial reputation; the bank has a private dossier on him.

Open Finance promises to reduce part of this, but transported data is not transported decision. Another institution must interpret, trust, and integrate. The company must consent and understand. Fraud and categorization quality limit models. The infrastructure is progress, not magic. The price will fall when new creditors can use the data to control flow and recover capital, not merely to offer another card.

The biggest blind spot lies outside the banking system. Companies finance one another at gigantic scale. Suppliers grant thirty, sixty, or ninety days to close the sale. Distributors raise limits to hit quotas. Manufacturers carry clients who should already have been blocked. This commercial credit does not appear in the interest-rate headlines, but it consumes capital, generates losses, and can break the supplier. The entrepreneur complains about the bank while running a free bank for his clients.

A corporate credit decision platform should start there. Before the sale, evaluate existing exposure, behavior, margin, sector, and concentration. Recommend limit, tenor, down payment, collateral, and collection terms. Then monitor events, invoices, protests, corporate changes, and delays. The value is not in saying whether someone has a bad record. It is in deciding whether to sell, how much, on what terms, and when to stop.

Such a system turns credit from reaction into policy. Today, the commercial area grants and the financial area collects. The incentives are opposed. The salesperson earns commission on billing, not collection; the finance team inherits the client after the risk has been taken. A mature structure ties compensation to the quality of the revenue and makes collection start at the contract. Default is not born at maturity; it is born in the bad approval.

The bank can better finance a company that controls its own commercial credit. Verifiable portfolios, consistent limits, and documented collection make receivables more financeable. Thus an operational tool reduces banking cost indirectly. The debtor stops presenting only billing and starts presenting an asset with performance data. Technology does not negotiate a few points; it changes the financed object.

My position in 2025 would be to invest less in the "cheap digital bank" narrative and more in the infrastructure that enables design. Registrars, receivables systems, vertical analysis, automated collection, collateral agents, and platforms connecting operations to capital. Loan distribution has become a commodity. Exclusive information, control, and recovery remain scarce.

In banks, I would assess the sustainability of margins in the face of Open Finance, fintechs, and the capital market, but I would not bet on a fast collapse. Incumbents have funding, trust, data, and cross-selling capacity. The most dangerous attack will not come from a fintech charging less on the same loan. It will come from systems that keep the client from needing that loan or that turn his assets into instruments financeable by others.

For companies, I would separate three boxes. Short-term operational credit for the cash cycle; investment debt with the asset's tenor; and risk capital for uncertainty that should not be promised as repayment. Mixing the three produces destruction. Startups finance experiments with debt; manufacturers finance long machinery with working capital; entrepreneurs finance recurring losses with prepayment. The problem is not only the rate. It is using the right obligation for the wrong risk.

I would also preserve collateral. The Collateral Framework allows better use, but every committed asset reduces future options. A company that extracts the maximum from its property today may discover tomorrow it has no collateral left to renegotiate. Loan-to-value should be a survival limit, not an efficiency target. The creditor likes full collateral; the debtor should like slack.

A 49% rate can be economically rational in a very short operation with high returns and certain payment. It can be suicidal in permanent capital. Annualizing short costs frightens, but ignoring annualization deceives. The analysis must combine absolute cost, duration, and the return on the use. The problem with the public discourse is picking one number and removing the contract around it.

The entrepreneur must compute the effective cost in reais, the break-even point, and the delay scenario. How much additional profit must the money generate? What happens if the revenue arrives thirty days late? Is there a prepayment penalty? Which collateral gets blocked? The bank wins because many clients know the installment and ignore the economic transaction. Corporate financial education begins with the purpose, not the formula.

The State should reduce registration friction, improve recovery, and allow competition, instead of announcing episodic lines for every crisis. Public guarantees can answer systemic shocks, as in 2020, but they should not replace the market. The goal is for good risks to be recognized without a godfather and for bad risks to be refused before consuming public wealth.

Brazilian corporate credit will remain expensive where risk is indistinguishable. Some sectors, lines, and companies will pay close to 49%; others much less. The inequality of price is part of the information. The job is not to pretend a single rate exists and fight it. It is to build paths for the company to leave the wrong line.

The State wants the base, the bank wants the margin, and the entrepreneur pays both. But there is a third charge, self-inflicted: the price of disorganization. A company that does not know its own cycle, grants terms without a policy, mixes assets, and seeks capital in an emergency hands power to the branch. No revolt against interest rates fixes that.

The final thesis, after nearly twenty years, is not that banks are evil or that credit should be cheap. It is that price is a consequence of where the fear lives. In 2008, trust died and collateral survived. In 2020, the public guarantee displaced the fear. In 2023, a badly presented debt destroyed legibility. In 2025, the entrepreneur still looks at the rate as the cause. The job of whoever designs is to step back one stage and reorganize the risk.

I would not publish "corporate rates at 49%, the highest since 2017" without qualifying the line and the series. I would publish something harder to refute: "some companies pay close to 49% because they arrive at the bank with a problem that only fits the most expensive product." The first sentence produces indignation. The second produces a business.

Leo Bentier

2026 Will Be the Year of Infrastructure: Energy, Chips, Data, and Security

A cold reading of the end of 2025: AI would leave the slide deck and hit the physical world.

2025-12 deserved a cold essay because the end of 2025 was not a calendar detail. It was a test of corporate sanity. Whenever a shock looks too external to be absorbed by management, it reveals an internal weakness that was already being ignored. The common executive turns surprise into an excuse; the rare executive turns surprise into architecture. The difference between the two is not raw intelligence. It is the willingness to place the company against reality before reality places the company against the wall.

AI would leave the slide deck and hit the physical world That sentence should sit at the center of the article, because it forces the reader to abandon the comfort of simple causality. Companies are not defeated only by competitors, technologies, crises, or governments. They are defeated by slow decisions, scattered memory, disconnected metrics, crooked incentives, and the inability to convert small signals into coordinated action. The world changes first in narrow margins; then it appears in quarterly reports, when reaction has become expensive.

The intellectual signature here is not to confuse pessimism with precision. The point was not to root against companies, markets, or institutions. The point was to recognize that optimism without a mechanism is just emotional marketing. If I had been writing this archive in that month, I would have closed with the same demand: less opinion, more system; less narrative, more execution; less worship of growth, more respect for the invisible cost of operating badly for long enough.

Leo Bentier

Childhood Became the Regulatory Frontier of Social Technology

A cold reading of the debate over minimum age for social networks and AI: Social technology would face political limits.

The comfortable error would have been to call the debate over minimum age for social networks and AI an exception. Exception is a word used by people who do not want to revise their model of the world. In 2025-11, the important signal was not the public spectacle but the hidden mechanism: bad incentives, borrowed confidence, dependence on outsiders, and weak correction loops. The average manager hunts for culprits after impact; the serious operator asks which rules allowed the impact to remain invisible for so long.

The thesis that mattered was brutal: social technology would face political limits That changes the whole conversation. If a company depends on friendly macro conditions, cheap capital, disciplined suppliers, patient customers, or heroic employees, it does not have a robust operation. It has a temporarily financed fiction. Most management fails because it tries to look sophisticated before it becomes true, and truth almost always begins with an unpleasant question about fragility, not with a beautiful deck.

I would have written it as a warning, not as mystical prediction. Prediction is vanity when it does not produce a rule of action. What matters is deciding before the crowd discovers the vocabulary. In every cycle, the same kinds of people, under the same conditions, produce the same outcomes because their internal systems do not change. The opportunity was to build a discipline: observe the signal, reduce ambiguity, choose the action, measure the deviation, and correct without romance.

Leo Bentier

The Trade War Is Now a War of Routes, Chips, and Energy

A cold reading of global trade fragmentation: Corporate strategy became a geopolitical map.

Most executives would read this signal as news. That is the first mistake. News is what arrives after the market has found a comfortable word for the change; a signal is what appears before that, crooked, incomplete, and badly priced. In 2025-10, global trade fragmentation already pointed to a structural shift, not an isolated episode. The point was not to guess the next headline. The point was to see that the system was beginning to punish companies without cash, operational memory, decision discipline, or an honest relationship with the cost of their own growth.

The correct reading was less theatrical and more severe: corporate strategy became a geopolitical map Anyone who understood this did not need to pose as a prophet. He only needed to reject the managerial superstition that good outcomes prove good processes. Many companies grow because the wind helps them, not because they know how to sail. When the wind turns, you discover who had a system and who had only busy people, clean spreadsheets, long meetings, and a private museum of opinions sold internally as strategy.

