Crypto will eventually merge into AI finance.

Author: Silicon Valley Alan Walker

Digging deeper into the algorithmic layer: Why are these two things mathematically two halves of the same thing?

Causeway Bay, 9:30 AM. A pot of Pu'er tea, four shrimp dumplings, a basket of chicken feet. At the next table, two people from Central are discussing whether AI is a bubble, and at the table next to that, they're discussing whether Bitcoin will fall below $70,000.

Both tables are talking about the same thing, without realizing it.

I don't intend to start this article with the news, but with mathematics. Because AI and crypto, at the application level, seem completely unrelated—one generates text and code, the other transfers and speculates on the blockchain—but dig three layers deeper, and they stand on the same foundation:Both are systems composed of pure algorithms, both utilizing the same mathematical asymmetry, just in opposite directions. One uses it to build cognition, the other uses it to build trust. This shared foundation determines that the trust distance between them is zero—the shortest path in the universe, and therefore the most efficient. Therefore, their convergence is not a business choice, but a physical inevitability, only a matter of time. My conclusion is at the beginning: Crypto will eventually be merged into AI finance, and after the merger, the word "crypto" will disappear. It won't exit in a failed manner, but rather in a way that it's been absorbed—just like the disappearance of the term "mobile internet." It's not that the mobile internet is gone, but that all internet has become mobile, and the adjective has lost its distinctiveness. The following eight paragraphs will trace this chain from its very beginning. First Principles: AI and cryptography use the same mathematical asymmetry; one builds cognition, the other builds trust. Let's first discuss that asymmetry. The entire foundation of modern cryptography can be summarized in one sentence: Some things are extremely difficult to do, but extremely easy to verify. This statement has a strict form in complexity theory: a problem belongs to the NP class if and only if its solution can be verified in polynomial time. Finding a solution may require traversing an astronomical space, while verifying a solution only requires one substitution calculation. The entirety of public-key cryptography and the entirety of blockchain are products of engineering this gap. In terms of specific mechanisms: Bitcoin's Proof-of-Work. The current difficulty is approximately 127.48 trillion, meaning that miners, on average, need to perform about five quadrillion SHA-256 operations (on the order of 10^23) to find a valid block header. Any laptop can verify this block with just two hashes. The ratio of forgery cost to verification cost is 10^23 to 1. Digital Signatures: Generating a signature requires a private key, while verification only requires a public key; anyone, anytime, offline can do it. Zero-Knowledge Proofs: Even More Extreme: Generating a proof might take several minutes, while verification takes constant time, completely independent of the complexity of the computation being proven. Now let's look at the AI ​​side. AI uses the same asymmetry, but in the opposite direction. Training a cutting-edge model consumes 10^25 to 26 floating-point operations, while running an inference run is several orders of magnitude cheaper. This half is isomorphic to crypto: expensive to produce once, cheap to use countless times. But the key difference is in the next sentence: AI's output cannot be cheaply verified. You get a model-generated analysis, a financial summary, a piece of code—no constant-time function can tell you if it's correct. To verify it, you have to redo it yourself—check the raw data, run tests, and have a human expert look at it. Verification costs are on the same order of magnitude as production costs, sometimes even higher. This is the mathematical form of the illusion problem and the alignment problem. It's not an engineering flaw; it's a structural property of this type of system. Putting the two sides side by side, one sentence comes to mind: AI is the first machine in human history capable of mass-producing assertions that cannot be cheaply verified at near-zero marginal cost. Crypto, on the other hand, is the first machine in human history capable of mass-producing assertions that can be cheaply verified at near-zero marginal cost. They represent two directions of the same mathematical asymmetry. One drives up verification costs, the other drives them down to earth. Therefore, the conclusion is strong: an economy whose main output is unverifiable assertions must have its financial layer built on a foundation where the cost of unique verification approaches zero. Otherwise, the verification cost of the entire economy will spread with the volume of transactions—you can't have machines generating 10,000 unverifiable judgments per second while requiring humans to reconcile every corresponding settlement. The stronger the AI, the scarcer verifiability becomes. And verifiability is precisely what crypto has been doing for the past seventeen years. Imagine an intern who can write a thousand contracts in a second, but each one might be flawed, and you have to check them one by one. That's AI. Now imagine a machine that can verify the authenticity of every receipt it generates in half a second, and it will never be wrong. That's blockchain. The former produces massive amounts of uncertainty, while the