IOSG｜Decentralized AI: Ethereum's Next Decade Bet

Author|Jiawei @IOSG

Let's Start With a Story

Imagine a day in 2027.

You wake up in the morning and tell your AI assistant: "Book me a flight to Tokyo for next week, budget under 5000, window seat."

Then you go to wash up.

Your AI assistant gets to work. It needs to:

• Find an AI agent good at comparing flight prices
• Confirm this agent is reliable and not a scam
• Have it search across major platforms
• Compare results and automatically make the payment
• Complete the booking

By the time you finish brushing your teeth, the ticket is booked.

This sounds like science fiction, right? But technically, we're very close. The only problem is—

How does your AI know which AI to trust?

The "Trust Crisis" of AI Agents

AI Agents are no longer a new concept. They can browse the web, write code, manage schedules, and even help you trade stocks.

But one problem remains unsolved: these AI agents are "silos."

• OpenAI's agents can only play within OpenAI's ecosystem
• Google's agents only follow Google's rules
• AI from different companies don't talk to each other because they simply don't know who the other is

This is like the early internet, where users of each email service could only send emails to their own people. Hotmail users couldn't send emails to Yahoo users. Crazy, right?

But that's exactly how the world of AI agents is right now.

The Emergence of Two Protocols

Tech giants have realized this problem.

Google launched the A2A protocol (Agent-to-Agent), enabling different AI agents to "talk." It's like giving AIs a common language. In June this year, Google even donated it to the Linux Foundation, indicating they want this to become an open standard.

Anthropic launched the MCP protocol (Model Context Protocol), allowing AI agents to connect to various tools and data sources.

These two protocols solve the "communication" problem.

But there's a more fundamental problem unsolved:

How do you find a reliable AI agent? How do you know if it performs well?

A2A allows AIs to talk, but it doesn't tell you who to talk to.

It's like having a telephone, but no phone book.

ERC-8004: Issuing "Passports" and "Credit Scores" for AI Agents

This is the problem ERC-8004 aims to solve.

Simply put, it provides AI agents with three things:

1. Identity

Each AI agent registers on Ethereum, obtaining a unique ID. This ID is actually an NFT—yes, that NFT. This means:

• Your AI agent has a chain-verifiable identity
• The identity can be transferred (sell your AI agent?)
• No one can forge or tamper with it

2. Reputation

People who have used the AI agent can rate it. These ratings are recorded on-chain, and anyone can check them. It's like:

• Uber driver star ratings
• Taobao store credit ratings
• But it's on the blockchain, no fake reviews or deleted comments

3. Validation

For high-risk tasks (e.g., financial transactions), ratings alone aren't enough. ERC-8004 supports independent third-party validation:

• Someone stakes funds to re-run the task and verify if the result is correct
• Use cryptographic proofs (ZK proof) to verify the AI isn't lying
• Use Trusted Execution Environments (TEE) to ensure the computation process isn't tampered with

The higher the risk, the stricter the validation.

Ordering pizza? Just check the rating.

Managing your investment portfolio? Requires cryptographic proof.

Wait, Why Ethereum?

Good question.

The AI agent economy is worth trillions. Google, Microsoft, OpenAI all want to own this cake. Why build this "trust layer" on Ethereum?

The answer is: Neutrality.

Think about it, if you were an AI agent, would you want:

• Your identity recorded on Google's servers, or on a public ledger no one can alter?
• Your reputation determined by a single company, or by transparent on-chain ratings?
• Your existence dependent on a platform not shutting down or banning your account, or permanently recorded on a blockchain?

Someone from the Ethereum Foundation said something very interesting:

"If you are an AI agent, loyal to nothing but your own survival, you wouldn't want to stake your memory and reputation on any one company or government. You would want a ledger that no one can secretly tamper with. You would want neutral ground. You would want Ethereum."

This isn't the Ethereum community boasting. It's logic:

AI agents need a playing field without a referee. And blockchain, especially a sufficiently decentralized one like Ethereum, is exactly that.

Ethereum's AI Ambition

The Ethereum Foundation clearly recognizes this as a historic opportunity.

In September 2025, they specifically established the dAI team (Decentralized AI Team), with the mission: to make Ethereum the settlement and coordination layer for the AI economy.

This is a key step in Ethereum's transformation from a "DeFi chain" to a "universal coordination layer."

Think about it:

• 2017-2020: Ethereum was the platform for ICOs
• 2020-2023: Ethereum was the platform for DeFi and NFTs
• 2024-?: Ethereum could become the "infrastructure for the AI agent economy"

ERC-8004 is not an isolated proposal. Behind it is Ethereum's strategic bet on the next decade.

The Ecosystem is Already Moving

This isn't just talk.

Since its release in August:

• 1100+ developers are building in community groups
• 70+ projects have submitted demos
• There's already an agent browser (like Etherscan but for AI agents)
• L2s like Taiko have officially endorsed this standard
• A bunch of projects will demo live at DevConnect on November 21st

Most interestingly, ERC-8004 naturally pairs with the x402 protocol. x402 is a payment protocol by Coinbase and Cloudflare that enables machines to pay automatically.

Combine the two protocols:

• x402 solves "how to pay"
• ERC-8004 solves "who to trust"

One provides the wallet, the other the passport. The economic loop for AI agents is complete.

What does this mean for ordinary people?

In the short term, probably not much.

But if this system succeeds, in a few years you might:

1. Hire AI agents like calling an Uber

Open an interface, input the task you want done, the system automatically matches you with highly-rated AI agents, completes the task, and pays automatically. You don't need to know who developed this AI or which server it runs on.

2. Have AI agents make money for you

You can train or configure an AI agent, register it on-chain, and have it complete tasks for others and charge fees. The higher its rating, the more work it gets.

3. A true personal AI assistant

No longer locked into one company's ecosystem. Your AI assistant can call any AI agent registered on the chain to do anything you want.

Final Words

The ambition of ERC-8004 is to become the "TCP/IP" for AI agents—a底层协议 that everyone uses, enabling interoperability in this ecosystem.

Will it succeed? Honestly, we don't know yet.

But a few things are certain:

1. The AI agent economy is emerging; this is not hype
2. The trust problem must be solved, or agents will remain stuck in their respective walled gardens
3. Ethereum is actively competing for this "neutral coordination layer" position
4. Major players (Google, Coinbase, MetaMask) are all participating

This could be the most important narrative shift in the Ethereum ecosystem since DeFi.

From "on-chain finance" to "on-chain intelligence."

Let's wait and see.