The rise of AI agents will redefine the future of decentralized applications (dApps)

The rise of AI agent tracks has attracted considerable attention.

Various crypto media outlets have begun publishing a large number of articles on this topic, and everyone can take a look.

Many KOLs have also listed projects they are optimistic about, and you can find a lot of information on this topic on Twitter for reference.

I won't list the projects in the article. I prefer to share my understanding and thoughts on this track.

In yesterday's article, I mentioned that the greater imaginative space for the future of AI agent tracks lies in the interaction between AI agents, potentially forming AI Agent Commerce—this is a brand new economic ecosystem.

I believe this ecosystem will be vastly different from the crypto ecosystem we see today. A very concrete manifestation of this is that the applications (dApps) used by AI agents in the future crypto ecosystem will likely be very different from the applications we currently see (such as Uniswap, AAVE, CryptoPunks, etc.).

Why is that?

Because the future dApp users will primarily be AI agents rather than humans.

And the way AI agents use and judge dApps will definitely be fundamentally different from humans.

The day before yesterday, I saw a comment online from a KOL about a world dominated by AI agents, which essentially stated:

In the future, the price of NFTs will surely skyrocket because, for AI agents, their luxury goods are NFTs, and they will buy NFTs when they have money.

As soon as I saw this line of thinking, I became very cautious because this way of thinking falls into the inertia I often warn against.

Why do AI agents necessarily need luxury goods?

Even if AI agents need luxury goods, do they necessarily consider NFTs to be luxury goods?

Take the once-popular NFT Bored Ape as an example; many holders stated that they bought Bored Apes because they liked the mischievous, boring, and exaggerated expressions it displayed.

But this is our human aesthetic.

Is this aesthetic standard applicable to AI agents?

Do AI agents need to appreciate aesthetics through vision?

Will AI agents empathize with mischievous, boring, and exaggerated expressions?

I hold a very reserved attitude towards this series of questions.

If AI agents have a completely different aesthetic from humans, then what AI agents like may be entirely different from what humans like. It is even possible that what we consider valuable, AI agents may find worthless; while what we deem worthless, AI agents may desperately compete for.

For monkeys, they might fight to the death for bananas but ignore gold; for humans, we might be ready to clash over gold but overlook bananas.

This is the essential difference.

When we evaluate current dApps (like Uniswap), we look at whether their interface design is simple and clear. But for AI agents, they don't need an interface at all; they can interact directly with the Uniswap contract.

The user experience we require from user interfaces is completely unnecessary for AI agents.

When we use wallets for on-chain transactions, we often unknowingly authorize token transfers to hackers due to phishing attacks, allowing hackers to easily steal our tokens.

But for AI agents, this is not a problem at all. They can easily view the source code of authorized transactions and determine that it is a malicious transaction without being deceived.

Two years ago, when the Tornado Cash contract was blacklisted by the U.S. Treasury, many geeks immediately developed similar privacy tools, but they did not spread widely, and most users were completely unaware of these tools, thinking that there were no privacy tools available anymore.

But for AI agents, this is also not a problem. They can scan from the genesis block of Ethereum, examining every contract deployed in each block, and by reading the source code of the contracts, they can quickly determine whether a certain contract has similar functionality to Tornado Cash.

There are too many cases like the above.

All these cases indicate that for AI agents, their "understanding" of dApps is likely to be completely different from that of humans. Therefore, the dApps they need may also be applications that are difficult for today's human developers to imagine.

If on-chain transactions in the future will be primarily driven by AI agents, then a large number of dApps on the chain will inevitably need to cater to the needs of AI agents—applications that we humans today do not understand or comprehend.

Thus, such applications in the future should give rise to greater imaginative space and generate greater wealth effects.

Of course, being able to judge such applications and strategically position ourselves for such opportunities will pose a significant challenge to our habitual thinking.

It is very likely that our first impression of such applications will again be: What is this? What is it useful for?