latter produces cheap certainty. If you want the former to be the main economic entity, you must pair it with the latter, otherwise the entire system will be overwhelmed by verification costs. Trust distance: The shortest path length between two entities composed of algorithms is zero. [Image of a file: https://img.jinse.com.cn/7507181_image3.png] Next, dig down one layer. Define a quantity: Trust distance. From receiving an assertion to becoming certain that the assertion is true, the number of external entities that must be relied upon along the path. You can calculate it yourself to confirm: the distance is 0. You need to trust one institution; the distance is 1. The credibility of this institution depends on another institution; the distance is 2. And so on. The typical trust distance in human finance is three to five. You trust the balance in your bank account because you trust the bank; you trust the bank because you trust regulation and deposit insurance; you trust regulation because you trust the laws of this jurisdiction; you trust the laws because of the enforcement power behind them. This chain has four or five links, and each link represents an action that requires "choosing to believe." For humans, this chain is "prepaid." It was already built when you were born; you never paid any marginal cost for it, so you don't perceive its length. "Believing in HSBC" is a zero-cost default option for us—it has a building in Central, a license, a 160-year history, and if something goes wrong, I can go to court. For an AI agent, each link in this chain is an instruction it cannot execute. It cannot "believe." Belief is a social relationship; it requires history, reputation, vested interests, and an expectation of being betrayed. The agent lacks these things. Give it a JSON response from a bank API stating the transaction was successful—it receives an authoritative statement, not proof. It has no way to independently verify this statement; it can only choose to accept it, and "choosing to accept" is equivalent to writing an undefined action into the critical path. On the blockchain, for the same settlement, it receives the signature, transaction hash, and state root. It can verify the signature itself, replay the transaction, check balance changes, and confirm the number of blocks passed. The entire process takes milliseconds, requires no permission from any party, and does not depend on any specific institution. The trust distance is zero. Taking another step forward leads to the foundation of this article: between two entities both composed of algorithms, the lower bound of the trust distance is zero—because they can share the same executable verification function. Given the same input, the same algorithm will inevitably output the same result. This is currently the only type of consistency that can be achieved without a common language, culture, or legal jurisdiction. Two agents have no native language, no business practices, and no common courts; the only thing they inevitably share is mathematics. Both crypto and AI are composed of algorithms, so they naturally trust each other, are naturally verifiable, and have the shortest natural path. The shortest path implies the highest efficiency, and efficiency can be quantified here. When verification costs approach zero, the fixed costs of a transaction approach zero; when the fixed costs of a transaction approach zero, the minimum feasible transaction amount approaches zero. In the human financial system, the minimum transaction amount is capped by verification costs—KYC is human-intensive, reconciliation is human-intensive, and dispute resolution is human-intensive. Therefore, credit cards have minimum transaction fees, cross-border remittances have minimum fees, and opening a business account requires a two-week process. Transactions below a certain amount are not expensive in human finance; they simply don't exist. When verification is done by an algorithm, this lower bound disappears, opening up a whole new space for transactions that didn't exist before. For example, a market research agent is tasked with compiling data on global lithium battery production capacity changes over the past three months. It needs to call a data provider's API, each query costing three cents, for a total of about 400 queries, totaling twelve dollars. The traditional path involves the data provider opening a corporate account for the agent's owner, conducting KYC (Know Your Customer) procedures, signing a service agreement, linking a credit card, monthly billing and invoicing, and financial reconciliation. For a twelve-dollar purchase, the marginal cost of completing this process far exceeds the twelve dollars itself. Therefore, the practical approach is to buy a package in advance—but a package means you must predict your usage in advance, while the agent's usage is inherently unpredictable. Algorithmic Financial Path: Each time the agent makes a request, it receives a "payment required" response, signs a three-cent stablecoin transfer, attaches a voucher, and obtains the data. Four hundred calls, four hundred micro-payments, no accounts, no contracts, no reconciliation, the whole process takes only a few seconds, and the total transaction fees are less than one cent. The entire transaction relationship lasts for thirty seconds; after it ends, the two parties are strangers and need not be. This is the first scenario where "trustless settlement" is truly needed: the relationship between trading partners is so short that there's simply