That is the kind of reading I would have recorded without asking consensus for permission. Businesses do not break on the day the market notices; they break when the organization loses the ability to turn context into decision, decision into execution, and execution into correction. The event of 2025-10 would have been used as a lens, not as a historical fetish. The lesson was simple, therefore almost always ignored: the company that does not build a system for deciding will be decided by the environment.

Leo Bentier

Bolsonaro will be judged less as a man than as a precedent

The law that does not reach the former ruler teaches the next ruler that office is license.

On September 11, 2025, the event looked like news. A serious letter would have treated it as a signal. News tells you what happened; a signal reveals what may repeat later with another name, another face, and another flag. Politics is a machine for recycling old passions with new vocabulary. Whoever reads only the event becomes hostage to the calendar. Whoever reads the incentive begins to see the future while it is still disguised as exception.

Bolsonaro's trial would decide not only a man's fate, but the future price of attempting to stay in power after defeat.

The argument would not be born from party preference, but from an older question: what kind of regime, elite, and people does this event reveal? Small politics asks who won today. Good politics asks which vice was rewarded, which virtue was punished, and which cost was pushed onto the next generation. This is where almost everyone fails. The commentator looks for convenient culprits; the statesman looks for mechanisms. The cheap moralist points the finger; the serious moralist asks why that finger found so many accomplices.

Brazil and the modern world suffer from the same disease with different accents: the confusion between form and substance. We have elections, but not always trust. Courts, but not always perceived justice. Markets, but not always responsibility. State, but not always common good. Communication, but not always truth. Data, but not always wisdom. Modernity multiplied instruments and impoverished judgment. The result is a civilization technically equipped to err at speed.

What could be seen on this date was less an isolated fact than a movement of plates. The surface would still allow calm speeches. It always does. Before ruptures, ministers guarantee normality, analysts calculate small probabilities, newspapers choose moderate adjectives, and investors pretend political risk fits in a spreadsheet. But history does not move only when there is an explosion. It moves when important men begin to lie differently. A change in the lie often precedes a change in the regime.

The letter would avoid the vice of the cheap prophet. The cheap prophet wants to look like owner of the future. The serious observer knows only that certain arrangements are fragile, certain incentives rot institutions, and certain accumulated humiliations look for a date. It is not necessary to predict the exact lightning strike when one knows the forest is dry. The vanity of guessing the day is inferior to the discipline of recognizing flammable wood.

There would also be a warning to the Brazilian reader. Brazil likes importing foreign explanations without importing the structures that make them true. It speaks of liberalism without impersonality, strong state without efficiency, sovereignty without strategy, democracy without civilized losers, market without competition, conservatism without self-control, progressivism without fiscal responsibility. That is why the country spends decades debating the vocabulary of great politics while preserving the habits of the small court.

Cicero would have seen in this episode a question about the republic: does law serve the city or the factions? Augustine would ask whether power, separated from justice, is truly government or only a gang with protocol. Aquinas would remind us that the common good is not the sum of private appetites, but a moral order that allows the city to seek higher ends. Marcus Aurelius would distrust applause, panic, and indignation, for all are poor emotional states for government. Churchill would look for will in the event, not excuse. Plato would ask what kind of collective soul chose this moment.

The forecast, then, would be simple and hard. The event recorded here would not end in itself. It would create habit, precedent, resentment, architecture, or language. The first error of observers would be to treat it as a period. The second would be to judge it only by the immediate winner. The third would be to believe the institution touched by the event would emerge unchanged. Institutions are not buildings; they are memories of behavior. When behavior changes, the building may remain standing while the republic has already changed skin.

The conviction would make Brazil a rare case of accountability for a coup attempt, while keeping alive the accusation of persecution among his followers.

In retrospect, the essential point would be this: almost every major political event first appears as anomaly, then as repetition, then as doctrine. The mistake is laughing at the anomaly. The opportunity is asking why it appeared now, why it found an audience, and who gains from its normalization.

There is a difference between good politics and cheering. Cheering needs innocents and monsters. Good politics needs causes, limits, and consequences. Cheering wants to absolve its own and condemn the others. Good politics asks which sin is common to both sides. Cheering calls prudence cowardice when it is winning and persecution when it is losing. Good politics knows that every faction, when it believes itself redemptive, begins rehearsing small tyrannies.

This is why the letter would not ask the reader for comfort. It would ask for maturity. The world is not governed by declared intentions, but by tolerated incentives. An elite can speak of democracy while despising the people. A people can speak of liberty while desiring revenge. A market can speak of efficiency while asking for public rescue. A state can speak of justice while expanding its own appetite. A power can speak of order while surrounding rivals. A rival can speak of security while preparing aggression.

The reader who wants to understand the future should abandon the superstition of proper names. Names matter, but less than mechanisms. The man passes; the incentive remains. The party changes; the method persists. The crisis ends; the learned fragility moves elsewhere. The modern world is a school of disguises. Whoever clings to the disguise always arrives late.

A good letter does not end by asking for agreement. It ends by leaving useful discomfort.

The law that does not reach the former ruler teaches the next ruler that office is license.

Leo Bentier

The Next Advantage Will Be Proprietary Operational Memory

A cold reading of the partial commoditization of general models: Context will be the differentiator.

2025-09 deserved a cold essay because the partial commoditization of general models was not a calendar detail. It was a test of corporate sanity. Whenever a shock looks too external to be absorbed by management, it reveals an internal weakness that was already being ignored. The common executive turns surprise into an excuse; the rare executive turns surprise into architecture. The difference between the two is not raw intelligence. It is the willingness to place the company against reality before reality places the company against the wall.

Context will be the differentiator That sentence should sit at the center of the article, because it forces the reader to abandon the comfort of simple causality. Companies are not defeated only by competitors, technologies, crises, or governments. They are defeated by slow decisions, scattered memory, disconnected metrics, crooked incentives, and the inability to convert small signals into coordinated action. The world changes first in narrow margins; then it appears in quarterly reports, when reaction has become expensive.

The intellectual signature here is not to confuse pessimism with precision. The point was not to root against companies, markets, or institutions. The point was to recognize that optimism without a mechanism is just emotional marketing. If I had been writing this archive in that month, I would have closed with the same demand: less opinion, more system; less narrative, more execution; less worship of growth, more respect for the invisible cost of operating badly for long enough.

Leo Bentier

The Biggest Risk in Corporate AI Is Decision Without Responsibility

A cold reading of the advance of autonomous agents: Agents require accountability.

The comfortable error would have been to call the advance of autonomous agents an exception. Exception is a word used by people who do not want to revise their model of the world. In 2025-08, the important signal was not the public spectacle but the hidden mechanism: bad incentives, borrowed confidence, dependence on outsiders, and weak correction loops. The average manager hunts for culprits after impact; the serious operator asks which rules allowed the impact to remain invisible for so long.

The thesis that mattered was brutal: agents require accountability That changes the whole conversation. If a company depends on friendly macro conditions, cheap capital, disciplined suppliers, patient customers, or heroic employees, it does not have a robust operation. It has a temporarily financed fiction. Most management fails because it tries to look sophisticated before it becomes true, and truth almost always begins with an unpleasant question about fragility, not with a beautiful deck.

I would have written it as a warning, not as mystical prediction. Prediction is vanity when it does not produce a rule of action. What matters is deciding before the crowd discovers the vocabulary. In every cycle, the same kinds of people, under the same conditions, produce the same outcomes because their internal systems do not change. The opportunity was to build a discipline: observe the signal, reduce ambiguity, choose the action, measure the deviation, and correct without romance.

Leo Bentier

Brazilian sovereignty now runs through Pix

The modern empire does not collect tribute only in gold; it collects through payment rails.

On July 10, 2025, the event looked like news. A serious letter would have treated it as a signal. News tells you what happened; a signal reveals what may repeat later with another name, another face, and another flag. Politics is a machine for recycling old passions with new vocabulary. Whoever reads only the event becomes hostage to the calendar. Whoever reads the incentive begins to see the future while it is still disguised as exception.

The Pix dispute showed that sovereignty in the twenty-first century lives in payment systems, data, currency, credit, digital infrastructure, and regulatory capacity.

The argument would not be born from party preference, but from an older question: what kind of regime, elite, and people does this event reveal? Small politics asks who won today. Good politics asks which vice was rewarded, which virtue was punished, and which cost was pushed onto the next generation. This is where almost everyone fails. The commentator looks for convenient culprits; the statesman looks for mechanisms. The cheap moralist points the finger; the serious moralist asks why that finger found so many accomplices.

Brazil and the modern world suffer from the same disease with different accents: the confusion between form and substance. We have elections, but not always trust. Courts, but not always perceived justice. Markets, but not always responsibility. State, but not always common good. Communication, but not always truth. Data, but not always wisdom. Modernity multiplied instruments and impoverished judgment. The result is a civilization technically equipped to err at speed.

