Whether AI Agent can become the next crypto outlet is the key to moving from narrative to practicality

By Alec Goh, Principal, HTX Ventures

Building a Sustainable Crypto + AI Value System In the

Over the past six months, I've come to realize that the combination of AI Agents and Web3 has the potential to do just that. However, the premise of all this is that the industry must move beyond the early hype stage to truly build long-term value for the end user.

We're familiar with stories like this. In 2017, the ICO craze was far ahead of the launch of products; In 2021, NFTs exploded into the mainstream, and most platforms still have incomplete infrastructure. Both cycles have one thing in common – the bubble comes quickly, and the pullback is just as fast.

explosion of AI Agent tokens is the same. Dozens of projects were launched under the name "Crypto + AI", which quickly attracted a lot of money and attention. Some projects were valued at billions of dollars in a matter of weeks. But as always, it is difficult to build a sustainable ecology with speculation alone. Once the investment enthusiasm has waned, user engagement has declined, and most early-stage projects have begun to lose their long-term relevance.

But in this round of fluctuations, some essential changes have indeed occurred.

In the fourth quarter of 2024, DeepSeek achieved a pivotal technological breakthrough that caught the attention of us at HTX Ventures. The project uses reinforcement learning to train large-scale models without relying on manual annotated data. This means that the cost of developing autonomous intelligent agents has dropped significantly, and for the first time, AI Agents have the possibility of large-scale, low-barrier deployment in Web3.

This shift is significant. If crypto is to deliver on its promise of a "decentralized, programmable financial infrastructure," intelligent agents with autonomous decision-making capabilities will become an indispensable underlying component. But at present, we are still far from this vision.

I fear that the industry may once again be focused on catching up on trends and forgetting to keep its feet on the ground. Indeed, a few promising frameworks have emerged, such as Eliza, which is building composable agent tools that adapt to multi-chain ecosystems such as Ethereum, Solana, and TON. Launchpads such as Virtual and Clanker accelerated token offerings; APPLIED PRODUCTS SUCH AS GRIFFAIN AND NEUR ARE ALSO EXPERIMENTING WITH ASSET MANAGEMENT AND GOVERNANCE TOOLS.

However, these are only early signs, and have not yet formed a well-defined system ecology.

Too many AI Agent projects are still speculatively oriented rather than sustainability-oriented. If this continues, we may once again face the old problems of capital flight, disappointed expectations and slowing project schedules.

It would be a great pity.

Because these application requirements are real. DeFi strategies that are too complex for most users can be managed automatically through intelligent agents; On-chain governance can be analyzed by the Agent, comparing the history, and voting according to user values; Cross-chain automation can also be performed directly by the agent based on the user's intent.

These scenarios are no longer hypothetical – they are real needs that have arrived. What the current industry lacks is not imagination, but stable execution, user retention and a viable token economic model.

What's next?

First, we must look at AI Agent projects in the same way that we look at DeFi protocols.

This means moving beyond the vague narrative to the rigorous logic of the project, asking the key question:

• Who are the actual users?
• Is it of indispensable practical value?
• Where does the income come from?
• Will the protocol survive after the token incentives disappear?

The protocol should not only demonstrate the potential of the technology, but also demonstrate real-world use cases, stable revenue models, and quantifiable impact. This is the fundamental difference between a "proof of concept" and a "worth building".

Second, in addition to applications, the industry needs to build the infrastructure that supports the successful deployment of AI agents, including the agent framework, data layer, and computing resources. HTX Ventures will continue to closely monitor the development of this type of infrastructure, as some of it will become a necessity in the future, just as EVM compatibility is for L1 public chains.

Finally, we need to build an industry culture that encourages iteration and transparency.

Not every project can get it right the first time, but those teams that can consistently optimize and polish are likely to be frontrunners in the next phase of Web3.

The convergence of AI and encryption is not a "concept", but an "evolution". But if we are to truly seize this opportunity, the industry must shift from "competing for attention" to "creating value".

This is not only the investment logic of HTX Ventures, but also our firm belief formed through years of experience and future judgment.

We have the tools, and we have the Builder. Now, what we need most is self-discipline.