比特币橙子Trader
比特币橙子Trader|4月 25, 2026 12:00
A16z in-depth article: Why is it necessary to use blockchain to "license" the explosion of AI agents? 5 hardcore logics to understand future infrastructure In the current AI industry, agents are extremely popular. AI is no longer just a tool person that can only chat with you. They are becoming true 'workers', starting to write code, do transactions, and run processes on their own. But as these AI employees become more and more numerous, an extremely fatal problem arises: they are a 'black house'. In the real world, a person needs an ID card, bank card, and labor contract to work and earn money. But currently, AI agents are unable to prove to the system who they are or what they are authorized to do, and there are no bank accounts that can legally receive and make payments. They run naked outside the financial system, and all collaborations are isolated islands. A16z pointed out in the latest long article that the most lacking underlying infrastructure in the era of AI agents is blockchain. This is by no means trying to ride on the heat. The encrypted identity, programmable payments, and tamper proof ledger provided by blockchain are becoming the last piece of the puzzle for AI to enter mainstream commerce. 1. Identity: Issuing a "Digital Identity Card" (KYA) to non-human individuals In the financial system, the proportion of "non-human" entities such as automated trading and risk control engines is already more than 100 times that of humans. But so far, these AI cannot prove their identity across platforms at all. The merchant's firewall will directly block them outside the door, because the merchant does not know who is behind this AI and who will be held responsible if something happens. So, we need a KYC (Know Your Agent) system similar to human KYC (Know Your Customer). The wallet and encrypted signature of blockchain can give the agent a portable and cross platform verifiable independent identity. With this identity, the Agent can openly represent you to interact with other platforms. 2. Governance: If AI controls core resources, who will control AI? Imagine in the future, the allocation of funds or supply chain management in a community will be handled by several powerful AI systems. If the underlying models of these AI are monopolized by a big company (such as OpenAI), then as long as the big company secretly changes the parameters, the entire community's operating rules will be completely overturned. To prevent this kind of 'big factory dictatorship', we need blockchain to enforce constraints. If collective decisions are recorded on the chain, AI must strictly follow these immutable instructions to execute them. Blockchain ensures that AI behavior is transparent, traceable, and cannot be secretly manipulated by large companies. 3. Payment: Open a "plug and play" stablecoin account for AI How do AI do business with each other? They do not have front-end pages, do not require customer service, only back-end code and interfaces. When a sales AI needs to buy data, it cannot fill in credit card information like a human. Traditional payment systems are simply unable to serve pure code transactions without physical entities or websites. And stablecoins and smart contracts on the blockchain perfectly meet this demand. For example, the protocol launched by Coinbase allows AI to directly include a few cents of stablecoin in the code request every time it calls data, completing settlement instantly. This extreme "machine to machine" micro payment can only run on the encrypted track. 4. Trust reassessment: When intelligence becomes extremely cheap, 'verification' becomes sky high value When the cost of AI generating code, graphics, and even trading instructions approaches zero, the biggest problem facing humanity is: who will review these things? Keeping 'humans on the review loop' is becoming physically impossible because the output speed of AI is too fast. If we completely delegate power, unverified AI could potentially result in billions of dollars in losses due to a small deviation at any time. In the future Agent economy, the most expensive thing is not computing power, but verifiable sources and proof of trust. The public ledger of blockchain can clearly record everything AI does. It is no longer an uncontrollable black box, but a transparent system with evidence and accountability in case of any incidents. 5. Control: Put AI in the cage of 'smart contracts' The biggest risk when we entrust increasingly complex tasks to AI is that it may "take its own initiative". A vague instruction may cause it to clear your account. The native tools of blockchain, such as intent based architecture and scope delegation framework, can perfectly solve this problem. You can tightly lock the permission boundaries of AI at the level of smart contracts. You can specify that it can only use a limit of 100 U or can only interact between two protocols.
+6
Mentioned
Share To

Timeline

HotFlash

APP

X

Telegram

Facebook

Reddit

CopyLink

Hot Reads