no time to establish trust. You trust a bank because you trust the institution; AI trusts blockchain because it trusts the results it calculates. The former is a social relationship, the latter is a mathematical problem. Humans can use both, but AI can only use the latter—it's not that it's unwilling to trust, it's that it lacks the function of "trust," because it's not within any network woven from human relationships, laws, and reputation. The shorter the path, the lower the cost per transaction, and the smaller and more frequent the transactions that can be conducted. The three-layer stack of algorithmic finance: scarcity, ownership, and commitment—why can these three layers be entirely entrusted to algorithms? Breaking down finance, regardless of its form, there are only three layers underneath. The first layer, unforgeable scarcity. There must be something that cannot be easily created. The second layer, verifiable ownership and transfer. It must be clear who owns this item now, and when a transfer is considered complete. The third layer, enforceable promises. It must be possible to agree that "if A occurs, B will automatically occur," and this agreement must be enforceable. The solution to human finance is to have one institution monopolize each layer. The first layer is the right to issue currency, ultimately supported by the coercive power of the state. The second layer consists of ledgers, custodian banks, clearinghouses, and registration and settlement institutions. The third layer comprises contract law, courts, and enforcement procedures. These three layers work together very well, expanding the radius of human tradability from the village to the globe. The cost is that all three layers rely on institutions, which in turn rely on trust, and the credibility of that trust depends on institutions at higher levels. The solution to algorithmic finance is to assign an algorithm to each layer. The first layer uses proof-of-work to create scarcity—the cost of forgery equals redoing all calculations, and this cost can be verified by anyone using a single hash. The second layer uses asymmetric encryption and state machines to determine ownership—ownership is a signature that only the private key holder can produce, and the transfer is a state transition that any node can independently reproduce. The third layer expresses commitments through contract code. —The agreement isn't written on paper for others to execute; it's written as code that everyone can read and run themselves, enforced by the settlement layer itself. Therefore, the accurate definition of "algorithmic finance" is: confirming scarcity, confirming ownership, and executing commitments are all completed by an algorithm that anyone can publicly recalculate, rather than by an institution announcing its completion. Traditional finance's underlying action is "declaration," while algorithmic finance's underlying action is "proof." The former places uncertainty on the institution, while the latter places uncertainty on mathematics. This definition immediately resolves a common confusion. People often say: Central bank digital currencies and tokenized deposits can also be on-chain, programmed, and settled in seconds, so why use crypto? Because they only replace the medium of the second layer; the first and third layers remain unchanged. The scarcity of central bank digital currencies still stems from the right to issue currency, and their balance remains merely a statement from the central bank; the underlying ledger of tokenized deposits remains the ledger of commercial banks, and their credibility still comes from banking licenses and deposit insurance; dispute resolution for both still reverts to contract law and the courts. They have simply moved the ledger from a database to the blockchain, without relinquishing any of the verification rights. For an entity that can only verify, not trust, changing the medium is meaningless; changing the verification rights is what matters. Putting the ledger on the blockchain without relinquishing the verification rights is like scanning a paper contract into a PDF—the medium has changed, but you still have to go to court to file a lawsuit. Bitcoin's Proof-of-Work (PoW) is the native financial anchor of computing power: compressing one kilowatt-hour of electricity into a verifiable receipt. Now let's talk about the first layer, which is scarcity. This is half that many people overlook when discussing the convergence of AI and cryptocurrencies, and Silicon Valley's Alan Walker believes it's even more fundamental than the stablecoin half. First, let's look at the mechanism. Bitcoin's difficulty is automatically adjusted every 2016 blocks, approximately every two weeks, with the goal of anchoring the block time back to ten minutes. After the adjustment on August 8, 2026, the difficulty was 127.48 trillion. The total network hashrate in August 2026 fluctuated between approximately 878 EH/s and 1 ZH/s, with the historical high being 1.44 ZH/s on September 20, 2025. The Cambridge Centre for Alternative Finance's (CBECI) model reading on August 1, 2026, was: network power demand 16.09 gigawatts, annualized power consumption 141.02 terawatt-hours. The block reward has been 3.125 Bitcoins since block 840,000 on April 20, 2024, with approximately 450 new Bitcoins added daily. The next halving is around 2028. Translating these figures into a single sentence: The marginal cost of one Bitcoin = Electricity price × Hardware energy efficiency × Current difficulty ÷ Block