What could be seen on this date was less an isolated fact than a movement of plates. The surface would still allow calm speeches. It always does. Before ruptures, ministers guarantee normality, analysts calculate small probabilities, newspapers choose moderate adjectives, and investors pretend political risk fits in a spreadsheet. But history does not move only when there is an explosion. It moves when important men begin to lie differently. A change in the lie often precedes a change in the regime.

The letter would avoid the vice of the cheap prophet. The cheap prophet wants to look like owner of the future. The serious observer knows only that certain arrangements are fragile, certain incentives rot institutions, and certain accumulated humiliations look for a date. It is not necessary to predict the exact lightning strike when one knows the forest is dry. The vanity of guessing the day is inferior to the discipline of recognizing flammable wood.

There would also be a warning to the Brazilian reader. Brazil likes importing foreign explanations without importing the structures that make them true. It speaks of liberalism without impersonality, strong state without efficiency, sovereignty without strategy, democracy without civilized losers, market without competition, conservatism without self-control, progressivism without fiscal responsibility. That is why the country spends decades debating the vocabulary of great politics while preserving the habits of the small court.

Cicero would have seen in this episode a question about the republic: does law serve the city or the factions? Augustine would ask whether power, separated from justice, is truly government or only a gang with protocol. Aquinas would remind us that the common good is not the sum of private appetites, but a moral order that allows the city to seek higher ends. Marcus Aurelius would distrust applause, panic, and indignation, for all are poor emotional states for government. Churchill would look for will in the event, not excuse. Plato would ask what kind of collective soul chose this moment.

The forecast, then, would be simple and hard. The event recorded here would not end in itself. It would create habit, precedent, resentment, architecture, or language. The first error of observers would be to treat it as a period. The second would be to judge it only by the immediate winner. The third would be to believe the institution touched by the event would emerge unchanged. Institutions are not buildings; they are memories of behavior. When behavior changes, the building may remain standing while the republic has already changed skin.

The American dispute over Pix, tariffs, and trade in 2026 would confirm that the financial battlefield had become political.

In retrospect, the essential point would be this: almost every major political event first appears as anomaly, then as repetition, then as doctrine. The mistake is laughing at the anomaly. The opportunity is asking why it appeared now, why it found an audience, and who gains from its normalization.

There is a difference between good politics and cheering. Cheering needs innocents and monsters. Good politics needs causes, limits, and consequences. Cheering wants to absolve its own and condemn the others. Good politics asks which sin is common to both sides. Cheering calls prudence cowardice when it is winning and persecution when it is losing. Good politics knows that every faction, when it believes itself redemptive, begins rehearsing small tyrannies.

This is why the letter would not ask the reader for comfort. It would ask for maturity. The world is not governed by declared intentions, but by tolerated incentives. An elite can speak of democracy while despising the people. A people can speak of liberty while desiring revenge. A market can speak of efficiency while asking for public rescue. A state can speak of justice while expanding its own appetite. A power can speak of order while surrounding rivals. A rival can speak of security while preparing aggression.

The reader who wants to understand the future should abandon the superstition of proper names. Names matter, but less than mechanisms. The man passes; the incentive remains. The party changes; the method persists. The crisis ends; the learned fragility moves elsewhere. The modern world is a school of disguises. Whoever clings to the disguise always arrives late.

A good letter does not end by asking for agreement. It ends by leaving useful discomfort.

The modern empire does not collect tribute only in gold; it collects through payment rails.

Leo Bentier

The AI Economy Must Prove It Is Not Just Hyperscaler Capex

A cold reading of massive spending on AI infrastructure: Infrastructure only matters if it creates distributed productivity.

Most executives would read this signal as news. That is the first mistake. News is what arrives after the market has found a comfortable word for the change; a signal is what appears before that, crooked, incomplete, and badly priced. In 2025-07, massive spending on AI infrastructure already pointed to a structural shift, not an isolated episode. The point was not to guess the next headline. The point was to see that the system was beginning to punish companies without cash, operational memory, decision discipline, or an honest relationship with the cost of their own growth.

The correct reading was less theatrical and more severe: infrastructure only matters if it creates distributed productivity Anyone who understood this did not need to pose as a prophet. He only needed to reject the managerial superstition that good outcomes prove good processes. Many companies grow because the wind helps them, not because they know how to sail. When the wind turns, you discover who had a system and who had only busy people, clean spreadsheets, long meetings, and a private museum of opinions sold internally as strategy.

That is the kind of reading I would have recorded without asking consensus for permission. Businesses do not break on the day the market notices; they break when the organization loses the ability to turn context into decision, decision into execution, and execution into correction. The event of 2025-07 would have been used as a lens, not as a historical fetish. The lesson was simple, therefore almost always ignored: the company that does not build a system for deciding will be decided by the environment.

Leo Bentier

The CFO Returned to the Center Because Capital Has a Cost Again

A cold reading of the still-restrictive rate environment: Growth must justify the balance sheet.

2025-06 deserved a cold essay because the still-restrictive rate environment was not a calendar detail. It was a test of corporate sanity. Whenever a shock looks too external to be absorbed by management, it reveals an internal weakness that was already being ignored. The common executive turns surprise into an excuse; the rare executive turns surprise into architecture. The difference between the two is not raw intelligence. It is the willingness to place the company against reality before reality places the company against the wall.

Growth must justify the balance sheet That sentence should sit at the center of the article, because it forces the reader to abandon the comfort of simple causality. Companies are not defeated only by competitors, technologies, crises, or governments. They are defeated by slow decisions, scattered memory, disconnected metrics, crooked incentives, and the inability to convert small signals into coordinated action. The world changes first in narrow margins; then it appears in quarterly reports, when reaction has become expensive.

The intellectual signature here is not to confuse pessimism with precision. The point was not to root against companies, markets, or institutions. The point was to recognize that optimism without a mechanism is just emotional marketing. If I had been writing this archive in that month, I would have closed with the same demand: less opinion, more system; less narrative, more execution; less worship of growth, more respect for the invisible cost of operating badly for long enough.

Leo Bentier

The Resilient Company Will Be Less Efficient in Excel and More Alive in Reality

A cold reading of the return of the resilience debate: Intelligent redundancy would become virtue again.

The comfortable error would have been to call the return of the resilience debate an exception. Exception is a word used by people who do not want to revise their model of the world. In 2025-05, the important signal was not the public spectacle but the hidden mechanism: bad incentives, borrowed confidence, dependence on outsiders, and weak correction loops. The average manager hunts for culprits after impact; the serious operator asks which rules allowed the impact to remain invisible for so long.

The thesis that mattered was brutal: intelligent redundancy would become virtue again That changes the whole conversation. If a company depends on friendly macro conditions, cheap capital, disciplined suppliers, patient customers, or heroic employees, it does not have a robust operation. It has a temporarily financed fiction. Most management fails because it tries to look sophisticated before it becomes true, and truth almost always begins with an unpleasant question about fragility, not with a beautiful deck.

I would have written it as a warning, not as mystical prediction. Prediction is vanity when it does not produce a rule of action. What matters is deciding before the crowd discovers the vocabulary. In every cycle, the same kinds of people, under the same conditions, produce the same outcomes because their internal systems do not change. The opportunity was to build a discipline: observe the signal, reduce ambiguity, choose the action, measure the deviation, and correct without romance.

Leo Bentier

Tariffs Show Globalization Now Has an Explicit Political Price

A cold reading of the return of tariff and industrial policies: Supply chains would become strategic design again.

Most executives would read this signal as news. That is the first mistake. News is what arrives after the market has found a comfortable word for the change; a signal is what appears before that, crooked, incomplete, and badly priced. In 2025-04, the return of tariff and industrial policies already pointed to a structural shift, not an isolated episode. The point was not to guess the next headline. The point was to see that the system was beginning to punish companies without cash, operational memory, decision discipline, or an honest relationship with the cost of their own growth.

The correct reading was less theatrical and more severe: supply chains would become strategic design again Anyone who understood this did not need to pose as a prophet. He only needed to reject the managerial superstition that good outcomes prove good processes. Many companies grow because the wind helps them, not because they know how to sail. When the wind turns, you discover who had a system and who had only busy people, clean spreadsheets, long meetings, and a private museum of opinions sold internally as strategy.

That is the kind of reading I would have recorded without asking consensus for permission. Businesses do not break on the day the market notices; they break when the organization loses the ability to turn context into decision, decision into execution, and execution into correction. The event of 2025-04 would have been used as a lens, not as a historical fetish. The lesson was simple, therefore almost always ignored: the company that does not build a system for deciding will be decided by the environment.

