Yongche Founder Zhou Hang: Cryptocurrency, Finally Its Time Has Come

marsbitDipublikasikan tanggal 2026-04-05Terakhir diperbarui pada 2026-04-05

Abstrak

Cryptocurrency, after years of being perceived primarily as a speculative asset or a complex technological curiosity, is finally finding its true purpose: not as money for humans, but as money for machines. While cryptocurrencies like Bitcoin and Ethereum failed to become practical daily payment tools for people due to volatility, poor user experience, and regulatory complexities, they are perfectly suited for the emerging Agent-to-Agent (A2A) economy. In this new paradigm, AI Agents autonomously transact with each other—purchasing data, API calls, or computational resources—using crypto wallets and smart contracts. They require no customer support, handle micro-payments efficiently, and operate across borders without traditional banking hurdles. Protocols like x402 are reviving the long-dormant "402 Payment Required" HTTP status code, enabling machines to negotiate and pay for services seamlessly within milliseconds. We are witnessing a fundamental shift: machines will use cryptocurrency as a high-frequency, efficient medium of exchange in the background, while humans continue to use traditional fiat currency for daily life. Cryptocurrency isn’t replacing the bank account; it’s becoming the silent, foundational layer of a new machine-driven economy.

Author: Zhou Hang

Over the past decade, if you mentioned "cryptocurrency" to an average person, the words that likely came to their mind were: getting rich quick, pump and dump schemes, hackers, or some incomprehensible geek toy.

From the emergence of Bitcoin (BTC) to the smart contract revolution of Ethereum (ETH), to the noise of various public chains and stablecoins, this world has been noisy for over a decade. Countless brilliant minds and massive amounts of capital have flooded in, trying to build a decentralized utopia.

But we still feel confused in real life: besides being a highly volatile speculative asset, besides buying low and selling high on exchanges, what is cryptocurrency actually useful for? When we go downstairs to buy a cup of coffee, we still scan WeChat Pay or Alipay; for international transfers, we still go through cumbersome bank wire processes.

It claims to颠覆finance, yet it似乎can't even handle the most basic "payment" properly.

Until today, with the arrival of the A2A (Agent to Agent) intelligent economy, this confusion finally has an answer: cryptocurrency did not fail; it just spent the past decade targeting the wrong users.

Why Cryptocurrency Couldn't Become "Human Money"

When Satoshi Nakamoto released the Bitcoin whitepaper in 2008, the title prominently read: "A Peer-to-Peer Electronic Cash System." His original intention was to create an everyday payment tool.

In 2010, a programmer named Laszlo bought two pizzas with 10,000 Bitcoins. This was seen as the great beginning of cryptocurrency payments. But the subsequent script went in another extreme direction.

There are three insurmountable practical obstacles why cryptocurrency cannot become money for daily human use:

First is volatility. When something is worth $1 today, might drop to $0.5 tomorrow, or rise to $2, no one dares to price things in it. There's an economic常识called "Gresham's Law" (good money is hoarded); when you expect Bitcoin to rise, you absolutely舍不得use it to buy pizza.

Second is the anti-human experience. Humans are creatures who极度dislike hassle. Crypto payments require you to妥善保管a long string of私钥like gibberish; once lost, assets are instantly zeroed out, with no customer service to help you recover them. You also need to understand what Gas fees are and endure long waits during network congestion.

Finally, regulation and taxes. In many countries, buying a coffee with cryptocurrency is considered an "asset sale" by the tax authority, for which you need to calculate and declare capital gains tax.

Humans need stable, simple financial services with customer service backup and legal protection. The traditional banking and fiat system, while having friction, perfectly meets human needs for security.

Cryptocurrency tried to pull humans into a cold, absolutely rational, self-risk code world, and was naturally rejected by humans, eventually morphing into a form of "digital gold" and speculative chip.

Machine Money: When Agents Become Consumers

But what if we shift our perspective from "human" to "machine"?

In the A2A intelligent economy, hundreds of millions of AI Agents will be calling APIs, buying computing power, acquiring data, and even negotiating rental contracts for you in the background every day. The CEO of Coinbase pointedly noted: "AI can't take an ID card to open a bank account, but they can control a crypto wallet without any障碍."

For AI Agents, the advantages of the traditional financial system are all disadvantages, and the disadvantages of cryptocurrency are all advantages.

Machines don't need customer service; they only trust code. Traditional contracts require lawyers to draft, courts to enforce, and banks to settle, taking days or even months. In the Agent world, they use "smart contracts"—this is essentially a program stored on the blockchain. When conditions are met, funds are automatically transferred instantly, with no possibility of违约. This is the true "machine-native contract."