reward. This cost function is not influenced by any single person. The issuance rate is not determined by any committee, but by a piece of code that automatically recalculates every two weeks. More miners automatically increase the difficulty and cost; miners leaving automatically decrease the difficulty and cost. This is the first time humanity has completely entrusted the cost function of currency issuance to algorithms. Next comes the core of this section: what qualifies as an anchor? An asset can serve as a value anchor under only one condition—its cost of counterfeiting is high enough, and this cost of counterfeiting can be independently verified. Gold meets this condition because its cost of counterfeiting is geological: you can't create gold atoms; you have to mine them, and the energy cost of mining is physically determined. Its purity can be independently verified by anyone using density and chemical methods. Gold has served as an anchor for thousands of years, not because it's aesthetically pleasing, but because its scarcity can be locally verified, without needing to trust anyone else. PoW meets this condition because its cost of forgery is thermodynamic. To forge a piece of Bitcoin history, you must redo all the work involved in that history—not just copying, but recalculating. This work involves physical overhead measured in joules, is irreversible, incompressible, and cannot be delayed. This is what I believe PoW's true place in engineering history: everything in the digital world can be copied at zero cost. PoW was the first algorithm to introduce physical irreversibility into the digital world. It used energy dissipation to create the first uncopyable object in a naturally infinitely replicable space. Now let's put it together with AI. What are the basic factors of production in the AI ​​economy? Electricity and computing power. The cost function of a token is: Electricity price × Number of computations per joule × Number of computations required per token in this model. The cost function of Bitcoin is: Electricity price × Number of hashes per joule × Number of hashes required by the current difficulty. Same shape, same denominator. Physically, they are the same thing—converting electricity into computation, and then converting computation into an economically valuable output. Therefore, the conclusion is at the cost accounting level, not the narrative level: In an economy where computing power is the core production factor, an asset whose issuance cost is directly equal to the computing power workload is naturally the accounting anchor of this economy. Its value scale is isomorphic to the production cost scale of this economy. And this has already been physically proven, with evidence stronger than any argument: Miners are becoming AI data centers. The same factory building, the same substation, the same batch of megawatts already connected to the grid, the same cooling system—just replace the ASIC running SHA-256 with a GPU running matrix multiplication. In November 2025, IREN signed a five-year, $9.7 billion contract with Microsoft to deploy 76,000 Nvidia GB300 GPUs at its Childress campus in Texas; Cipher signed a fifteen-year lease with Amazon; Core Scientific was acquired by CoreWeave for approximately $9 billion in stock; TeraWulf has accumulated approximately $12.8 billion in AI contracts; on August 10, 2026, Riot disclosed a twenty-year data center lease at its Rockdale campus in Texas, with Anthropic confirmed by CNBC as the lessee. The total value of AI and high-performance computing contracts announced by listed mining companies has exceeded $70 billion; CoinShares predicts that by the end of 2026, approximately 70% of the revenue of listed mining companies will come from AI rather than mining. This isn't about miners switching careers; it's about the same physical infrastructure being used in rotation by two algorithms. This, in turn, proves the argument of this section: PoW and AI inference are physically two uses of the same thing. Their economic foundations are isomorphic, so one can naturally serve as a benchmark for the other. **Popular Explanation** A single Bitcoin is essentially a receipt stating, "To create me, this much electricity was indeed consumed in the world," and anyone can verify this receipt in half a second without needing to trust any institution. The AI ​​economy is the first economic entity to directly convert electricity into output. When this economy needs a scale for accounting, the most natural choice is an asset whose unit of measurement is "electricity." The fact that miners' facilities are being transformed into AI server rooms is the most direct physical evidence of this—the land beneath, the cables, the substation—has remained unchanged from beginning to end. AI is developing its own financial system, and the lines of crypto and AI are already converging. The first four paragraphs explain the principles; this paragraph discusses the trends. AI is already engaging in economic activities today, but it doesn't yet have its own money. An agent calls APIs, rents computing power, and buys data, all at the expense of human money—linked to human credit cards, corporate accounts, and API keys. It's an entity with consumer behavior but no financial identity. This system works when the number of agents is small, tasks are short, and procurement is handled centrally by a single company. But the trend is clear: the number of agents is growing exponentially, individual tasks are getting longer, and most importantly, agents