The Pix Principle: Why Artificial Intelligence Will Become Invisible

In November 2020, Brazil introduced a new payment system called Pix.

It was not a new credit card. It was not a new bank. It was not another financial app competing for space on the home screen. Pix was public payment infrastructure that connected the banks and payment institutions people already used.

Less than six years later, Pix is the dominant payment method in Brazil. A Central Bank survey published in December 2024 found that 76.4 percent of Brazilians used it. By July 2026, roughly 167 million individuals had registered a Pix key — about 186 million accounts counting businesses — and Pix accounted for more than half of all transactions in the country by volume. Brazilians now use it for everything from buying food from a street vendor to purchasing clothing, paying suppliers, shopping online, and buying cars.

That is an extraordinarily short adoption window.

A Brazilian walking down the street today might be surprised to learn how recently Pix arrived. It no longer feels like a new technology. It feels like part of the environment.

That is what is about to happen with artificial intelligence.

Pix succeeded because its fundamentals were better.

For individuals it is free by regulation — the Central Bank prohibits participating institutions from charging natural persons to send or receive. For businesses, its cost is typically lower than other electronic payment methods, and the fees institutions may charge them are capped. Transactions operate around the clock, including weekends and holidays, and the system is specified to settle in ten seconds or less. The seller does not wait several business days for funds.

These are not cosmetic advantages. They improve the underlying economics of a transaction.

The buyer gets simplicity and speed. The seller gets faster access to money and lower acceptance costs. Small merchants can receive digital payments without purchasing elaborate payment infrastructure. A person can send money using a phone number, an email address, a tax identification number, or a randomly generated key rather than entering a long set of banking details — and can initiate the payment by scanning a QR code instead of typing anything at all.

Pix also avoided one of the most common obstacles to adoption: it did not require people to establish a separate digital relationship with Pix.

There is no central Pix app that every Brazilian must download and learn. People access Pix through their existing bank, fintech, digital wallet, or payment institution. Pix appears inside the financial environment they already use.

That distinction is enormously important.

Pix is not primarily a destination. It is a capability.

You do not go to Pix because you want to shop. You go to the merchant. Pix is simply available when it is time to pay. The consumer experiences the benefit without having to think deeply about the infrastructure producing it.

Pix did not enter an empty market.

It entered a global financial system containing some of the largest and most established companies in the world. Every transaction moved onto Pix represented a transaction that might otherwise have passed through a card network, a bank transfer system, or another payment intermediary. Visa and Mastercard have both flagged instant-payment networks of this kind as a competitive risk in their securities filings.

The economic resistance was therefore predictable.

In July 2025, the Trump administration initiated a Section 301 investigation into several Brazilian policies, including digital trade and electronic payment services. On June 1, 2026, the Office of the United States Trade Representative determined that those practices were actionable, finding that Brazil had unfairly disadvantaged American electronic-payment companies through policies favoring what it called its “national champion Pix.” The determination objected specifically to the Central Bank’s dual role as both regulator and operator, to the mandate that large institutions participate, to the requirement that Pix appear prominently in their apps, and to the free-for-individuals rule and the caps on business fees.

On July 15, 2026, USTR took final action: a 25 percent tariff on substantially all Brazilian goods, effective July 22, with carve-outs for beef, orange juice, energy, and civil aircraft. This is a separate track from the 50 percent tariff imposed on Brazil under emergency economic powers in July 2025, which the Supreme Court invalidated in February 2026.

One can debate whether the structure of Pix is fair, whether a central bank should operate infrastructure it also regulates, or whether Brazil gave its domestic system preferential treatment. Those are legitimate policy questions, and the administration has been careful to say it is not asking Brazil to get rid of Pix.

But they do not alter the more important observation.

Pix has already become useful enough that it cannot simply be argued out of existence.

The political power of the United States is considerable. The economic interests attached to global card networks are considerable. The institutional power of incumbent banks is considerable.

Yet there is Pix.

The system survives because millions of people and businesses experience its advantages every day. Political pressure can affect its regulation, pricing, governance, and international expansion. It is much harder to persuade an entire population to abandon something that is faster, cheaper, and already woven into ordinary life. Brazil’s central bank governor, Gabriel Galípolo, has pointed out that card usage rose in absolute terms because Pix brought more than seventy million people into the financial system in the first place — complaining about it, he said, would be “kind of like saying that creating basic sanitation hurt the revenues of those who own water trucks.”

Fundamentals eventually become their own form of political power.

Artificial intelligence is following a similar trajectory.

Artificial intelligence itself is not new, but the consumer generative-AI era has a useful symbolic starting point: November 30, 2022, when OpenAI publicly introduced ChatGPT.

