Alliance: AI Agents Are the Missing Piece of Web3

Original Author: Imran Khan, Co-founder of Alliance

Compiled by: Jiahua, ChainCatcher

I believe Chris Dixon's judgment is correct. Cryptographic technology, or what people used to call Web3, has filled the economic and ownership layer of the internet.

Marc Andreessen hoped to embed payment capabilities directly into browsers during the Netscape era. The early web even reserved the HTTP 402 status code, which means "Payment Required"; Coinbase's x402 is named after this.

This name also pays tribute to the pioneers who repeatedly tried to embed financial capabilities into the internet infrastructure in the early years. Thanks to Andreessen, who fought this battle for us back in those days.

Netscape once collaborated with Visa to try to integrate payment functions directly into the browser, but it never truly materialized. While Netscape laid an important foundation for modern internet commerce through SSL, currency itself was left outside the early internet infrastructure.

Later, people gradually realized that commerce would inevitably migrate to the internet, but the means of online payment would still be credit cards and payment platforms like PayPal and Stripe, rather than an internet-native currency system.

Returning to the early days of the crypto industry, concepts such as user-owned assets, permissionless markets, tokenization, decentralized identity, open social graphs, cross-platform transferable reputation, decentralized storage, and programmable finance all hinted at the possibility of redesigning the internet.

A group of early Ethereum community participants proposed that we could reconstruct parts of the internet's structure, giving users greater autonomy on the economic level. This was also one of the reasons I entered the crypto industry.

However, over time, this original intention has gradually been diluted due to multiple factors such as malefactors, security threats, and policy restrictions.

It is in this wave that Chris Dixon proposed his core judgment: decentralization and ownership are the true core of the crypto industry.

His logic is that centralized platforms hold the power of commission and control. Over time, the incentive mechanisms of these platforms will gradually tilt towards the platforms themselves and their shareholders, rather than the product users.

These platforms control distribution, identity, payments, APIs, data, and social relationships, even deciding how economic benefits are distributed between users and developers.

Cryptographic technology offers another possibility: allowing key elements like identity, payments, and data to be carried by open protocols, no longer controlled by a single company; enabling all participants to share in the economic benefits while truly owning their data and assets.

But it turns out that most people do not care whether a platform is centralized. They care more about convenience and whether they can get what they want in the simplest way.

As long as the product is good enough, people do not mind giving their data to companies like Google, Facebook, Apple, Amazon, and OpenAI.

People like me, who are crypto-punks, care deeply about privacy, anti-censorship, decentralization, and personal sovereignty. But most people are not willing to endure a significantly worse user experience for these principles.

Most people wake up every day without first thinking about whether they own a social graph. They just want the product to work well and provide entertainment.

After years of practice, reality has not developed completely along the path Chris Dixon initially envisioned. Nevertheless, I still believe his direction is correct; it just lacks the most crucial piece of the puzzle: the rise of AI.

AI and cryptographic technology are highly complementary. AI needs an open economic infrastructure, and cryptographic technology needs a new large-scale demand side. The combination of the two fills in the previously missing key link: demand.

AI Agents Are the True Demand Side of Open Infrastructure

AI Agents are becoming a new type of participant in the internet. Cloudflare disclosed in 2026 that automated traffic has surpassed human activity for the first time, accounting for about 57% of all web requests. I believe a significant portion of this is related to AI Agents.

Moreover, this trend will continue to accelerate. In the future, most of our interactions with the internet may no longer wait for humans to issue commands one by one but will be initiated proactively by AI Agents.

However, to become a "first-class citizen" on the internet, AI Agents must be able to access thousands of services, datasets, and markets without permission and interact directly with counterparties and other Agents without having to establish customized relationships with each party.

This is not about decentralization for the sake of decentralization, nor is it about retracing old paths. It is simply that when the scale of collaboration expands from the human level to the machine level, portability, open access, shared states, and interoperability will generate direct economic value.

What they need is a set of open infrastructure that can be used anytime without permission. They need to establish identities on protocols like Handshake and require programmable currencies and asset ownership.

In addition, they may also need persistent storage, transferable identities, open social and reputation graphs, verifiable information, permissionless markets, and programmable contracts, as well as open networks or markets for mutual discovery and communication, such as Moltbook, Farcaster, or some entirely new form.

Therefore, my judgment is that most of the concepts Chris Dixon proposed in "Read Write Own" are directionally correct; it’s just that the target users are not humans but AI Agents.

The Next Generation of the Internet Is More Than Just Cryptographic Technology

The original internet, or Web1, was built on open protocols. Information could be accessed freely, anyone could publish and obtain content, websites could link freely, and developers could build products without needing permission. At that time, users were mostly just readers of information.

The subsequent arrival of Web2 turned these readers into creators. Social networks, smartphones, broadband, cloud services, online markets, and large platforms emerged one after another. Billions of people began to create and publish content, moving their businesses online and connecting more closely with each other.

However, platforms also controlled all key aspects, including identity, payments, social relationships, data, distribution, and the economic connections between people.

So, what will the next step be?

I do not believe that Web3, or what I call the "next generation of the internet," is merely another term for cryptographic technology.

I believe the next generation of the internet will be composed of cryptographic technology, AI, robotics, energy, computing power, and ultimately biotechnology, intertwined in some way.

AI will become the layer of intelligence and interaction; cryptographic technology will become the layer of economy and ownership; robotics will connect the intelligent layer to the physical world; computing power and energy will become the resource layer that supports everything; and biotechnology will increasingly resemble programmable systems, with Neuralink representing one direction of this.

Take decentralized identity, reputation, and social graphs as an example. In the future, AI Agents may need Moltbook, Farcaster, or some new open platform. They can provide services on these platforms, complete tasks, establish reputations based on actual delivery, form persistent identities on the internet, and establish relationships with other Agents.

Decentralized storage is similar. Humans usually do not mind putting files on Google Drive, iCloud, or other cloud services. However, AI Agents may require greater autonomy and control, and need to ensure that information exists long-term or has redundant backups, so they may use both decentralized and centralized storage services simultaneously.

For example, IPFS and Filecoin can store content, allowing another Agent to verify whether the retrieved content is identical to the original content.

I am not saying that everything will be decentralized. What I mean is that some of the underlying infrastructure may indeed be useful and can work in conjunction with centralized services like AWS and GCP.

Looking at ownership, I believe that ownership may be even more important for AI Agents than for humans. If the entire world is tokenized in some form in the future, the available assets will be numerous and highly dispersed, making it difficult for humans to quickly understand and configure these assets.

Currently, AI Agents are still trapped in closed ecosystems controlled by individual MCP services or applications. This does not mean that similar Agents cannot be built using MCP, OAuth, A2A, Stripe, Visa, AWS, and traditional databases.

But with on-chain infrastructure, more assets and markets in the financial world will be able to flow across platforms, be open to the global market, and be directly read by machines. AI Agents can directly access assets and markets and make financial decisions based on human-set goals and risk boundaries.

Therefore, I believe Dixon accurately pointed out the layer that the internet is missing: native ownership.

However, the next generation of the internet is not a single technology but the result of the interplay of on-chain infrastructure, robotics, AI, biotechnology, computing power, and energy, just as Web2 was the product of smartphones, high-bandwidth networks, centralized applications, and cloud storage.

The truly missing piece in this new architecture is the AI Agent, not humans.