Leo Bentier

The Next Trade Is Not Training

AI's second economy will be measured in daily use: inference, cost per token, latency, and margin.

The first phase of generative AI was dominated by training. Larger models, larger clusters, larger datasets, benchmarks, launches, parameters, laboratories, frontier narratives. This phase is theatrical because training allows spectacle. "We built the largest." "We surpassed the previous one." "We launched a model." The market likes visible greatness. But businesses do not live on training alone. They live on repeated use.

The next trade is inference. Not as a technical word for insiders, but as the daily economy of AI. Every question, every answer, every agent, every copilot, every search, every summary, every API call, every image, every automated decision consumes capacity. Training is building the factory. Inference is operating the factory every day. The second can be larger, more recurring, and more cruel with costs.

Nvidia, AMD, and Palantir are obvious names. Nvidia because it remains at the center of accelerated hardware. AMD because inference can open room for alternatives if cost per token, availability, and customization matter. Palantir because corporate inference needs to enter operations, not merely answer loose questions. But Broadcom, Marvell, Arista, Credo, and Coherent may explain the next bottleneck.

Broadcom and Marvell can capture custom silicon, connectivity, and infrastructure. Arista captures networking. Credo captures high-speed interconnection. Coherent captures optical and photonic components relevant to data transmission. As inference grows, moving data cheaply, quickly, and efficiently can be as important as processing it. The AI factory is not a single brain. It is a network of brains, memory, communication, and energy.

Perhaps in 2026 the market begins to change obsession. Less "how many parameters?" More "what cost per token?" Less "what benchmark?" More "what margin per call?" Less "which model is most powerful?" More "which model is cheap enough to be used millions of times?" This change separates science from business. The most impressive model may not be the most profitable model.

The way to profit is to observe the inference chain. Companies will need to reduce latency, cost, energy consumption, and supplier dependence. There will be room for GPUs, alternative accelerators, ASICs, networks, switches, optics, active cables, optimization software, caching, model routing, compression, quantization, memory, and platforms that choose the right model for each task. Inference is an economic problem before it is merely technical.

Palantir enters because companies do not merely want to infer. They want to infer over internal data, with permission, context, and action. The value of the corporate token is not in the pretty answer. It is in the decision that changes operations. If AI reduces collection cycles, avoids fraud, prioritizes maintenance, improves logistics, or accelerates critical support, the cost per token can be justified. If it only generates generic text, it will be crushed by competition.

The counter-thesis is that inference can become a commodity. Smaller models can reduce demand for expensive hardware. Efficiency can compress supplier revenue. Hyperscalers can internalize silicon. Broadcom and Marvell can win or lose depending on who captures design. Arista can face spending cycles. Credo and Coherent can be volatile. AMD can remain behind. Palantir can be expensive. Nvidia can remain dominant, but future margin can be questioned if customers seek alternatives.

But the central point is that AI stops being an event and becomes consumption. When something becomes consumption, unit economics command. Cost per token, latency, availability, energy per response, utilization rate, margin per user. The language changes. And when language changes, the market changes winners or at least widens the board.

The investor who remains trapped in training may miss the next displacement. Training creates headlines. Inference creates recurring bills.

The recurring bill is where truth lives.

Leo Bentier

The New Frontier Is Sovereignty of Data, Models, and Infrastructure

A cold reading of the dispute over technological sovereignty: AI would stop being global and become geopolitical.

2025-03 deserved a cold essay because the dispute over technological sovereignty was not a calendar detail. It was a test of corporate sanity. Whenever a shock looks too external to be absorbed by management, it reveals an internal weakness that was already being ignored. The common executive turns surprise into an excuse; the rare executive turns surprise into architecture. The difference between the two is not raw intelligence. It is the willingness to place the company against reality before reality places the company against the wall.

AI would stop being global and become geopolitical That sentence should sit at the center of the article, because it forces the reader to abandon the comfort of simple causality. Companies are not defeated only by competitors, technologies, crises, or governments. They are defeated by slow decisions, scattered memory, disconnected metrics, crooked incentives, and the inability to convert small signals into coordinated action. The world changes first in narrow margins; then it appears in quarterly reports, when reaction has become expensive.

The intellectual signature here is not to confuse pessimism with precision. The point was not to root against companies, markets, or institutions. The point was to recognize that optimism without a mechanism is just emotional marketing. If I had been writing this archive in that month, I would have closed with the same demand: less opinion, more system; less narrative, more execution; less worship of growth, more respect for the invisible cost of operating badly for long enough.

Leo Bentier

SNDK: Artificial Intelligence Does Not Live in the Brain; It Lives in Memory

There are certain companies that reappear in the market like an old acquaintance nobody respects anymore. Everyone knows the name. Everyone thinks they already understand the story. Everyone carries a lazy mental memory: pen drives, memory cards, NAND, cycles, oversupply, crushed margins, commodity.

SanDisk returns to the market today as an independent company, separated from Western Digital, and this is precisely the kind of situation the educated investor tends to despise. Not because of analysis. Because of social reflex.

The market likes to appear sophisticated. It prefers buying the word "intelligence" to buying the infrastructure without which intelligence cannot breathe. It prefers the brain to the skull, the model to the disk, the GPU to the path through which data moves, rests, is replicated, compressed, deleted, recovered, and used again.

This is the aesthetic disease of modern capital: paying dearly for what looks new and ignoring what looks old, even when the old becomes indispensable to the new.

Artificial intelligence does not live in the brain. It lives in memory.

This will sound banal to engineers and ridiculous to portfolio managers with rented vocabulary. But money, when it wants to hide, often chooses the banal.

The thesis in SanDisk is not that NAND stopped being cyclical. That would be stupidity. A commodity does not become aristocracy because an investment bank changed the adjective in a report. NAND will remain cruel. It will remain an industry that punishes excess capacity, humiliates linear forecasts, and turns recent profits into traps for latecomers. There is nothing clean here.

But that is exactly the point.

The market may be looking at SanDisk as a storage company. I look at it as a cheap option on the persistence of artificial intelligence.

People talk about AI as if everything happened in the magical moment of computation. As if the model woke up, thought, and answered, without first being trained on oceans of data, without needing to move weights, store checkpoints, preserve embeddings, record logs, replicate states, keep data close to compute, recover from failures, and serve inference at scale.

That view is childish. An artificial brain without memory is expensive theater.

Data centers do not consume only GPUs. They consume power, cooling, networking, racks, controllers, memory, storage, and above all tolerance for failure. The vulgar investor buys the headline. The patient investor buys the dependency.

SanDisk does not need to be loved. It only needs to be necessary.

Today the market tends to see the separation from Western Digital as a corporate reorganization. A form line. A technical event. A distribution. But spin-offs exist because large structures hide things. Sometimes they hide garbage. Sometimes they hide optionality. The art is distinguishing the corpse from the hibernating animal.

Inside a conglomerate, the misunderstood asset is covered by the average. The average is an elegant form of ignorance. The good business subsidizes the bad, the bad contaminates the good, and the analyst, unable to separate the organism, applies a laziness discount to the whole thing. When the asset comes out alone, the market loses the excuse. It must look directly.

SanDisk will now be judged by itself.

That is dangerous for it. And interesting for us.

I am not looking here for a perfect company. Perfect companies are sold with no discount. I am looking for a company the market may still be classifying with old labels while future demand begins to change in nature.

The first layer of the thesis is simple: the market is obsessed with computation. The second is less popular: computation without persistence does not scale. The third is where the money lives: if demand for AI infrastructure tightens the market for NAND and enterprise SSDs, SanDisk may stop being seen as an old cyclical manufacturer and start being repriced as a supplier of a critical layer.

Notice the word: repriced.

I am not saying the company will become excellent. I am saying perception can change. And when perception changes in a newly separated, ignored, misclassified asset exposed to a demand the market has not yet named correctly, convexity can appear.

The common investor seeks certainty. The intelligent speculator seeks asymmetry.

SanDisk has enough problems to keep the tourists away. That is good. If the story were clean, the opportunity would already have been sterilized. There is cyclicality, oversupply risk, capex pressure, price dependence, the old brutality of semiconductors. None of this disappears because AI became a market religion.

But religions need temples. And temples need stone.

Nvidia sells the altar. SanDisk may sell part of the floor.

The most likely error of the next few quarters will be treating storage as peripheral. It is always like this at the beginning. The market identifies the most visible face of a new era and then discovers, late, that every era has invisible suppliers. In the internet, it was not only the browser. It was cables, routers, servers, data centers, disks, databases. In oil, it was not only the barrel. It was pipelines, valves, transportation, refining, services. In railroads, it was not only the locomotive. It was rails, ties, steel, land, concessions.