That means that in August 2026, the generative-AI era is not yet four years old.

We are still early enough that people think of AI as something they consciously visit. They open ChatGPT, Claude, Gemini, or another AI product. They type a prompt. They wait for a response. They copy the result into another application.

That is the equivalent of imagining that Pix would remain a separate payment destination.

It will not.

The enduring form of AI is not the chatbot. It is not the prompt box. It is not even the AI app. The enduring form of AI is generation — the production of something that did not previously exist.

• TravelThe system generates the receipt and the expense report.
• CommerceThe storefront generates product descriptions, merchandising images, and customer responses.
• WebThe site generates and tests its own new landing pages.
• EmailThe inbox generates the replies.
• MusicThe environment generates a playlist for a particular person, activity, and moment.
• Everything elseA personal operating system generates a book, a presentation, a research report, a schedule, a budget, a dashboard, a photograph, a lesson plan, or a piece of software.

The user does not necessarily open an AI application to request each artifact. The generation occurs inside the environment where the artifact is needed.

There may be no visible prompt. There may be no explicit instruction. There may be no moment when the person says, “I am now using artificial intelligence.”

AI will simply be there.

This is the central mistake many companies are making.

They are treating AI as a new category of software application. They believe success means convincing users to download their AI app, create another account, learn another interface, and consciously return whenever they want intelligence.

That may describe the transitional period. It does not describe the destination.

Pix became ubiquitous partly because it did not ask Brazilians to organize their financial lives around Pix. It entered the institutions and commercial experiences they already used. AI will do the same thing.

You will not necessarily go to an AI app to create a website. You will tell your personal system that you need a website, and the website will appear.

You will not open an AI music application and carefully construct a prompt describing your mood. Your environment will understand the activity, the people present, the time available, your previous choices, and the atmosphere you prefer. The playlist will already be playing.

You will not ask an AI to reply to each routine email. The replies will already exist. Your attention will be requested only when the system encounters something surprising, consequential, or genuinely ambiguous.

You will not ask AI to produce every receipt, report, illustration, product description, forecast, or schedule. Those artifacts will be generated as natural consequences of the work being performed.

The intelligence will remain present while the experience of “using AI” disappears.

Artificial intelligence will encounter enormous resistance.

Software companies will attempt to preserve the application as the primary unit of computing. Employers will attempt to confine workers to approved corporate systems. Professional organizations will defend existing definitions of expertise. Governments will regulate where and how generation may occur. Copyright owners, technology platforms, universities, labor groups, publishers, and established service providers will all attempt to shape the transition.

Some of those interventions will be necessary. Some will protect legitimate rights and reduce real dangers. Others will primarily protect incumbent economic arrangements.

But none of them changes the fundamentals.

A person can describe what they need in ordinary language and receive the finished artifact.

That is the fundamental advantage. The artifact might be a website, a book, a contract summary, a software tool, a marketing campaign, a research report, a lesson, a video, a financial model, a business plan, a medical explanation, or a response to an email.

Previously, producing that artifact required the person to locate an appropriate application, learn its interface, understand its commands, assemble the necessary inputs, and perform the work according to the application’s predetermined structure.

Generation reverses that relationship. The person describes the desired reality. The system constructs it.

That reduction in cost, time, friction, and required expertise is too fundamental to remain confined inside a collection of AI applications.

Just as Pix compresses the distance between the decision to pay and the arrival of the money, AI compresses the distance between an intention and an artifact.

That is why it will spread.

Pix began full operation on November 16, 2020. Four years later, Brazil’s central bank reported that the system had processed roughly 63.5 billion transactions during 2024 alone, moving approximately 26.5 trillion Brazilian reais. In the first five months of 2026 it moved another 36 billion transactions — on the order of three thousand operations per second.

Five years was enough for a new infrastructure to become an ordinary part of national life.

The equivalent five-year anniversary of the ChatGPT moment will arrive on November 30, 2027.

By then, the important question will no longer be how many people “use AI.” That question will begin to sound like asking how many Brazilians “use Pix.”

People will use AI without identifying the activity as AI use. They will simply create, communicate, decide, shop, learn, travel, work, and entertain themselves inside environments where generation has become a standard capability.

The transition will not be complete. Institutions will still resist. Regulations will remain unsettled. Many organizations will continue operating through conventional applications and interfaces.

But the direction will be difficult to reverse.

The decisive moment occurs when a technology stops feeling like a technology and starts feeling like part of reality.

Pix has already crossed that threshold in Brazil. Artificial intelligence is approaching it everywhere.

Its final form will not be an app sitting on the screen. It will be the seemingly invisible capacity of the world around us to make what we ask for.