Vitalik Talks AI Agent: Building Trust with ERC-8004, Suggested Applications for Real-time Translation and More

Ethereum co-founder Vitalik Buterin and Davide Crapis, head of the dAI team at the Ethereum Foundation, deeply explored the infrastructure, trust models, and privacy protection of the AI agent economy at Devconnect Trustless Agent Day, detailing the x402, ERC-8004 standards, and future computational visions. This article is sourced from a piece by ChainFeeds, organized, compiled, and written by Foresight News. (Background: Vitalik questions X platform’s privacy policy: Mandatory disclosure of user country/region has harmed the Anonymity of the crypto community) (Context: Vitalik's Shanghai speech: The history of Cryptography and the significance of ZK) During Ethereum Devconnect, an event called Trustless Agent Day gathered leading thinkers at the intersection of Web3 and AI. This concluding panel was hosted by Tina from Flashbots, with core guests being Ethereum co-founder Vitalik Buterin and Davide Crapis, head of the dAI team at the Ethereum Foundation. This dialogue not only concerns technical standards but also serves as a simulation of the future structure of the digital society: What kind of infrastructure, trust model, and privacy protection do we need when AI agents become the primary participants in economic activities? The dual walls of infrastructure: Payment and discovery (x402 and ERC-8004) The conversation began with two core protocols: x402 for payment and ERC-8004 for service discovery, both of which constitute the foundation of the agent economy. Vitalik's vision for the reconstruction of micropayments Vitalik first expressed his excitement about micropayments in the age of AI. He believes that AI's involvement makes micropayments truly feasible. In the human world, judging 'whether to pay 4 cents or 11 cents for this service' is not only time-consuming but also mentally burdensome, while for AI agents, this is a millisecond-level decision. Vitalik emphasized that 'Pay for what you consume' is the most efficient economic model. However, he also pointed out that such high-frequency payments must be built on privacy protection. Without protection, an agent's thousands of query records would completely expose user behavior patterns. Thus, integrating ZK (zk-SNARKs) technology is critical, for example, users can prepay a sum (like $5) in exchange for a voucher for 5000 queries, while on-chain, these 5000 queries are unlinkable. Davide and ERC-8004: From payment to trust If x402 addresses the question of 'how to pay,' Davide's ERC-8004 attempts to solve the question of 'who to pay.' Davide stated that when he saw people starting to send micropayments to web services or AI via x402, a fundamental question arose: How do you trust these services? ERC-8004 (Trustless Agent standard) was thus born. It is not a simple allowlist but a decentralized service discovery mechanism. It allows service providers to register and showcase their capabilities on-chain. Davide categorized trust into two types: * Soft Trust: Based on past performance, reputation accumulation, and audit results. * Hard Trust: Based on cryptographic proofs or guarantees from cryptoeconomics. ERC-8004 standardizes the interaction format of this information, allowing agents to autonomously search for and verify service providers in a decentralized network. The elephant in the room: The gap between ideals and reality Before discussing the future, host Tina handed the microphone to the audience to initiate a discussion on 'finding the elephant in the room,' meaning those industry pain points that are obvious yet overlooked. The 'role-playing' crisis of agents Developer Shaw raised a sharp point: We do not yet have truly usable agents. He pointed out that current agents are mostly trained on textual data from sources like Reddit; they know 'the theoretical steps to make a cake,' but have never 'baked a cake' in the real world. Current agents attempting to trade or navigate prediction markets are operating 'out of distribution.' To some extent, the current industry is conducting an expensive LARP (live-action role-playing), lacking agents that have true end-to-end execution capabilities. The dual impact of costs and biases Another developer, Tim, pointed out the unsustainability of economic accounts: The reasoning costs are too high. Every tiny decision invocation is burning funding, and to realize the vision of x402, the cost of a single decision must be reduced to below 10% of transaction fees. Currently, many startups survive solely on the free tiers provided by cloud service providers. Additionally, Andrew Miller poured cold water on reputation systems, arguing that historically proven reputation systems tend to favor incumbents and are prone to failure. He proposed that the only solution might be to utilize TEE (Trusted Execution Environment) sandboxes, allowing open-source agents to assess the security of those closed-source agents. Why blockchain? The native habitat for agents Given the numerous issues, why do we insist on building the agent economy on the blockchain? Vitalik and Davide provided answers that go beyond 'payment tools.' On-chain games and synthetic assets Vitalik offered an interesting perspective: The blockchain is the natural soil for on-chain games, where 'games' refer to market interactions in the sense of game theory. He believes that agents do not need to establish trust through identification like humans; they are more suited to game in an anonymous, trustless environment. More importantly, agents can understand and handle extremely complex synthetic assets, which are financial products composed of a basket of commodities that are difficult for humans to intuitively understand but make logical sense to machines. This could give rise to an agent-exclusive market that is entirely different from human financial markets. Constrained delegation Davide added from a security perspective that blockchain provides 'hard rules.' As humans delegate more decision-making power to AI (i.e., agentification), we need a safety valve. Smart contracts can achieve constrained delegation; for example, I allow my DeFi agent to mobilize funds for arbitrage, but the underlying smart contract code explicitly states 'prohibit withdrawals' to external addresses. This code-based constraint provides a level of security that traditional Web2 APIs cannot offer. Privacy as a 'hygiene habit' In terms of privacy…