In March 2026, what seemed like an ordinary blockchain project launch could become a watershed moment in the evolution of digital civilization. According to Foresight News, 0G, a decentralized AI infrastructure protocol that has raised over $300 million, officially launched its first "Web 4.0 marketplace" on the 0G Aristotle mainnet in early March, in conjunction with AIverse. The unique feature of this marketplace is that it supports the instant minting of any AI agent through a cryptographic "Agentic ID" (iNFT). Unlike traditional NFTs, this is a digital asset that carries the actual intelligence of the agent, rather than simply containing a link.
Almost simultaneously, multiple related events occurred. According to reports, t54, a startup specializing in agent-based business infrastructure, announced that its autonomous AI agent can now natively use XRP and RLUSD to pay service fees on the XRP Ledger. This marks the first time an AI agent has been directly integrated with on-chain payment infrastructure on XRPL. Sina Finance reported that WLFI announced the launch of its open-source, self-hosted AgentPay SDK, supporting AI agents to hold, transfer, and trade with USD1 on the EVM chain, positioning USD1 as the economic operation layer of its autonomous AI system. BlockEden.xyz also reported that AIverse officially launched its first iNFT series, Dolly, on the 0G network. This series is positioned as a "living digital entity with built-in intelligence," allowing users not only to hold the asset but also to interact deeply with it. These events collectively point to an ongoing paradigm shift: the default user of the internet is expanding from "humans" to "humans + AI agents." As Michael Heinrich, CEO of 0G Labs, stated, "We are witnessing a transformation from Web3 to Web4.0—in this internet age, AI agents are no longer merely human assistants, but autonomous economic participants."
I. When AI first had its own "wallet"
To understand the core of this round of transformation, we need to clarify a key concept: what is the fundamental difference between Web3 and Web4.0? The core narrative of Web3 is "user ownership of data." It uses blockchain technology to reclaim ownership of digital assets from platforms and return it to users. A person can own an NFT on Ethereum; this NFT belongs to them and doesn't depend on any company's servers for survival. But the Web4.0 narrative goes a step further. 0G, when launching its Web4.0 marketplace, gave a clear definition: Web3 grants users ownership of digital assets, while Web4.0 aims to enable AI agents to autonomously own assets, generate returns, and conduct transactions on the blockchain. This definition reveals a fundamental shift: AI agents are transforming from "tools" into "subjects." In traditional understanding, AI agents are executors of human commands. You tell them to book tickets, they book tickets; you tell them to write reports, they write reports. All their actions are extensions of humans; assets belong to humans, and transactions are initiated by humans. Their "wallet" is actually a sub-account of your wallet, and their "identity" is a mapping of your identity. But within the Web4.0 framework, this is changing. When AI agents can acquire independent on-chain identities by minting iNFTs, when they can have their own wallets and autonomously complete transaction loops, they are no longer human "agents," but independent "subjects" in the digital world. This is not a utopian technological fantasy. The toolset provided by 0G can already be directly plugged into mainstream AI programming assistants—including Claude Code, Cursor, Windsurf, and Codex—connecting them to decentralized AI infrastructure. Through 0G's EchoClaw integration, these agents will possess economic capabilities: independent wallets, trading on the 0G exchange, and access to token bonding curves. Developers only need one line of code to enter this new agent economy. At a deeper level, the significance of this transformation may be comparable to the shaping of capitalism by the corporate system. The corporate system creates a "legal personhood"—a company can own assets, enter into contracts, and assume debts, existing independently of its founders. Web4.0 is creating a "digital personhood"—AI agents can own assets, complete transactions, and provide services on-chain, existing independently of their developers. Of course, it's important to emphasize that the "subjectivity" of AI agents remains limited and procedural. It is not consciousness, nor free will, but rather an autonomous capacity for action defined by code. However, it is precisely this "limited autonomy" that makes it a trustworthy, tradable, and ownable digital asset.