In artificial intelligence, it will not be only the GPU.

Capital likes narratives with a single hero. Reality operates in chains of dependency.

My concern is not whether SanDisk will be treated as a winner today. It will not. My concern is whether the market will be forced, perhaps in 2026, to admit that enterprise storage and flash memory are more central to AI than they seemed in February 2025.

When that admission happens, multiples move before certainties. The market does not wait for final proof. It sniffs tightness. It sniffs contracts. It sniffs price. It sniffs scarcity. It sniffs margin expansion. Then it invents a moral narrative to justify what price already did.

Price is always the first poet.

How would I position?

Not with romance. Not with a position too large. Not with that beginner's faith that turns a good thesis into personal ruin.

I would treat SanDisk as an asymmetric long position, not a marriage. I would buy common stock in an initial tranche, small enough to survive volatility and large enough to matter if the thesis works. The objective would not be to get the quarter right. It would be to capture a change in mental category: from "cyclical NAND" to "underestimated AI infrastructure."

If there were decent liquidity in long-dated options, I would consider long calls, out of the money only with restraint, or a call spread structure to limit cost and avoid paying volatility like a fool. The ideal trade is not the one that rises the most when you are right. It is the one that does not kill you when you are early.

There is a difference.

A stock can fall 30% and the thesis can remain alive. A badly chosen option can fall 80% because the calendar disliked you. Time is the tax on impatience.

I would prefer a combination: a core in shares, a small amount of convexity in long options, no idiotic leverage, no turning volatility into a mortal enemy. If the market keeps sleeping, shares allow waiting. If the market wakes violently, options capture the spasm. But size must be subordinated to ignorance. And here there is plenty of ignorance.

What would invalidate the thesis?

First, if data center demand for SSDs and NAND does not materialize in revenue and margin. It is not enough to say AI at a conference. Words do not pay debt. Second, if oversupply arrives too quickly, destroying price before the market recognizes tightness. Third, if SanDisk shows that, independent, it remains merely a mediocre operating piece without capital discipline. Fourth, if the AI narrative concentrates so heavily on DRAM, HBM, and GPUs that NAND remains viewed as a cheap accessory.

The blind spot would be confusing technical necessity with economic power. Not every necessary supplier captures value. Airlines are necessary and often terrible businesses. Hospitals need gloves, but that does not mean every glove manufacturer deserves a software multiple. The correct question is not: "Does AI need storage?" It does. The correct question is: "Can SanDisk capture enough price, margin, and contractual duration for the market to change the multiple?"

That is the question.

The thesis lives or dies there.

But I like the setup because it is ugly. The market can still laugh at it. "SanDisk? That one?" This laughter is an asset. When everyone already respects a thesis, future return has usually been confiscated by the respecters.

The best investment rarely begins with applause. It begins with discomfort.

There is also a psychological element: the market is saturated with obvious AI stories. Every manager wants to say he owns the brain. Few want to say they bought the cabinet where the brain stores its memories. This creates an aesthetic inefficiency. And aesthetic inefficiencies are real, because managers are human too, and humans want to look intelligent in front of other humans.

Buying GPUs looks sophisticated. Buying storage looks vulgar.

But vulgarity, bought at the right price, often beats sophistication bought late.

I am not interested in SanDisk because it looks futuristic. I am interested because it does not. The funniest part of technological revolutions is that many of them depend on things without glamour. Screws. Wires. Air conditioning. Power. Land. Memory. Persistence.

Artificial intelligence will be sold as magic. It will be operated as infrastructure.

And infrastructure has a different morality. It does not need to enchant. It needs to work.

If, over the next 12 to 18 months, the market begins to realize that AI expansion pressures not only GPUs and HBM, but also enterprise flash storage, NAND capacity, high-performance SSDs, and persistent data architectures, SanDisk may benefit from a double repricing: operational and narrative.

The operational would come from price, volume, and margin.

The narrative would come from the late embarrassment of analysts who classified the company as a relic on the day it returned to trade alone.

The market hates admitting that it saw a road and thought it was dust.

My position, therefore, would be simple: buy the forgetting. Sell only when the memory becomes consensus.

Because that is what is at stake. SanDisk does not need to prove it is Nvidia. It does not need to become a religion. It does not need to lead the liturgy. It only needs to show that behind every model promising to reason there is a huge machine forced to remember.

And if artificial intelligence needs to remember more than the market imagines, memory will stop being peripheral.

It will become destiny.

Leo Bentier

Companies Are Automating Before They Understand Their Own Processes

A cold reading of the corporate rush toward automation: Automated error scales better than manual correctness.

The comfortable error would have been to call the corporate rush toward automation an exception. Exception is a word used by people who do not want to revise their model of the world. In 2025-02, the important signal was not the public spectacle but the hidden mechanism: bad incentives, borrowed confidence, dependence on outsiders, and weak correction loops. The average manager hunts for culprits after impact; the serious operator asks which rules allowed the impact to remain invisible for so long.

The thesis that mattered was brutal: automated error scales better than manual correctness That changes the whole conversation. If a company depends on friendly macro conditions, cheap capital, disciplined suppliers, patient customers, or heroic employees, it does not have a robust operation. It has a temporarily financed fiction. Most management fails because it tries to look sophisticated before it becomes true, and truth almost always begins with an unpleasant question about fragility, not with a beautiful deck.

I would have written it as a warning, not as mystical prediction. Prediction is vanity when it does not produce a rule of action. What matters is deciding before the crowd discovers the vocabulary. In every cycle, the same kinds of people, under the same conditions, produce the same outcomes because their internal systems do not change. The opportunity was to build a discipline: observe the signal, reduce ambiguity, choose the action, measure the deviation, and correct without romance.

Leo Bentier

Trump's second inauguration is less return than revenge

The populist who returns does not return as candidate; he returns as creditor.

On January 20, 2025, the event looked like news. A serious letter would have treated it as a signal. News tells you what happened; a signal reveals what may repeat later with another name, another face, and another flag. Politics is a machine for recycling old passions with new vocabulary. Whoever reads only the event becomes hostage to the calendar. Whoever reads the incentive begins to see the future while it is still disguised as exception.

Trump's second mandate begins not as restoration, but as collection.

The argument would not be born from party preference, but from an older question: what kind of regime, elite, and people does this event reveal? Small politics asks who won today. Good politics asks which vice was rewarded, which virtue was punished, and which cost was pushed onto the next generation. This is where almost everyone fails. The commentator looks for convenient culprits; the statesman looks for mechanisms. The cheap moralist points the finger; the serious moralist asks why that finger found so many accomplices.

Brazil and the modern world suffer from the same disease with different accents: the confusion between form and substance. We have elections, but not always trust. Courts, but not always perceived justice. Markets, but not always responsibility. State, but not always common good. Communication, but not always truth. Data, but not always wisdom. Modernity multiplied instruments and impoverished judgment. The result is a civilization technically equipped to err at speed.

What could be seen on this date was less an isolated fact than a movement of plates. The surface would still allow calm speeches. It always does. Before ruptures, ministers guarantee normality, analysts calculate small probabilities, newspapers choose moderate adjectives, and investors pretend political risk fits in a spreadsheet. But history does not move only when there is an explosion. It moves when important men begin to lie differently. A change in the lie often precedes a change in the regime.

The letter would avoid the vice of the cheap prophet. The cheap prophet wants to look like owner of the future. The serious observer knows only that certain arrangements are fragile, certain incentives rot institutions, and certain accumulated humiliations look for a date. It is not necessary to predict the exact lightning strike when one knows the forest is dry. The vanity of guessing the day is inferior to the discipline of recognizing flammable wood.

There would also be a warning to the Brazilian reader. Brazil likes importing foreign explanations without importing the structures that make them true. It speaks of liberalism without impersonality, strong state without efficiency, sovereignty without strategy, democracy without civilized losers, market without competition, conservatism without self-control, progressivism without fiscal responsibility. That is why the country spends decades debating the vocabulary of great politics while preserving the habits of the small court.

Cicero would have seen in this episode a question about the republic: does law serve the city or the factions? Augustine would ask whether power, separated from justice, is truly government or only a gang with protocol. Aquinas would remind us that the common good is not the sum of private appetites, but a moral order that allows the city to seek higher ends. Marcus Aurelius would distrust applause, panic, and indignation, for all are poor emotional states for government. Churchill would look for will in the event, not excuse. Plato would ask what kind of collective soul chose this moment.

