The Global On-chain Asset Summit, hosted by HashKey Group, was held today in Singapore. During the summit, Ethereum co-founder Vitalik Buterin joined Dr. Xiao Feng, Chairman and CEO of HashKey Group, for a fireside chat to discuss the future of Ethereum applications and blockchain development.

Below is the full transcript of their dialogue:

Dr. Xiao Feng: Today in Singapore, we want to talk about how Ethereum can confront or drive the development of applications on the Ethereum network. In my view, blockchains, despite their thousands of variants, can be divided into two categories. One is blockchains that do not support third-party applications, like Bitcoin, which has already become a killer application for blockchain, and after 15 years is globally accepted. The other category, represented by Ethereum, was born to attract and empower developers and users. My first question is: what is your current outlook on applications on Ethereum or blockchain in general? Do you think the time for applications has arrived, and in which direction will they evolve?

Vitalik Buterin: I think we can talk about two main directions. The first is what I called low-risk DeFi in an article I wrote two weeks ago. Low-risk DeFi refers to the types of applications that have existed on Ethereum for a long time, and we already know they are relatively safe. You can hold tokens, make trades, do lending and borrowing, or swaps—these are all quite simple. They also exist in traditional finance, but blockchain adds advantages like being global and always open. Examples include Morpho and stablecoins, which already bring users 3% to 4% returns in USD, EUR, or other fiat. For many people without access to reliable banking services, that’s extremely valuable.

Three years ago, low-risk DeFi didn’t exist—DeFi did, but always with the problem that a new technology comes with many risks: bugs, hacks, and vulnerabilities. Back in 2019, the amount of funds hacked was 5% of total DeFi TVL. This year, it has dropped to about 0.02%. That shows it takes time for a new financial ecosystem to mature, and now we’ve reached a stage where it is actually working successfully.

The second direction is more creative applications. I mean things like decentralized prediction markets, ENS, ZKID, and others. We don’t know yet which of these will become the next big success, but some could suddenly explode. For example, last year Polymarket gained huge traction—before 2024, prediction markets were just something economists liked to talk about, but last year mainstream media and large numbers of people were sharing Polymarket screenshots. So Ethereum is important both for low-risk DeFi and for creative, experimental apps. Both matter a lot.

Dr. Xiao Feng: You just mentioned prediction markets. I remember very clearly—in 2015 you introduced us to them, and we even invested at the time, in Algo.

Vitalik Buterin: I still have 50,000 of their tokens.

Dr. Xiao Feng: Since then I’ve been following prediction markets closely, but after 2017, the conversation faded. Then suddenly in 2024 Polymarket exploded, even successfully predicting Trump’s election. My first thought was: Vitalik’s prediction markets have finally arrived—they are mature and useful now. It seems it really takes patience to wait for a non-financial application to break out. So far, apart from prediction markets, do you think there are other non-financial applications that have already proven successful, or—like you said about Algo in 2015—at least show promise in the right direction?

Vitalik Buterin: Recently I’ve been talking a lot about ZKID. Why? Because identity and privacy are crucial in today’s world. Centralized databases will inevitably get hacked, and data will leak—we’ve seen that many times. The value of ZKID is that you can prove key information without revealing all your data. For example, in under-collateralized lending, the hardest problem is knowing who is likely to repay and who is not. ZK combined with AI can help provide this proof, enabling people globally to access loans.

I think this is an interesting example of combining financial and non-financial elements. In the future, many applications will have both. Decentralized social networks are another case: they started as non-financial, but now many platforms are experimenting with financial features. Sure, 90% will fail within five years—that’s normal—but the remaining 10% could be very exciting.

Dr. Xiao Feng: That reminds me of an economic observation: every industrial revolution has been accompanied by a financial revolution. Britain’s industrial revolution relied on banks and bonds; the information revolution relied on venture capital. Now AI and crypto are seen as the fourth industrial revolution, and this too should come with new financial services. In recent years, VC hasn’t succeeded much in crypto. Perhaps new fundraising models will appear. Imagine entrepreneurs using ZK-based identity proofs and token mechanisms to raise $100,000 globally. This could replace traditional VC and become the financial revolution of the digital era.

Vitalik Buterin: Yes, I also think the combination of financial and non-financial will drive this wave of innovation.

Dr. Xiao Feng: Let me ask a question developers often face. In Web2, everyone asks whether your model is ToB or ToC. Do you think in blockchain there’s still such a distinction? Is Ethereum ToB, ToC, or a third way?

Vitalik Buterin: In blockchain, the distinction between B2B and B2C is much smaller than before. If you issue an ERC-20 token, both individuals and institutions can hold it. On-chain, they are equal—just addresses. The only difference is institutions have larger amounts and need higher security, so they use multisig wallets. But fundamentally, they’re still addresses. Similarly, the difference between L1 and L2 is not about one being ToB and the other ToC. It depends on the application: some are more suited to L1, others to L2.

Dr. Xiao Feng: Users also face another confusion at the application layer: decentralization boundaries. For example, stablecoins cannot be said to be 100% decentralized—there’s always an issuer like Circle. What’s your view?

Vitalik Buterin: It’s a complex issue. Many applications inevitably require partial centralization and trust. For example, under-collateralized lending is essentially about trusting someone. But if people say “L1 is decentralized, so we don’t care what happens at the application layer,” that could be dangerous. Take email: in theory, it’s a decentralized protocol, but in practice a few big platforms dominate. Blockchain, however, still offers important advantages: even if an L2 is semi-centralized, users can withdraw their assets themselves through L1 contracts. That’s a very real benefit.

Dr. Xiao Feng: One last question. Recently, Google and EF have been working on an Agent-to-Agent Payment standard. What do you think of AI payments?

Vitalik Buterin: In the next five to ten years, AI will handle many payments. But I don’t think large payments will be handled entirely by AI—it’s too risky. AI is too easy to attack. But for small payments, it’s fine, and AI can also serve as a risk control tool for large ones—checking if a $100,000 transaction looks suspicious or not. I think we’ll gradually discover the right ways to combine AI with payments. There will be errors, but also successes.

Dr. Xiao Feng: Thank you very much, Vitalik. Today you’ve helped us address many questions from the perspective of Ethereum’s founder.

Vitalik Buterin: Thank you.