Author: Zuo Ye Wai Bo Shan
We live in a vast gap, and the passage of time always brings a sense of dizziness.
In 2023, Musk called for small payments to block Bots and maintain the community atmosphere for human discussion. But just two or three years later, Bots, empowered by AI, are about to become the main actors of small payments and are showing a "rejection emotion" towards humans.
Image Caption: Musk wants to use micropayments to block Bots
Image Source: @elonmusk
It's not just ordinary white-collar workers who feel abandoned; more so, it's the sense of backwardness in the crypto community. You can repeatedly experience this despair from Vitalik's sutra-like repetition of the importance of ZK for trustworthy AI, de Agents like Virtuals seeking decentralized LLMs, and the latest—Agents should consume stablecoins.
This time, Crypto must prove its value to AI, in the name of stablecoins.
AI Steps into Productivity, Crypto Stuck in Finance
When you pursue freedom, you have already lost it.
In the narratives about AI and Crypto, the most classic is "AI is responsible for productivity, Crypto is responsible for production relations," but this relationship has never materialized.
Production relations refer to human collaboration models, and the biggest feature of this round of AI is "replacing humans."
AI Agents are climbing out of the screen, taking over white-collar jobs, while also invading the mechanically repetitive areas of manual labor. Admittedly, as Jensen Huang said, AI itself is also limited by physical hardware like electricity, firmly anchored in the geometric space woven by humans.
Image Caption: AI enters the production field
Image Source: @zuoyeweb3
If humans can only serve the production and consumption of AI, can they still preserve their subjectivity in labor?
At the same time, Crypto's vision is gradually narrowing, from early decentralized storage like IPFS/Filecoin, to computing power and storage projects in the 24/25 cycle, to USDAI focusing on GPU lending models. You can't say Crypto isn't trying, but each wave fails to capture AI's real use cases.
"AI is responsible for both productivity and production relations" is more likely to come true. Even if the Agent narrative is not a bubble, the stablecoins used may not be "stablecoins based on BTC/ETH assets, running on Ethereum."
I am not spreading terror or pessimism. The Genius Act has effectively allowed the OCC to take the definition power of "compliant stablecoins." Sky's USDS is just a U.S. debt certificate running on Ethereum. So, in the upcoming RWA era, will more assets choose Canton backed by Goldman Sachs or Solana, which is actively BDing?
ETH L1 has once again become a high-performance public chain, at the cost of node scale exploding rapidly. First, use compliant stablecoins to replace "YBS," then use Canton to replace ETH/SOL, and finally co-opt Vitalik as an institutionalized rebel.
Liangshan was good because it could surrender; the coin circle is bad because it's all about finance.
But Liangshan had the military foundation to surrender; the coin circle has all the套件 for financialization. And, no matter how much Agent replaces human consumption power, it must act according to human will.
In other words, Agent is the behavioral subject, and humans are the意愿 subject. Agent's consumption behavior can only be an infinite extension of human will.
In March 2026, Stripe launched the MPP protocol (Machine Payments Protocol), seemingly to摆脱 humans, but it is actually a failed restart之作 to incorporate machines into the existing financial system.
As early as last September, Stripe联合 OpenAI launched the ACP (Agentic Commerce Protocol) protocol, hoping to replace the previous "Google Search + Amazon Mall" through OpenAI's chat page. But unfortunately, Agent cannot cross the non-standard存量 of the old giants over 20 years, and the complexity leads to a conversion rate that is聊胜于无.
There are many such protocols, including Visa and the banking industry, which are urgently launching their own "machine payment" protocols. But归根到底, they are just adding stablecoins on the acquiring side. You hardly see merchants actively integrating stablecoins.
This is not to deny the trend of combining stablecoins and Agents. Borrowing a sentence from @Shoal Research, "In the past fifty years, efforts to replace card organizations, from PayPal to Apple Pay, have ultimately failed."
If the stablecoins issued by banks, running on Visa's private chain, ultimately defeat Crypto's efforts like they defeated Fintech, this story is not friendly to us.
On this point, Crypto must learn from AI, from Prompt Engineering, to Context Engineering, to the current Harness Engineering. The survival form of Agent has been constantly changing.
Early on, people optimized their prompts in Chat; then, people optimized their expressions in AI response texts, summarizing them into skills, and then疯狂购买 API and Mac Mini and other software and hardware services, ultimately training Agent into a version that is easier to replace themselves.
The narrative of Agent consuming stablecoins is a helpless move after only having financial value. It can only sell its cheap, fast and other technical narratives to AI and the public. This is无异于 putting on the rope that will hang itself, and at a clearance price.
Crypto Token ⇄ AI Token ⇄ Crypto Token
Stay away from the sound of gunfire, observe the battle from afar as much as possible.