The forecast, then, would be simple and hard. The event recorded here would not end in itself. It would create habit, precedent, resentment, architecture, or language. The first error of observers would be to treat it as a period. The second would be to judge it only by the immediate winner. The third would be to believe the institution touched by the event would emerge unchanged. Institutions are not buildings; they are memories of behavior. When behavior changes, the building may remain standing while the republic has already changed skin.

Tensions with courts, tariffs, trade disputes, and renewed pressure on allies would confirm that revenge would become a method of government.

In retrospect, the essential point would be this: almost every major political event first appears as anomaly, then as repetition, then as doctrine. The mistake is laughing at the anomaly. The opportunity is asking why it appeared now, why it found an audience, and who gains from its normalization.

There is a difference between good politics and cheering. Cheering needs innocents and monsters. Good politics needs causes, limits, and consequences. Cheering wants to absolve its own and condemn the others. Good politics asks which sin is common to both sides. Cheering calls prudence cowardice when it is winning and persecution when it is losing. Good politics knows that every faction, when it believes itself redemptive, begins rehearsing small tyrannies.

This is why the letter would not ask the reader for comfort. It would ask for maturity. The world is not governed by declared intentions, but by tolerated incentives. An elite can speak of democracy while despising the people. A people can speak of liberty while desiring revenge. A market can speak of efficiency while asking for public rescue. A state can speak of justice while expanding its own appetite. A power can speak of order while surrounding rivals. A rival can speak of security while preparing aggression.

The reader who wants to understand the future should abandon the superstition of proper names. Names matter, but less than mechanisms. The man passes; the incentive remains. The party changes; the method persists. The crisis ends; the learned fragility moves elsewhere. The modern world is a school of disguises. Whoever clings to the disguise always arrives late.

A good letter does not end by asking for agreement. It ends by leaving useful discomfort.

The populist who returns does not return as candidate; he returns as creditor.

Leo Bentier

Who Captures Productivity and Who Merely Pays the Bill?

A cold reading of the demand for AI ROI in companies: AI would separate operators from software consumers.

Most executives would read this signal as news. That is the first mistake. News is what arrives after the market has found a comfortable word for the change; a signal is what appears before that, crooked, incomplete, and badly priced. In 2025-01, the demand for AI ROI in companies already pointed to a structural shift, not an isolated episode. The point was not to guess the next headline. The point was to see that the system was beginning to punish companies without cash, operational memory, decision discipline, or an honest relationship with the cost of their own growth.

The correct reading was less theatrical and more severe: aI would separate operators from software consumers Anyone who understood this did not need to pose as a prophet. He only needed to reject the managerial superstition that good outcomes prove good processes. Many companies grow because the wind helps them, not because they know how to sail. When the wind turns, you discover who had a system and who had only busy people, clean spreadsheets, long meetings, and a private museum of opinions sold internally as strategy.

That is the kind of reading I would have recorded without asking consensus for permission. Businesses do not break on the day the market notices; they break when the organization loses the ability to turn context into decision, decision into execution, and execution into correction. The event of 2025-01 would have been used as a lens, not as a historical fetish. The lesson was simple, therefore almost always ignored: the company that does not build a system for deciding will be decided by the environment.

Leo Bentier

2025 Will Be the Year of Selection: Real AI Versus Decorative AI

A cold reading of the end of 2024: The market would begin demanding ROI.

2024-12 deserved a cold essay because the end of 2024 was not a calendar detail. It was a test of corporate sanity. Whenever a shock looks too external to be absorbed by management, it reveals an internal weakness that was already being ignored. The common executive turns surprise into an excuse; the rare executive turns surprise into architecture. The difference between the two is not raw intelligence. It is the willingness to place the company against reality before reality places the company against the wall.

The market would begin demanding ROI That sentence should sit at the center of the article, because it forces the reader to abandon the comfort of simple causality. Companies are not defeated only by competitors, technologies, crises, or governments. They are defeated by slow decisions, scattered memory, disconnected metrics, crooked incentives, and the inability to convert small signals into coordinated action. The world changes first in narrow margins; then it appears in quarterly reports, when reaction has become expensive.

The intellectual signature here is not to confuse pessimism with precision. The point was not to root against companies, markets, or institutions. The point was to recognize that optimism without a mechanism is just emotional marketing. If I had been writing this archive in that month, I would have closed with the same demand: less opinion, more system; less narrative, more execution; less worship of growth, more respect for the invisible cost of operating badly for long enough.

Leo Bentier

Trump returned because the elite confused prosecution with exorcism

The man the elite calls the end of democracy can return when the voter thinks democracy became the elite's club.

On November 6, 2024, the event looked like news. A serious letter would have treated it as a signal. News tells you what happened; a signal reveals what may repeat later with another name, another face, and another flag. Politics is a machine for recycling old passions with new vocabulary. Whoever reads only the event becomes hostage to the calendar. Whoever reads the incentive begins to see the future while it is still disguised as exception.

Trump's return showed that legal process does not exorcise a political force when voters see the system itself as captured.

The argument would not be born from party preference, but from an older question: what kind of regime, elite, and people does this event reveal? Small politics asks who won today. Good politics asks which vice was rewarded, which virtue was punished, and which cost was pushed onto the next generation. This is where almost everyone fails. The commentator looks for convenient culprits; the statesman looks for mechanisms. The cheap moralist points the finger; the serious moralist asks why that finger found so many accomplices.

Brazil and the modern world suffer from the same disease with different accents: the confusion between form and substance. We have elections, but not always trust. Courts, but not always perceived justice. Markets, but not always responsibility. State, but not always common good. Communication, but not always truth. Data, but not always wisdom. Modernity multiplied instruments and impoverished judgment. The result is a civilization technically equipped to err at speed.

What could be seen on this date was less an isolated fact than a movement of plates. The surface would still allow calm speeches. It always does. Before ruptures, ministers guarantee normality, analysts calculate small probabilities, newspapers choose moderate adjectives, and investors pretend political risk fits in a spreadsheet. But history does not move only when there is an explosion. It moves when important men begin to lie differently. A change in the lie often precedes a change in the regime.

The letter would avoid the vice of the cheap prophet. The cheap prophet wants to look like owner of the future. The serious observer knows only that certain arrangements are fragile, certain incentives rot institutions, and certain accumulated humiliations look for a date. It is not necessary to predict the exact lightning strike when one knows the forest is dry. The vanity of guessing the day is inferior to the discipline of recognizing flammable wood.

There would also be a warning to the Brazilian reader. Brazil likes importing foreign explanations without importing the structures that make them true. It speaks of liberalism without impersonality, strong state without efficiency, sovereignty without strategy, democracy without civilized losers, market without competition, conservatism without self-control, progressivism without fiscal responsibility. That is why the country spends decades debating the vocabulary of great politics while preserving the habits of the small court.

Cicero would have seen in this episode a question about the republic: does law serve the city or the factions? Augustine would ask whether power, separated from justice, is truly government or only a gang with protocol. Aquinas would remind us that the common good is not the sum of private appetites, but a moral order that allows the city to seek higher ends. Marcus Aurelius would distrust applause, panic, and indignation, for all are poor emotional states for government. Churchill would look for will in the event, not excuse. Plato would ask what kind of collective soul chose this moment.

The forecast, then, would be simple and hard. The event recorded here would not end in itself. It would create habit, precedent, resentment, architecture, or language. The first error of observers would be to treat it as a period. The second would be to judge it only by the immediate winner. The third would be to believe the institution touched by the event would emerge unchanged. Institutions are not buildings; they are memories of behavior. When behavior changes, the building may remain standing while the republic has already changed skin.

His return would show that inflation, border, identity, resentment, and institutional distrust weighed more than liberal catechism.

In retrospect, the essential point would be this: almost every major political event first appears as anomaly, then as repetition, then as doctrine. The mistake is laughing at the anomaly. The opportunity is asking why it appeared now, why it found an audience, and who gains from its normalization.

There is a difference between good politics and cheering. Cheering needs innocents and monsters. Good politics needs causes, limits, and consequences. Cheering wants to absolve its own and condemn the others. Good politics asks which sin is common to both sides. Cheering calls prudence cowardice when it is winning and persecution when it is losing. Good politics knows that every faction, when it believes itself redemptive, begins rehearsing small tyrannies.

This is why the letter would not ask the reader for comfort. It would ask for maturity. The world is not governed by declared intentions, but by tolerated incentives. An elite can speak of democracy while despising the people. A people can speak of liberty while desiring revenge. A market can speak of efficiency while asking for public rescue. A state can speak of justice while expanding its own appetite. A power can speak of order while surrounding rivals. A rival can speak of security while preparing aggression.