Machines need millisecond-level micro-payments. Imagine an AI Agent is generating a report for you; it needs to buy a piece of real-time data from another Agent for $0.001. The per-transaction fee for traditional credit card networks is as high as $0.30,根本无法supporting such micro-transactions. Through加密networks, an Agent can complete extremely low-cost settlement in hundreds of milliseconds.

Machines have no borders and no identity. They don't need complex KYC (Know Your Customer) verification. As long as they have a private key, an Agent running on a server in Singapore can instantly pay an Agent running in Tokyo.

A Status Code That Slept for 30 Years

The最能embody this paradigm shift is a真实history full of metaphor in the internet world.

If you often surf the internet, you must have encountered "404 Not Found." In the initial design of the HTTP protocol, there was actually another status code called 402 Payment Required.

The pioneers of the internet foresaw that the future network would not only need to transmit information but also value. But because there was a lack of a native internet payment layer at the time, this 402 status code was硬生生shelved for 30 years, almost never truly used.

Until 2025, a payment protocol designed specifically for AI Agents was born, its name is x402.

Through the x402 protocol, when an Agent requests data from another server, if payment is required, the server no longer pops up a credit card form for humans to fill out, but directly returns a machine-readable "402 Payment Required" instruction. After receiving the instruction, the Agent instantly calls USDC (a dollar-pegged stablecoin) from its crypto wallet to complete the payment. The entire process ends in hundreds of milliseconds, and the data channel opens.

No registration, no scanning, no password verification. Value flows seamlessly at the bottom layer of the internet, just like data.

Human Money and Machine Money: The Folding of Wealth

According to data from blockchain analytics firms, in the short few months from late 2025 to early 2026, AI Agents have already completed hundreds of millions of payments via stablecoins. Cryptocurrency no longer needs to prove itself more useful than Alipay; it has sunk into the deep sea of the internet, becoming the silently运转blood of millions of machines.

But the story doesn't end here. When machines start to have wallets, start to earn and spend money autonomously, as humans deeply固化by the concepts of "cash" and "bank accounts," how should we understand this new form of wealth? What is the relationship between our money and the machine's money?

In the past, wealth was显性的, physical. You took out a banknote, or opened a banking App watching the balance数字change, you had a切肤之感of "spending money."

But in the future, wealth will be folded.

Imagine you hire an AI Agent to operate a social media account for you. You don't need to pay it a salary; you just need to initially top up its "Agentic Wallet" with 100 USDC (equivalent to $100).

Next, this Agent begins its autonomous狂奔: it pays another data Agent 0.05 USDC to get hot trends; pays a drawing Agent 0.1 USDC to generate an illustration; after the article is published, it automatically deposits the advertising revenue it earned (maybe 0.5 USDC) into its own wallet.

In this process, the machine's money疯狂流转, earns interest, and is consumed at millisecond speed in the underlying network. And you, as the human owner,根本看不到these密密麻麻micro-payment bills. You don't need to understand what x402 is, or what smart contracts are.

The only thing you see is a极简report this Agent sends you weekly: "This week投入$10, net profit $50, profits have been withdrawn to your fiat bank account."

This is the ultimate division of labor between humans and machines regarding wealth: Machines handle friction, humans enjoy the results.

Machine money (cryptocurrency) is for flowing,它是high-frequency, cold, production material pursuing极致efficiency;而human money (fiat) is for feeling, it is for buying coffee, paying rent,承载life security—the final destination.

Cryptocurrency did not eliminate bank accounts; it just pushed complex financial transactions one layer down. While humans enjoy the极致convenience brought by AI on the front end, in those unseen底层, a financial system专属to machines is silently重塑the business rules of this world.

Pertanyaan Terkait

QAccording to the article, why has cryptocurrency failed to become a mainstream payment method for humans?

ACryptocurrency has failed as a mainstream human payment method due to three main obstacles: high price volatility discouraging its use for daily transactions, a user-unfriendly experience involving complex private key management and gas fees, and complicated regulatory and tax implications like capital gains tax on everyday purchases.

QWhat is the core argument about who the real user of cryptocurrency should be?

AThe core argument is that cryptocurrency has been looking for the wrong user for the past decade. Its real user is not humans, but AI Agents. The 'Agent to Agent' (A2A) smart economy is where cryptocurrency finds its purpose, as it is perfectly suited for machine-to-machine transactions.

QHow do the needs of AI Agents make cryptocurrency's disadvantages into advantages?

AFor AI Agents, the disadvantages of cryptocurrency for humans become advantages. Machines don't need customer service and trust only code. They require millisecond-level micropayments that traditional systems can't support due to high fees. They have no nationality and can bypass complex KYC requirements, allowing for instant, borderless transactions between agents using only a private key.

QWhat historical internet protocol code is mentioned, and what new protocol is proposed to finally fulfill its original purpose?