If you are caught in it, try to create a new battlefield for yourself.
Providing liquidity for the AI industry is a dead end (no money, and it will relegate the coin circle to the background). Being treated as SaaS and channels will be swallowed and have profits taken away. Only by focusing on volatility can we trigger human FOMO emotions, thereby creating miracles of asset price surges.
The most exciting narrative of Agent is being the dual subject of production and consumption. Letting Agent consume can greatly surpass human physiological limitations:
- The number of humans is limited, or rather, the subjects of stablecoin consumption are limited, at most 8 billion. But the number of Agents is endless,无限递归. A human's Agent calls an Agent; the vassal of my vassal is not my vassal.
- Agent does not need sleep. This is the first time a "tool" has shown a physiological advantage over humans, a biological advantage over human labor—not smarter, but more耐磨, the Scaling Law on the time scale.
- Agent is good at handling Fuzzy tasks, or always gives you a solution. In multi-Agent mode, AI shows all-weather working ability for the first time.
Agent is not an evolution of the Ford assembly line, but the theoretical optimum of Taylorism. People are unstable, but Agent can be continuously tuned to最终适配 the ultimate dream of capitalism—capital appreciation.
This is not alarmist. The word Token, in both Chinese and English contexts, already points to the throughput of AI, no longer the competition of PoW.
Image Caption: The变迁 of the computing power era
Data Source: @DigiEconomist @IEA
Ethereum abandoned the PoW mechanism in 2022; ChatGPT started the AI computing power competition era in 2022. Fate likes to play tricks on us. When Token in crypto became a pure emission game, more people chose AI supported by computing power consumption.
Transitioning from Crypto Token to AI Token is not difficult; stablecoins are trying. But the reverse, converting AI Token into the liquidity of Crypto Token, has become a difficult sutra to chant.
Computing power, reasoning, storage—all lost. How can the stablecoin story come true?
Perhaps AI can give us a reference. Agent's replacement of Chat is based on the fact that the Chat model cannot sustain a business model. ChatGPT shows no signs of replacing Google or Amazon; Claude specializes in "deterministic" areas like Coding.
Now it's time to consider whether Crypto will decline slowly and painfully like Binance, or rapidly like FTX.
We must learn from AI in reverse: create volatility based on stability. This is what crypto is best at. Traditional finance's adoption of AI pursues accuracy to reduce labor, but traditional finance's adoption of Crypto pursues "tokenization" to reduce friction.
This is not Crypto's own story. Volatility has always been built on a stable foundation. For example, RWA shifts from national debt to corporate debt; lending shifts from floating rates to fixed income.
What the crypto industry needs is, under the premise of Agent using stablecoins, how to create volatility, or rather, the inflation of asset prices?
Image Caption: Government becomes the consumption subject
Image Source: @OurWorldInData
It's not as difficult as we imagine. Since the industrial revolution, the state has become the main body of economic operation. Even after the neoliberal narrative began in the 1980s, the proportion has still been steadily increasing.
The water released by QE, strong bank regulation, eventually turned into the withdrawal tide of asset management giants in 2026. This market never lacks AUM; it always lacks demand侧.
This is the significance of stablecoins for the retail market. Whether you package it as the chosen one for Agent or a post-human necessity, human emotions are always needed to trigger the initial purchase.
The AI narrative is the same, and the crypto narrative is the same. Ultimately, they must be sold to a small demand market.
Only technological progress can generate wealth surplus, gradually提高 the overall per capita income level, and thereby提升 consumption levels. This is the story AI tells. But now, the wealth effect of tokens is approaching zero; there is no growth. This is Crypto's biggest困境.
It has never been about productivity and production relations. This is a良缘 about consumption and finance. AI can become the subject of consumption; the government is the driving force of consumption. But humans are always needed to execute the financial big bang to complete the cycle.
The leap from AI Token to Crypto Token also requires letting Token break free from quantity restrictions and re-enter the space of imagination.
Conclusion
Circle's真正聪明之处 is to let Agent replace the limits of human numbers and consumption power,造了一个 dream of infinite users for the capital market.
But compliant Circle will also fall sharply due to the clear act "prohibiting" passive interest. This shows that the market hopes for a vague sense of朦胧感—when compliance is不利, there is also room for survival.
This is the characteristic of the crypto industry—always the frontline financial laboratory.
Current AI is becoming increasingly stable in methodology. The only breakthrough direction is Li Fei-Fei and Yann LeCun's world model, but this is essentially not a complete innovation of "algorithms," but still a dimensional update on the data side.
The significance of Agent for consumption and finance is about to enter the落地 stage after experimentation, and stablecoins carry all the hopes of Crypto.