The reader who wants to understand the future should abandon the superstition of proper names. Names matter, but less than mechanisms. The man passes; the incentive remains. The party changes; the method persists. The crisis ends; the learned fragility moves elsewhere. The modern world is a school of disguises. Whoever clings to the disguise always arrives late.

A good letter does not end by asking for agreement. It ends by leaving useful discomfort.

The man the elite calls the end of democracy can return when the voter thinks democracy became the elite's club.

Leo Bentier

The Next Corporate Software Will Not Be a Dashboard. It Will Be an Operator

A cold reading of the evolution from copilots to agents: Data must become action.

The comfortable error would have been to call the evolution from copilots to agents an exception. Exception is a word used by people who do not want to revise their model of the world. In 2024-11, the important signal was not the public spectacle but the hidden mechanism: bad incentives, borrowed confidence, dependence on outsiders, and weak correction loops. The average manager hunts for culprits after impact; the serious operator asks which rules allowed the impact to remain invisible for so long.

The thesis that mattered was brutal: data must become action That changes the whole conversation. If a company depends on friendly macro conditions, cheap capital, disciplined suppliers, patient customers, or heroic employees, it does not have a robust operation. It has a temporarily financed fiction. Most management fails because it tries to look sophisticated before it becomes true, and truth almost always begins with an unpleasant question about fragility, not with a beautiful deck.

I would have written it as a warning, not as mystical prediction. Prediction is vanity when it does not produce a rule of action. What matters is deciding before the crowd discovers the vocabulary. In every cycle, the same kinds of people, under the same conditions, produce the same outcomes because their internal systems do not change. The opportunity was to build a discipline: observe the signal, reduce ambiguity, choose the action, measure the deviation, and correct without romance.

Leo Bentier

The municipality is where ideology discovers the pothole

Every grand political narrative dies a little before sewage, daycare, and asphalt.

On October 27, 2024, the event looked like news. A serious letter would have treated it as a signal. News tells you what happened; a signal reveals what may repeat later with another name, another face, and another flag. Politics is a machine for recycling old passions with new vocabulary. Whoever reads only the event becomes hostage to the calendar. Whoever reads the incentive begins to see the future while it is still disguised as exception.

Brazil's municipal elections reminded national politics that power is territorial before it is rhetorical.

The argument would not be born from party preference, but from an older question: what kind of regime, elite, and people does this event reveal? Small politics asks who won today. Good politics asks which vice was rewarded, which virtue was punished, and which cost was pushed onto the next generation. This is where almost everyone fails. The commentator looks for convenient culprits; the statesman looks for mechanisms. The cheap moralist points the finger; the serious moralist asks why that finger found so many accomplices.

Brazil and the modern world suffer from the same disease with different accents: the confusion between form and substance. We have elections, but not always trust. Courts, but not always perceived justice. Markets, but not always responsibility. State, but not always common good. Communication, but not always truth. Data, but not always wisdom. Modernity multiplied instruments and impoverished judgment. The result is a civilization technically equipped to err at speed.

What could be seen on this date was less an isolated fact than a movement of plates. The surface would still allow calm speeches. It always does. Before ruptures, ministers guarantee normality, analysts calculate small probabilities, newspapers choose moderate adjectives, and investors pretend political risk fits in a spreadsheet. But history does not move only when there is an explosion. It moves when important men begin to lie differently. A change in the lie often precedes a change in the regime.

The letter would avoid the vice of the cheap prophet. The cheap prophet wants to look like owner of the future. The serious observer knows only that certain arrangements are fragile, certain incentives rot institutions, and certain accumulated humiliations look for a date. It is not necessary to predict the exact lightning strike when one knows the forest is dry. The vanity of guessing the day is inferior to the discipline of recognizing flammable wood.

There would also be a warning to the Brazilian reader. Brazil likes importing foreign explanations without importing the structures that make them true. It speaks of liberalism without impersonality, strong state without efficiency, sovereignty without strategy, democracy without civilized losers, market without competition, conservatism without self-control, progressivism without fiscal responsibility. That is why the country spends decades debating the vocabulary of great politics while preserving the habits of the small court.

Cicero would have seen in this episode a question about the republic: does law serve the city or the factions? Augustine would ask whether power, separated from justice, is truly government or only a gang with protocol. Aquinas would remind us that the common good is not the sum of private appetites, but a moral order that allows the city to seek higher ends. Marcus Aurelius would distrust applause, panic, and indignation, for all are poor emotional states for government. Churchill would look for will in the event, not excuse. Plato would ask what kind of collective soul chose this moment.

The forecast, then, would be simple and hard. The event recorded here would not end in itself. It would create habit, precedent, resentment, architecture, or language. The first error of observers would be to treat it as a period. The second would be to judge it only by the immediate winner. The third would be to believe the institution touched by the event would emerge unchanged. Institutions are not buildings; they are memories of behavior. When behavior changes, the building may remain standing while the republic has already changed skin.

The 2026 presidential dispute would depend less on abstract slogans and more on local reach, mayors, machines, and territorial alliances.

In retrospect, the essential point would be this: almost every major political event first appears as anomaly, then as repetition, then as doctrine. The mistake is laughing at the anomaly. The opportunity is asking why it appeared now, why it found an audience, and who gains from its normalization.

There is a difference between good politics and cheering. Cheering needs innocents and monsters. Good politics needs causes, limits, and consequences. Cheering wants to absolve its own and condemn the others. Good politics asks which sin is common to both sides. Cheering calls prudence cowardice when it is winning and persecution when it is losing. Good politics knows that every faction, when it believes itself redemptive, begins rehearsing small tyrannies.

This is why the letter would not ask the reader for comfort. It would ask for maturity. The world is not governed by declared intentions, but by tolerated incentives. An elite can speak of democracy while despising the people. A people can speak of liberty while desiring revenge. A market can speak of efficiency while asking for public rescue. A state can speak of justice while expanding its own appetite. A power can speak of order while surrounding rivals. A rival can speak of security while preparing aggression.

The reader who wants to understand the future should abandon the superstition of proper names. Names matter, but less than mechanisms. The man passes; the incentive remains. The party changes; the method persists. The crisis ends; the learned fragility moves elsewhere. The modern world is a school of disguises. Whoever clings to the disguise always arrives late.

A good letter does not end by asking for agreement. It ends by leaving useful discomfort.

Every grand political narrative dies a little before sewage, daycare, and asphalt.

Leo Bentier

Founder Mode is not a management style. It's the refusal to become irrelevant.

There is an essay that distinguishes founder mode from manager mode. But founder mode is not a style you choose; it is the refusal to delegate yourself into irrelevance. The founder who delegates everything and stays only at the high level becomes a decorative figure who does not know their own company. Founder mode is refusing that irrelevance.

There is an essay that distinguishes founder mode from manager mode — two ways of leading a company. But there is a deeper reading of what founder mode means. Founder mode is not a management style chosen among others. It is the refusal to become irrelevant. The founder who delegates everything and stays only at the high level becomes a decorative figure who does not know their own company. Founder mode is the refusal of that irrelevance — the refusal to delegate yourself into ceasing to matter.

Start with the trap of total delegation. The conventional wisdom of management tells the leader to delegate — step away from the details, stay at the high level, let others operate. Taken to the extreme, that total delegation transforms the leader into a figure who no longer knows the details of their own company — who delegated everything and was left only with the high level, without the knowledge of what actually happens. Total delegation has a trap: it removes the leader from the details that matter, transforming them into a decorative figure who presides without knowing. The leader who delegates everything becomes irrelevant to the operation of their own company, because they no longer know what it actually does.

Here is why that irrelevance is the danger. A leader who does not know the details of their own company cannot decide well about it — their decisions are based on summaries, not on real knowledge; they preside, but do not understand. That leader becomes decorative: they have the title, but not the relevance, because they delegated the knowledge that would make them relevant. The irrelevance is the danger of total delegation: the leader becomes a figure who does not matter to the operation, because they stepped away from the details that matter. And a company led by a decorative figure, who does not know its details, loses the direction of whoever should understand it deeply. The irrelevance of the leader is a real cost to the company.

Here is what founder mode actually is. Founder mode is not a management style among others; it is the refusal of that irrelevance. It is the founder refusing to delegate themselves into ceasing to matter — staying in the details that matter, knowing their own company deeply, remaining relevant to its operation. Founder mode is not choosing a style; it is refusing the trap of total delegation, refusing to become a decorative figure who does not know the company. The founder in founder mode stays in the details that matter, not for a style, but for a refusal — the refusal to become irrelevant to the very company they founded.