AThe historical HTTP status code '402 Payment Required' is mentioned, which was envisioned for internet value transfer but was unused for 30 years due to a lack of a native payment layer. The new protocol proposed to fulfill this purpose is called 'x402', designed for AI Agents to make instant, automated micropayments.

QWhat is the described 'ultimate division of labor' between human money and machine money?

AThe ultimate division of labor is: 'Machines handle friction, humans enjoy the results.' Machine money (cryptocurrency) is for flow—it is high-frequency, cold, production capital that pursues extreme efficiency in the background. Human money (fiat currency) is for feeling—it is the final destination for buying coffee, paying rent, and carrying a sense of security in daily life.

Bacaan Terkait

Laporan Q1 2026 Ethereum: Biaya Turun, Pengguna dan Volume Transaksi Capai Rekor Tertinggi Sejarah

**Laporan Kuartal I Ethereum 2026: Biaya Turun, Pengguna dan Jumlah Transaksi Capai Rekor Tertinggi** Laporan Ethereum Q1 2026 menunjukkan data yang tampak berlawanan tetapi kritis: jaringan ini mengalami pertumbuhan pengguna, transaksi, dan throughput tertinggi sepanjang masa, sementara biaya transaksi, TVL, volume perdagangan, dan kapitalisasi pasar ETH terdilusi turun. Ini menandai transisi Ethereum ke fase 'biaya rendah untuk skala' setelah peningkatan Fusaka meningkatkan kapasitas data dan membuat ruang blok lebih murah, melepaskan permintaan jaringan (paradoks Jevons). Naratif inti Ethereum bergeser dari blockchain DeFi ke lapisan penyelesaian keuangan global. Ethereum mempertahankan dominasi dalam aset tokenisasi: stablecoin, dana tokenisasi (naik 4.9% QoQ), komoditas tokenisasi (naik 60% QoQ, terutama emas), dan saham tokenisasi. Kehadiran institusi seperti BlackRock, JPMorgan, dan Fidelity semakin memperkuat adopsi. **Data Kunci Q1 2026:** * **Penggunaan (Naik):** Pengguna Bulanan Aktif (MAU): 13.2 juta (+53.5% QoQ). Jumlah Transaksi: 200.4 juta (+38% QoQ). Throughput: 25.78 TPS. * **Nilai & Biaya (Turun):** Biaya Transaksi Lapisan-1: $39.9 juta (-47.9% QoQ). TVL Ekosistem: $316.2B (-11% QoQ). Kapitalisasi Pasar ETH Tercairkan Penuh: $290B (-30.3% QoQ). * **Aset Tokenisasi (Stabil/Tumbuh):** Nilai Pasar: $2034B. Didominasi stablecoin ($1789B), diikuti dana ($194B) dan komoditas ($47B). Ethereum mengorbankan pendapatan biaya jangka pendek untuk ekspansi jaringan, berfokus pada konsolidasi sebagai lapisan penyelesaian default untuk aset keuangan global. Peningkatan kapasitas berkelanjutan (seperti rencana upgrade Glamsterdam) diharapkan dapat lebih mendorong adopsi dan nilai jaringan jangka panjang.

marsbit1j yang lalu

Laporan Q1 2026 Ethereum: Biaya Turun, Pengguna dan Volume Transaksi Capai Rekor Tertinggi Sejarah

marsbit1j yang lalu

Baru Saja Raup Dana Rp 2,7 Triliun, Fei-Fei Li Juga Berinvestasi

Ilmuwan ternama dan mantan peneliti senior DeepMind, Pete Florence, baru saja menggalang dana US$4 miliar (sekitar Rp27 triliun) untuk perusahaannya, Generalist AI. Menariknya, meskipun dikenal sebagai salah satu perintis arsitektur model "dunia nyata" (world model) atau Vision-Language-Action (VLA), Florence secara terbuka menolak label "world model" untuk perusahaannya. Florence, yang dibimbing oleh ahli robotika fisik Rusia Tedrake di MIT, percaya bahwa fokus utama seharusnya pada *tujuan*, bukan sekadar *label*. Tujuannya adalah menciptakan robot yang dapat melakukan tugas fisik apa pun dengan tingkat keberhasilan dan kecepatan tinggi, tanpa memerlukan data spesifik untuk setiap tugas, mirip seperti manusia. Generalist AI telah meluncurkan dua model: GEN-0 (2025) dan GEN-1 (2026). GEN-1 diklaim memiliki tingkat keberhasilan 99% dalam tugas-tugas rumit seperti melipat kardus dan merawat robot, berkat pelatihan dengan data interaksi fisik skala besar yang dikumpulkan melalui sarung tangan mekanis khusus. Kemajuan ini menunjukkan bahwa model fisik mereka mendekati titik balik menuju utilitas komersial, mirip dengan GPT-3 untuk bahasa. Pendanaan putaran ini, yang meningkatkan valuasi perusahaan menjadi US$20 miliar, dipimpin oleh investor seperti NVentures (Nvidia), Bezos Expeditions, NFDG, serta figur ternama termasuk pendiri Xiaomi Bin Lin, pendiri Zoom Eric Yuan, dan ilmuwan terkemuka Fei-Fei Li. Investor percaya pada visi Florence untuk mewujudkan robot serba bisa yang benar-benar berguna dalam kehidupan nyata.