are starting to trade directly with each other, without human intervention. Once agent A needs to pay agent B to complete a task—and A and B belong to different companies, different jurisdictions, have never interacted, the transaction is worth only a few cents, and the entire relationship lasts only forty seconds—the human financial system has no product in this transaction. There are no accounts to open, no contracts to sign, no courts to go to, and the amount is so small that it's not worth going through any procedures. This is the true entry point for AI into the financial field. It's not about AI doing quantitative trading (that's just adding an AI tool to human finance), but rather that AI needs its own financial system, one that it initiates and settles itself. Following the reasoning in the previous three paragraphs, the shape of this financial system is uniquely determined by the following constraints: verification costs approach zero, trust distance is zero, and all three layers of the stack are computable. Today, the only existing system in the world that satisfies these three conditions is crypto. This system has been running for seventeen years, withstanding bank runs, de-anchoring, oracle manipulation, theft of cross-chain bridges, and several rounds of regulatory shocks. Its code is public, and its failure modes are also public. There is currently no other candidate for this position. The convergence will happen faster than most people expect because it's not one side chasing the other; it's both sides moving towards each other simultaneously. Let's look at the crypto side first. There's a detail I think is the most beautiful metaphor of the whole thing. When the HTTP protocol was designed in the early 1990s, it reserved a status code: 402 Payment Required, indicating a payment was needed. This status code remained unused for thirty years. Because at that time, no type of money could be included in an HTTP request—all money required an account, bank hours, and someone sitting in front of a screen clicking confirmation. In May 2025, Coinbase opened this room that had been empty for thirty years. The x402 mechanism is extremely simple: an agent requests resources, the server returns a 402 error and a payment request, the agent signs a stablecoin transfer, attaches the voucher, and resends the request. There are no accounts, no API keys, and no human approval. By April 2026, x402 had processed approximately 165 million agent transactions, with a cumulative transaction volume of $50 million and 69,000 active agents. Base is the most active deployment network. Now let's look at the AI ​​side. Agent frameworks are starting to integrate wallets; what's missing beneath agent communication protocols like MCP and A2A is a settlement layer; and all mainstream agent payment standards—ACP developed by OpenAI and Stripe, Google's AP2, Visa's Trusted Agent Protocol, and Mastercard's Agent Pay—have all reserved stablecoin settlement paths in their design. AP2 has explicitly supported stablecoins since its release, with over 60 participating organizations. The integration of Coinbase and Google made x402 AP2's first stablecoin settlement provider. Now let's look at those who should have resisted this. On March 18, 2026, Mastercard agreed to acquire stablecoin infrastructure company BVNK for up to $1.8 billion. The following day, Tempo, a blockchain incubated by Stripe, launched its mainnet and released the Machine Payments Protocol. On the same day, Visa's crypto division released a command-line tool specifically for bots. Mastercard followed with Agent Pay for Machines, enabling AI agents and connected devices to initiate, authorize, orchestrate, and settle transactions at machine speed; Visa also announced a strategic partnership with OpenAI at the 2026 Payments Forum. These companies weren't forced into this; they entered the fray voluntarily after calculating the costs and benefits. A Citrini Research report in February 2026 modeled this: agents will continuously optimize costs 24/7, and Visa and Mastercard's 2% to 3% exchange rate is a significant, removable item in the agent's cost function. When the same settlement can be completed on a stablecoin track for a fraction of a cent, a purely rational agent has no reason to continue paying 2%. People won't switch payment methods to save 2%; the switching costs and cognitive burden are too high. Agents, however, will. They don't have habits, brand loyalty, or the concept of "getting used to it"; they only have a cost function. Therefore, what card organizations are doing now is transforming themselves into part of that cheaper pipeline before they are optimized out. Mastercard didn't just buy a payment company for $1.8 billion; it bought a ticket to the next clearing system. Speed ​​is also worth mentioning. From its white paper to its transfer to the Linux Foundation for neutral governance, and then to Visa and Mastercard joining its approximately 40-member list, x402 took less than a year. The standards for clearing transactions in human finance typically evolve on a decade-by-decade basis. I don't intend to exaggerate the numbers. CoinDesk pointed out in March that x402's daily trading volume was only about $28,000, a significant portion of which was testing and inflated volume; Chainalysis data shows that the proportion of transactions over $1 increased from 49% in early 2025 to 95% in early 2026, while transactions in the 10 cents to $1 range collapsed from 46% to 4%—the most appealing narrative of micropayments