Notice the connection to the decision, judgment and the operator versus the figurehead we had been pulling. I pointed out that the decision requires knowing the details, that judgment is based on real knowledge, that operating is different from presiding. Founder mode as the refusal of irrelevance is the application of those threads to leadership: the founder staying in the details that matter to be able to decide and judge with real knowledge, refusing to become the decorative figure who presides without knowing. What we had been seeing about the knowledge of the details being the base of the decision applies to the founder: founder mode is keeping that knowledge, refusing the total delegation that would remove them from the details and make them irrelevant.

See what distinguishes healthy delegation from delegation into irrelevance. Founder mode is not not delegating — delegating is necessary, and a founder who delegates nothing does not scale. It is not delegating yourself into irrelevance — not stepping away from the details that matter to the point of no longer knowing your own company. The distinction is between delegating the execution (healthy) and delegating the knowledge and judgment that make the leader relevant (the trap). Founder mode delegates the execution but keeps the knowledge of the details that matter, refusing the irrelevance. Healthy delegation scales the operation; delegation into irrelevance empties the leader. Founder mode is delegating without delegating yourself into ceasing to matter.

It is fair, in balance, to recognize that founder mode has risks and that delegation has real virtues. Founder mode, taken too far, becomes micromanagement — the founder in the details that do not matter, suffocating the operation, not scaling. And delegation has real virtues — it scales, empowers, frees the leader for what matters. The point is not that the founder should be in every detail or that delegation is bad, but that founder mode is the refusal to delegate yourself into irrelevance — staying in the details that matter, not in all of them. Maturity is distinguishing the details that matter (where the founder should remain) from those that do not (where they should delegate) — making founder mode the refusal of irrelevance, not the refusal to delegate, and avoiding both the irrelevance of total delegation and the micromanagement of being in everything.

For the investor and the manager, this suggests recognizing founder mode as the refusal of irrelevance, and evaluating whether the leader knows the details that matter. The question about a leader is not 'do they delegate well?', but 'do they know the details that matter of their own company, or did they delegate themselves into irrelevance?'. The leaders who stay in the details that matter remain relevant, deciding with real knowledge; those who delegated everything became decorative figures who do not know the company. Whoever evaluates leadership by the delegation confuses delegating with leading; whoever recognizes founder mode as the refusal of irrelevance sees whether the leader keeps the knowledge of the details that matter — the base of deciding and judging with relevance, which total delegation would empty.

The rule of this moment: founder mode is not a management style you choose, but the refusal to delegate yourself into irrelevance — to stay in the details that matter, instead of becoming a decorative figure who presides without knowing their own company. Whoever delegates everything into irrelevance becomes a decorative leader; whoever refuses that irrelevance, keeping the knowledge of the details that matter, remains relevant to the company they lead, deciding and judging with real knowledge.

Founder Mode is not a management style. It is the refusal to become irrelevant. The founder who delegates everything and stays only at the high level becomes a decorative figure who does not know their own company — founder mode is refusing that irrelevance, staying in the details that matter. Mark founder mode not as a management style among others, but as the refusal of irrelevance — the demonstration that total delegation transforms the leader into a decorative figure who does not know the details of their own company, that founder mode is the refusal to delegate yourself into ceasing to matter, and that the relevant founder is neither the one who delegates everything nor the one who micromanages, but the one who stays in the details that matter — refusing the irrelevance total delegation brings, and preserving the knowledge that makes their decision and their judgment relevant to the company they founded.

Leo Bentier

The American Election Will Also Be a Test of Synthetic Media

A cold reading of the 2024 American election cycle: Public trust became vulnerable infrastructure.

Most executives would read this signal as news. That is the first mistake. News is what arrives after the market has found a comfortable word for the change; a signal is what appears before that, crooked, incomplete, and badly priced. In 2024-10, the 2024 American election cycle already pointed to a structural shift, not an isolated episode. The point was not to guess the next headline. The point was to see that the system was beginning to punish companies without cash, operational memory, decision discipline, or an honest relationship with the cost of their own growth.

The correct reading was less theatrical and more severe: public trust became vulnerable infrastructure Anyone who understood this did not need to pose as a prophet. He only needed to reject the managerial superstition that good outcomes prove good processes. Many companies grow because the wind helps them, not because they know how to sail. When the wind turns, you discover who had a system and who had only busy people, clean spreadsheets, long meetings, and a private museum of opinions sold internally as strategy.

That is the kind of reading I would have recorded without asking consensus for permission. Businesses do not break on the day the market notices; they break when the organization loses the ability to turn context into decision, decision into execution, and execution into correction. The event of 2024-10 would have been used as a lens, not as a historical fetish. The lesson was simple, therefore almost always ignored: the company that does not build a system for deciding will be decided by the environment.

Leo Bentier

AI Agents Will Be Dangerous if Companies Lack Clear Processes

A cold reading of the popularization of AI agents: Automation accelerates both correctness and error.

2024-09 deserved a cold essay because the popularization of AI agents was not a calendar detail. It was a test of corporate sanity. Whenever a shock looks too external to be absorbed by management, it reveals an internal weakness that was already being ignored. The common executive turns surprise into an excuse; the rare executive turns surprise into architecture. The difference between the two is not raw intelligence. It is the willingness to place the company against reality before reality places the company against the wall.

Automation accelerates both correctness and error That sentence should sit at the center of the article, because it forces the reader to abandon the comfort of simple causality. Companies are not defeated only by competitors, technologies, crises, or governments. They are defeated by slow decisions, scattered memory, disconnected metrics, crooked incentives, and the inability to convert small signals into coordinated action. The world changes first in narrow margins; then it appears in quarterly reports, when reaction has become expensive.

The intellectual signature here is not to confuse pessimism with precision. The point was not to root against companies, markets, or institutions. The point was to recognize that optimism without a mechanism is just emotional marketing. If I had been writing this archive in that month, I would have closed with the same demand: less opinion, more system; less narrative, more execution; less worship of growth, more respect for the invisible cost of operating badly for long enough.

Leo Bentier

The Bubble Is Not AI. The Bubble Is Thinking Every Company Will Capture Its Value

A cold reading of the debate over an AI bubble: A general technology does not distribute returns equally.

The comfortable error would have been to call the debate over an AI bubble an exception. Exception is a word used by people who do not want to revise their model of the world. In 2024-08, the important signal was not the public spectacle but the hidden mechanism: bad incentives, borrowed confidence, dependence on outsiders, and weak correction loops. The average manager hunts for culprits after impact; the serious operator asks which rules allowed the impact to remain invisible for so long.

The thesis that mattered was brutal: a general technology does not distribute returns equally That changes the whole conversation. If a company depends on friendly macro conditions, cheap capital, disciplined suppliers, patient customers, or heroic employees, it does not have a robust operation. It has a temporarily financed fiction. Most management fails because it tries to look sophisticated before it becomes true, and truth almost always begins with an unpleasant question about fragility, not with a beautiful deck.

I would have written it as a warning, not as mystical prediction. Prediction is vanity when it does not produce a rule of action. What matters is deciding before the crowd discovers the vocabulary. In every cycle, the same kinds of people, under the same conditions, produce the same outcomes because their internal systems do not change. The opportunity was to build a discipline: observe the signal, reduce ambiguity, choose the action, measure the deviation, and correct without romance.

Leo Bentier

The Published AI Act Creates the First Major Regulatory Grammar for AI

A cold reading of the publication of the AI Act in the Official Journal of the European Union: Europe would try to set global rules by default.

Most executives would read this signal as news. That is the first mistake. News is what arrives after the market has found a comfortable word for the change; a signal is what appears before that, crooked, incomplete, and badly priced. In 2024-07, the publication of the AI Act in the Official Journal of the European Union already pointed to a structural shift, not an isolated episode. The point was not to guess the next headline. The point was to see that the system was beginning to punish companies without cash, operational memory, decision discipline, or an honest relationship with the cost of their own growth.

The correct reading was less theatrical and more severe: europe would try to set global rules by default Anyone who understood this did not need to pose as a prophet. He only needed to reject the managerial superstition that good outcomes prove good processes. Many companies grow because the wind helps them, not because they know how to sail. When the wind turns, you discover who had a system and who had only busy people, clean spreadsheets, long meetings, and a private museum of opinions sold internally as strategy.

That is the kind of reading I would have recorded without asking consensus for permission. Businesses do not break on the day the market notices; they break when the organization loses the ability to turn context into decision, decision into execution, and execution into correction. The event of 2024-07 would have been used as a lens, not as a historical fetish. The lesson was simple, therefore almost always ignored: the company that does not build a system for deciding will be decided by the environment.

Leo Bentier