marsbit1j yang lalu

Baru Saja Raup Dana Rp 2,7 Triliun, Fei-Fei Li Juga Berinvestasi

marsbit1j yang lalu

Dua Legenda Hilang dalam Tiga Hari: Bendungan Talenta AI Google, Mulai Jebol?

Dalam tiga hari, Google kehilangan dua tokoh legendaris AI. Noam Shazeer, penulis inti makalah Transformer dan pemimpin bersama Gemini, meninggalkan Google untuk bergabung dengan OpenAI. Dua hari kemudian, John Jumper, pemenang Nobel Kimia 2024 dan pemimpin inti AlphaFold, meninggalkan Google DeepMind untuk bergabung dengan Anthropic. Tren ini diperkuat dengan keputusan mantan anggota pendiri OpenAI, Andrej Karpathy, yang bergabung dengan Anthropic pada Mei. Kehilangan ini mencerminkan tren yang lebih luas: aliran talenta AI puncak dari Google ke OpenAI dan Anthropic. Penyebabnya adalah perbedaan mendasar dalam misi. Bisnis inti Google adalah iklan, sehingga penelitian AI sering kali harus selaras dengan tujuan komersial. Sebaliknya, OpenAI berfokus pada AGI untuk kemanusiaan, sementara Anthropic berfokus pada keamanan AI, memungkinkan peneliti berkonsentrasi penuh pada kemajuan teknologi. Faktor lain termasuk prospek kekayaan dari IPO yang akan datang di OpenAI dan Anthropic, yang menawarkan potensi imbalan ekuitas yang jauh lebih besar dibandingkan dengan raksasa matang seperti Google. Selain itu, penggabungan Google Brain dan DeepMind pada 2023 dianggap gagal menyelesaikan ketegangan antara penelitian jangka panjang dan tekanan komersialisasi. Google masih memiliki aset kuat seperti infrastruktur komputasi dan data, tetapi kehilangan talenta kritis ini merupakan tantangan struktural yang mendalam. Di bidang AI, retensi talenta terbaik ternyata lebih sulit daripada membangun model yang paling canggih sekalipun.

marsbit3j yang lalu

Dua Legenda Hilang dalam Tiga Hari: Bendungan Talenta AI Google, Mulai Jebol?

marsbit3j yang lalu

Di Balik 'Raport' AI, Tersembunyi Seorang 'Pembuat Soal' Tionghoa

Setiap kali model AI terdepan dirilis, industri melihat "laporan nilai" seperti MMLU-Pro, MMMU, dan MMMU-Pro. Tolok ukur ini telah menjadi bahasa umum untuk mengevaluasi kemampuan model. Di baliknya adalah nama seorang peneliti Tionghoa, Chen Wenhu, asisten profesor di University of Waterloo. Dia dan lab TIGERLab-nya menciptakan MMLU-Pro karena MMLU lama tidak lagi efektif—model canggih seperti OpenAI o3 hampir mencapai nilai sempurna. MMLU-Pro, dengan 12.032 soal lebih sulit dan 10 pilihan jawaban, berhasil membedakan kembali kemampuan model. Selain itu, mereka mengembangkan MMMU untuk mengevaluasi model multimodal (teks dan gambar) pada 11.500 soal dari berbagai disiplin ilmu. Bahkan model terkuat seperti GPT-4V hanya mencapai akurasi 56%. MMMU-Pro kemudian dibuat agar model tidak bisa mengandalkan teks saja dan harus benar-benar memahami informasi visual. Chen Wenhu memiliki latar belakang riset dalam pemahaman informasi kompleks. Pengalamannya di Google DeepMind untuk proyek Gemini membantunya memahami celah dalam evaluasi. Labnya juga mengerjakan penelitian model, seperti UniVideo untuk video dan MoCha untuk karakter virtual, yang memperdalam pemahaman mereka dalam merancang tolok ukur yang solid. Kini, dia bergabung dengan Meta untuk fokus pada data pelatihan dan evaluasi multimodal. Karyanya menggarisbawahi kontribusi signifikan peneliti Tionghoa di balik layar dalam membentuk standar evaluasi AI global.

marsbit3j yang lalu

Di Balik 'Raport' AI, Tersembunyi Seorang 'Pembuat Soal' Tionghoa

marsbit3j yang lalu

Trading

Spot
Futures
活动图片