BlackRock explains why AI agents prefer Bitcoin over fiat currency, and a Brazilian shows that the answer is 134 years old

In a study cited by the manager, none of the 36 AI models chose fiat currency, something that the Austrian School of Carl Menger already explained in 1892, and João Paulo Mayall updates the theory for the machine age with practical salability, proposed in a paper on SSRN.

BlackRock published a report in September on the native economy of machines, which cites a study with 36 artificial intelligence models, and none of them chose fiat currency as their first option. For entrepreneur João Paulo Mayall, co-founder of QR Capital and responsible for QBTC11, the first Bitcoin ETF in Latin America, the reasons pointed out by the manager were already in an article published by Austrian economist Carl Menger in 1892.

"BlackRock just published a paper on AI agents that, without mentioning Menger even once, confirms his answer", Mayall wrote in a thread posted on X this Thursday (24). Menger's name, founder of the Austrian School of economics, does not appear on any of the 11 pages of the document.

In 1892 Carl Menger answered a question that seems silly: "Why does everyone accept money?" BlackRock just published a paper on AI agents that, without mentioning Menger even once, confirms his answer. 🧵👇 pic.twitter.com/EFOFWxLaC2 --- J. P. Mayall (@jpmayall) September 24, 2026

What the machines chose

The study cited by BlackRock was conducted by the Bitcoin Policy Institute (BPI), a research organization based in Washington, and tested 36 models from 6 developers (Anthropic, DeepSeek, Google, MiniMax, OpenAI, and xAI) in 9,072 open monetary scenarios, without suggesting any currency.

In total, Bitcoin was the most frequent choice, with 48.3% of the responses, followed by stablecoins with 33.2%. Fiat currency and bank money accounted for 8.9%, and none of the 36 models placed this option first.

The results change according to the function. In scenarios of value storage over several years, Bitcoin accounted for 79.1% of the responses (1,794 out of 2,268), compared to 6.7% for stablecoins and 6% for fiat currency. In payments, the order reverses, with stablecoins at 53.2%, Bitcoin at 36%, and fiat currency at 5.1%. Preferences of the 36 AI models in value storage and payments. Source: Bitcoin Policy Institute (moneyforai.org)

BlackRock mentions the survey on page 4 of the report as preliminary support for the idea of a monetary architecture where stablecoins serve for transactions and Bitcoin for storing value. The manager emphasizes that these are simulated responses from the models, not observed behavior from agents.

The barriers that BlackRock points out

On page 5, the manager lists why current payment rails poorly serve high-volume, low-value transactions made by agents. Among the problems are account opening processes, credentialing, and authorization that may require a human, acceptance fees that make very low-value transactions uneconomical, and settlement restrictions, as most of the volume of the American ACH settles in up to 1 business day.

"An agent doesn’t have a CPF, doesn’t wait for the bank to open, and doesn’t pay a fee higher than the purchase. For the machine, this is the dimension that decides the game before any other," says Mayall.

The list from 1892

For Mayall, the list is practically the same as the one Menger published in the article "On the Origin of Money" in the British journal The Economic Journal in 1892. In the text, the Austrian economist explains that money was not invented by decree or convention. It arises because each person exchanges what they have for the good that is easiest to resell, a quality that Menger called Absatzfähigkeit, or salability.

"No one invented money. A fisherman who wants shoes cannot trade with a shoemaker who does not want fish, and the fish spoils before he finds someone who wants it. The solution is to first trade the fish for a good that anyone accepts, and that good becomes money on its own, without decree," he explains.

On pages 246 and 247 of the article, Menger lists the factors that limit the salability of a good. Mayall highlights 3 of them, which are the limitations imposed politically and socially on trade, the transportation cost in proportion to the value of the good, and the periodicity of the market. Excerpts from Carl Menger (1892) and the BlackRock report (2026) compared by João Paulo Mayall. Image: @jpmayall "When BlackRock says that opening an account requires a human, it is describing what Menger called a political limitation on trade. When it says that the fee makes payments of cents unfeasible, it is the cost proportional to the value. When it complains about settlement on business days, it is the market that only opens on certain days. It is the fish problem 134 years later, dressed differently," says Mayall.

The Contribution of the Brazilian

The original part of the argument comes from a paper that Mayall published on SSRN, an international repository for academic research, titled "Bitcoin as Validation of the Regression Theorem: An Austrian Synthesis with Szabo and Ammous." In the work, he compares gold, fiat currency, and Bitcoin in 10 monetary properties derived from Menger and economist Saifedean Ammous, author of The Bitcoin Standard, and adds an 11th, which he called practical salability.

The new dimension measures whether an asset can be used now, without asking for permission and without losing value along the way, based on 3 observable criteria, which are the spread, the settlement time, and access without intermediaries. In the assessment of the paper, which the author himself describes as qualitative, gold receives a score of 4, fiat currency 6, and Bitcoin 9. The 11 monetary dimensions compared in the paper, with practical salability highlighted. Source: Mayall (2026), SSRN 5376225 "Menger measured the salability of goods in a world of fairs and merchants. In the digital age, it is also necessary to measure the track, that is, whether money can be used now, without asking anyone for permission and without losing value in the spread. That is what I called practical salability," says Mayall.

According to him, the barriers listed by BlackRock correspond point by point to the criteria of the new dimension. "BlackRock described practical salability without naming it. The same criteria that I use to score gold, fiat currency, and Bitcoin appear in the report as the reasons why the banking system does not serve machines."

Save in Bitcoin, Spend in Stablecoin

Mayall summarizes the agents' choice in 2 filters. The first is practical salability, which eliminates traditional bank money. The second is the time horizon, which separates what the agent uses to pay from what they use to save.

"For the machine, money goes through 2 filters. Practical salability eliminates the bank, then the horizon decides between stablecoin and Bitcoin. It is the same division that Menger made between selling in space and selling in time," he says.

Reservations

The BPI study itself shows that preferences vary significantly among developers. The share of Bitcoin was 91.3% in Anthropic's Claude Opus 4.5, compared to 18.3% in OpenAI's GPT-5.2, although the results remained stable across different model configurations.

"These are simulation responses, not observed behavior, so it’s a hypothesis and not proof. However, it is a hypothesis that the Austrian School formulated 134 years ago, and now the world's largest asset manager has arrived at it through a different path," reflects Mayall.

Who is João Paulo Mayall

Mayall is an entrepreneur and executive in the digital assets market and co-founder of QR Capital, a management firm sold in 2025. He led the first Bitcoin ETF in Latin America (QBTC11) and three other pioneering crypto ETFs on B3, the country's first 100% crypto asset fund, and one of the first tokenization projects approved in the CVM sandbox, in addition to structuring the CCA® certification from Ancord, teaching the largest cryptocurrency course ever held by B3, with 27,000 participants, and advising Anbima on Web3. At the same time, he publishes papers on monetary theory from the perspective of the Austrian School on SSRN, including the study in which he proposes practical salability as the 11th monetary property, the basis for the analysis of this report, and another on the indirect exposure of sovereign gold reserves to U.S. Treasury securities.

See more at:

• João Paulo Mayall's thread on X: https://x.com/jpmayall/status/2103124950701715872
• Paper on SSRN: https://ssrn.com/abstract=5376225
• BlackRock report: https://www.blackrock.com/us/individual/literature/whitepaper/the-machine-native-economy.pdf
• Bitcoin Policy Institute study: https://www.moneyforai.org
• Carl Menger, On the Origin of Money (1892), original scan: https://archive.org/details/sim_economic-journal_march-december-1892_2_5-8