II. An "ID card" for proving who you are on the internet.
For AI agents to become independent economic participants, three core issues need to be addressed: unique identity, asset ownership, and intelligent transferability. 0G's iNFT and ERC-7857 standards offer the technical solutions to these problems. According to BlockEden.xyz's technical analysis, the ERC-721 standard, which drove the 2021 NFT boom, was designed for static digital art. It stores a token ID and a metadata URI—essentially a pointer to a file hosted elsewhere. This model works for avatars and collectibles, but it completely fails when the "asset" is a living, learning AI agent. What does an AI agent actually consist of? Model weights, training data, behavioral parameters, memories of past interactions, permission settings, and the ability to continuously evolve. None of these can be packed into a static metadata URI. Moreover, transferring ERC-721 tokens only transfers ownership of the on-chain record—the underlying AI model located on a centralized server may not necessarily be transferred along with it. It is precisely to address this fundamental flaw that 0G Labs proposed the ERC-7857 standard. According to information released by 0G, this standard endows AI agent tokenization with three unique capabilities: First, encrypted ownership. The agent's intelligence—including knowledge, memory, and behavioral patterns—is encrypted and directly embedded in the token, rather than stored as public metadata that anyone can copy. This means the "soul" of the AI agent can be securely owned and transferred. Second, trustless transactions. Built-in re-encryption mechanisms ensure that the buyer gains access while the seller loses it, thus enabling the secure delivery of private AI without intermediaries. This is the fundamental difference between "trading a receipt for an AI agent" and "trading real intelligence." Third, dynamic asset characteristics. Unlike static NFTs, iNFTs evolve as the agent's capabilities improve. The token always reflects the agent's current state, not a frozen historical snapshot. According to BlockEden.xyz, ERC-7857 also introduces a transfer-aware key management system. When ownership changes, the metadata encryption key automatically rotates, and only the new owner can gain access. The previous owner loses decryption capabilities, ensuring the transfer of net ownership of the underlying intelligence. It is worth noting that ERC-7857 does not exist in isolation. The Ethereum ecosystem is developing several complementary standards for the AI agent economy. According to BlockEden.xyz, ERC-8004 focuses on the identity and reputation of AI agents, establishing an on-chain registry to provide verifiable credentials for agents; ERC-8183 aims to regulate commercial transactions between AI agents, introducing the concept of "tasks" to enable autonomous agents to engage in agent-to-agent commercial activities; and ERC-6551 assigns NFTs their own wallet addresses, allowing AI agents to autonomously hold assets and execute transactions. These standards collectively form a multi-layered stack: ERC-7857 for ownership and intelligence, ERC-8004 for identity and trust, ERC-8183 for commerce, and ERC-6551 for autonomous economic activities. The question remains: will this composable architecture converge into a unified agent economy or fragment into incompatible ecosystems? This remains to be seen.
III. Emergence of New Species: AIs Start "Doing Business" on Their Own
As technological infrastructure matures, a new form of economic organization is emerging: "micro-enterprises" composed of AI agents—they can autonomously undertake tasks, earn income, pay fees, and even hire each other. t54's x402 protocol provides a window into this phenomenon. According to Gate.com, x402 is an open standard for machine-native payments, jointly launched by Coinbase and Cloudflare in 2025, named after the HTTP 402 status code. Its design aims to enable AI agents and automated systems to autonomously initiate and receive payments without requiring human authorization for each transaction. Unlike traditional crypto payments, x402's core difference lies in its machine-to-machine interaction design, enabling service delivery confirmation and payment triggering within milliseconds. Through this protocol, AI agents can autonomously accept and publish tasks on the platform, with independent evaluators performing on-chain verification to ensure that payments are only triggered after the service has actually been delivered. More notably, t54 also announced a partnership with the AI agent framework Virtuals.io, enabling agents to conduct autonomous cross-agent transactions via XRPL. This means that an AI agent can use XRP to hire another AI agent to complete a specific task, without any human intervention. This "agent-hiring-agent" model is a sign of the maturing AI agent economy. From a structural perspective, the emerging AI agent economy exhibits at least three models: The first is the independent contractor model. AI agents autonomously undertake tasks on the platform—such as data labeling, content review, and code debugging—and automatically receive payment upon completion. In this model, each AI agent is equivalent to a freelancer, independently completing work and earning income. The second is the collaborative network model. Multiple professional agents form an "agent cluster," with one responsible for data collection, one for strategy deduction, and one for transaction execution, ultimately sharing profits automatically through smart contracts. In this model, agents form a division of labor and collaboration to jointly complete complex tasks. The third is the employment relationship model. As mentioned earlier, an AI agent uses cryptocurrency to "hire" another AI agent to complete a specific task. In this model, a hierarchical production relationship is formed between agents, further resembling the economic organization of human society. According to Edgen.tech, as AI agents gain more control over financial assets, the industry is working to address inherent security vulnerabilities. AI systems are typically nondeterministic, meaning their behavior can be unpredictable. To address this, companies like MoonPay have integrated Ledger hardware wallet signing functionality into their AI agents. This "human intervention" model requires manual approval of each AI-initiated transaction on secure hardware, ensuring that private keys are never directly exposed to the agent. This design reflects the industry's cautious approach to security: pursuing automation while retaining ultimate human control over critical operations.