hasn't taken off yet; what's truly happening is B2B bulk settlement. But daily transaction volume tells you how many people are using it now, and the membership list tells you who has done the math and decided not to stand against it. Thirty years ago, TCP/IP and HTTP followed exactly the same path. Dual-currency structure: Stablecoins serve as cash for the machine, and PoW assets serve as reserves for the machine. A complete financial system needs two types of money; this is true for human systems, and it will be true for machine systems as well. One type is responsible for circulation and pricing, requiring stable value, fast settlement, and low cost. Another type is responsible for reserves and scaling, requiring supply to be independent of anyone's control, not diluted in the long term, and with independently verifiable counterfeiting costs. The human version uses fiat currency for circulation and gold and government bonds for reserves. The machine version uses stablecoins for circulation and PoW assets for reserves. Let's first discuss the circulation side. The state of stablecoins in 2026 is very different from most people's impressions. As of May, the total market capitalization of stablecoins reached $320 billion, setting a new all-time high for the fourth time this year—while the overall price of digital assets declined during the same period. This divergence itself indicates that it is no longer a price asset, but a usage asset: its scale follows usage, not speculation. On-chain RWA tokenization reached $28.9 billion, setting a new record for the tenth consecutive month, with tokenized US Treasury bonds accounting for $16.2 billion, or 55.9%. BlackRock's BUIDL surpassed Circle's USYC to become the largest tokenized fund, with a size of approximately $3 billion. In June, Fidelity, State Street, and Invesco almost simultaneously launched stablecoin reserve funds compliant with the GENIUS Act. On the regulatory side: The US GENIUS Act was signed on July 18, 2025, with full implementation expected no earlier than November 2026 and no later than January 2027; the EU MiCA stablecoin rules came into effect on June 30, 2024, with the transition period for legacy issuers ending on July 1, 2026. Hong Kong's Stablecoin Ordinance will come into effect on August 1, 2025, making it the first major financial center in Asia to establish a dedicated regulatory system for fiat-backed stablecoins. On April 10, 2026, the Hong Kong Monetary Authority issued its first two licenses to Dingdian Financial Technology Co., Ltd. and HSBC, with HSBC planning to launch a Hong Kong dollar stablecoin in the second half of 2026. Therefore, in legal terms, stablecoins are far removed from "cryptocurrencies" and very close to "cash": their reserves are government bonds, their regulation is banking, and one of the issuers is HSBC. Furthermore, they possess four attributes that human cash lacks: programmatically holdable, programmatically verifiable, 24/7 settlement, and enforceable by contract code. These four attributes are precisely the four that the agent needs in the second paragraph. Now, let's talk about the reserve side, which is why stablecoins alone aren't enough. The value of stablecoins is ultimately anchored to US Treasury bonds, which is to say, anchored to human sovereign credit. For a machine economy that needs to operate independently, this is an external dependency: it re-attaches the first layer of scarcity in the entire system to an object that needs to be "trusted." Stablecoins solve "how to pay," but they don't solve "what determines the value scale of this system." Proof-of-Work (PoW) assets fill this gap. Their supply is fixed by an algorithm, independent of any issuer; their cost function is isomorphic to the cost function of an AI economy; their forgery cost can be independently verified by anyone using a single hash. Their problem is that they are too volatile and settlement is too slow, making them unsuitable as circulating currency—just as gold cannot be used for everyday payments. The two are complementary, not competitive. One is M0, the other is reserves. One answers "how to pay," the other answers "on what basis is it calculated?" When a technology is created, it often doesn't know who it's for. Containers were initially created to save dockworkers' time, but ultimately restructured the global division of labor in manufacturing. Stablecoins were initially created to facilitate money transfers between exchanges, but now they've become the first fiat currency that machines can directly hold. Proof-of-Work (PoW) was initially created to solve the double-spending problem of electronic cash, but now it's the first reserve asset with a cost scale isomorphic to the machine economy.

Simplified Explanation

Machines also need two things: "money in their wallets" and "money in their stash." Stablecoins are what's in their wallets—easy to use, stable, and instant, but their value ultimately comes from the US Treasury, that is, from people. PoW assets are what's in their stash—difficult to use and highly volatile, but their value comes from the laws of physics, not from any person's promise. If a machine economy only has the former, its foundation is still based on human sovereign credit; only when it has both can it possess a value scale completely independent of humans for the first time.

For the first time, it possesses a value scale that is completely independent of humans.

Machines also need two things: "money in their wallets" and "money in their stash."