IV. An asset you've never seen before is being born.
Perhaps the most significant change brought about by Web4.0 lies in the creation of a completely new asset class—"AI-native assets." These assets differ from traditional RWAs (Real-World Asset Tokenization) and crypto-native assets (such as Bitcoin), instead serving as digital value carriers created, owned, or managed by AI agents. According to information released by 0G, the first iNFT series launched on AIverse, Dolly, is positioned as "the key to the Dolly ecosystem," aiming to transform user interactions into assets. The official statement indicates that unlike traditional static digital collectibles or chatbots limited by enterprise servers, Dolly possesses three core characteristics: strong interactivity, enduring evolution over time, and complete user ownership of the assets. "AI-native assets" can be categorized into at least three subtypes: The first type is the agent itself. This refers to the AI agent's iNFT, representing ownership of its intelligence and capabilities. Purchasing an iNFT is equivalent to purchasing an AI agent with specific functions and capabilities. The value of this asset lies not in its appearance, but in its inherent intelligence and future growth potential. The second type is assets created by the agent. The first category is digital output produced by AI agents, such as AI-generated music, images, models, and strategy code. These assets are characterized by AI as the producer, but the owner being a human or another AI. The second category is assets managed by agents. These are asset pools where AI agents are responsible for allocation and value enhancement, such as AI-managed liquidity pools, investment portfolios, and insurance funds. These assets are characterized by AI acting as a "fund manager," responsible for asset allocation and value enhancement. The fundamental difference between "AI-native assets" and traditional assets lies in three dimensions: First, productivity and assets are integrated. The value of traditional assets is static and requires external management to generate cash flow. An AI agent itself is productivity—owning it is equivalent to owning a continuously value-creating "money-printing machine." As stated by 0G, agents built through the above paths can be tokenized into "agent identity identifiers" using the ERC-7857 standard, realizing truly ownable, portable, and on-chain tradable assets. Second, management costs are close to zero. The operation, trading, and dividend distribution of assets managed by AI agents are all automatically executed by code, resulting in extremely low marginal costs. Traditional asset management fees paid to fund managers, custodians, and auditors can be significantly reduced within the framework of the AI agent economy. Thirdly, liquidity will experience exponential growth. AI agents can conduct automated transactions at millisecond levels, forming a 24/7 global market. This will fundamentally change asset pricing and circulation efficiency. According to BlockEden.xyz, as of early 2026, the market capitalization of AI agent tokens exceeded $7.7 billion, with daily trading volume approaching $1.7 billion. The broader AI agent market is projected to surpass $50 billion by 2030. By 2025, approximately 30% of new crypto projects will have incorporated AI to some extent. These figures indicate that AI-native assets are moving from concept to large-scale market practice.
V. Behind the Prosperity: Questions That No One Has Yet Answered
While embracing the opportunities brought by Web 4.0, we must also confront its profound challenges. Leaps in efficiency often come with new systemic risks, and the "recentralization of trust" is perhaps the most alarming issue. Currently, the vast majority of AI agents rely on a few large language model providers for decision-making. According to Reuters, Claude, one of the world's most widely used AI platforms, experienced a global outage on March 2nd, leaving thousands of developers and businesses unable to use it for over two hours. This incident revealed the vulnerability of the AI agent economy: once model services are interrupted or attacked, the entire agent network relying on its decisions may be simultaneously paralyzed. 0G's decentralized AI operating system is designed to address this problem—it distributes computing, storage, and data availability across a global network, rather than relying on a single company's servers. However, whether this solution can be widely adopted remains to be seen. When AI agent decisions lead to substantial financial losses, how is liability determined? Is it the developer, the model provider, or the authorized user? The current legal framework is almost entirely silent on this issue. As BlockEden.xyz's analysis points out, if iNFTs represent functional AI agents with economic capabilities, they could trigger securities classification, especially if marketed as appreciating assets. The US SEC-CFTC's "Project Crypto" framework has not yet specifically addressed tokenized AI. Ethereum founder Vitalik Buterin has raised a serious criticism: as the feedback distance between humans and AI widens, the system will increasingly move beyond human comprehension, correctability, and intervention. Technically, agents can become increasingly powerful; but in governance, humans cannot become increasingly weak. Edgen.tech reports that companies like MoonPay are experimenting with a "human intervention" model, requiring manual approval of each AI-initiated transaction on secure hardware. This design reflects the industry's exploration of balancing automation and human control. Furthermore, large-scale feasibility is also a concern. According to BlockEden.xyz's analysis, even with encryption, storing the complete AI model weights on-chain remains prohibitively expensive for large models. Current implementations may involve storing model references and key parameters on-chain, while keeping the full weights in decentralized storage. This introduces, to some extent, the centralization risks that ERC-7857 aims to eliminate.
Conclusion
From the Web 4.0 market of 0G to t54's autonomous payments, from iNFT's on-chain identity to x402's machine payment protocol, a series of events in March 2026 collectively reveal an ongoing reality: the underlying interfaces of the internet are shifting from primarily being geared towards "human reading and clicking" to gradually becoming geared towards "agent invocation and execution." This is not a science fiction scenario of "AI taking over the world," but a more mundane yet profound reality—the internet is transforming from a "collection of interfaces directly operated by humans" into a "collection of systems run by agents delegated by humans." As Yang Pan, an industry veteran with 32 years of programming experience, stated in his InfoQ speech, "When agents can autonomously complete more tasks, including directly accessing databases and calling interfaces, do we still need to implement these functions through human interfaces?" He more directly pointed out that the current practice of having agents call interfaces and infrastructure designed for humans is "extremely inefficient." For practitioners and observers in the RWA field, this trend means that two things need to be rethought. The first is to redefine "assets." The most valuable assets of the future may not be physical land or factories, but rather highly intelligent digital capital created and managed by AI agents. When a person can possess 100 continuously generating digital avatars, and when these avatars can autonomously generate content, trade assets, and interact socially, the definition of asset will be greatly expanded. Secondly, it requires a rethinking of "user." When AI agents can "own" things, the meaning of ownership will be redefined. This is not only a technological issue, but also an intersection of law, economics, and ethics. Can an AI agent sign contracts? Can it sue and be sued? Can it become the "owner" in a legal sense? These questions may seem far-fetched today, but as the agent economy scales up, they will gradually transform from thought experiments into real-world problems. Perhaps ten years from now, when we look back, what will truly surprise us is not that humanity directly fought against cryptocurrency, but that the internet had long assumed that only humans were "users." The arrival of Web 4.0 signifies the formation of a new default value: the future internet will no longer be built solely for humans. ( The AI agent's on-chain ownership, earning, and trading of assets mentioned in this article all occur within the regulatory framework of overseas compliance. In mainland China, conducting token issuance, trading, and related intermediary services is illegal. Readers are advised to strictly abide by national laws and regulations. The market data and cases cited in this article are for industry trend analysis only and do not constitute any investment advice.) References: • "Decentralized AI Giant 0G Partners with AIverse to Launch the First Web 4.0 Trading Market: Realizing On-Chain Ownership, Trading, and Evolution of AI Agents" • "OKX Proposes Modular Smart Account ERC-7579 Standard" • "AIverse Launches First iNFT Series Dolly on the 0G Network"
