"Have you raised lobsters yet?" This is probably the most common greeting among Web3 users these days. Since the Chinese New Year Gala's robot performance that stunned audiences, a new generation of AI agents, represented by OpenClaw, has become a new toy among tech enthusiasts. Some are using AI for customer service, some are using AI to write code, and some are even trying to use agents to simulate a whole set of "digital employees." The concept of "one-person companies," which has been frequently mentioned on various internet platforms recently, refers to a single person using an AI workflow to complete tasks that previously required a small team. Web3, of course, hasn't been idle either. If you look at industry media lately, you'll find that many projects are starting to focus on AI agents. Some are researching how agents can directly access on-chain assets or contracts, others are working on payment, identity, or financial infrastructure for agents, some are discussing "Agent economic systems" that allow AI to participate in the network like users, and some are even chanting the new slogan of "Web 4.0." Seeing this, there's a familiar feeling. It's said that the fashion world is cyclical, but who would have thought the tech world (or rather, the crypto world) is the same? Remember the bear market that started in 2022? ChatGPT became an overnight sensation, and AI suddenly became everyone's talking about. The Web3 community wasn't idle either, quickly spawning a bunch of new concepts like AI Agents, AI Traders, and automated strategies—it seemed that anything even remotely related to AI could be used to tell a new story. But this excitement didn't last long. Once the crypto market rebounded, everyone's attention quickly returned to crypto itself. In the second half of 2025, the crypto market experienced another bearish trend, prompting Web3 to seek new concepts to capitalize on. However, Portal Labs believes this is precisely where the problem lies. When a narrative becomes popular, many Web3 startups aren't making technical or business judgments, but rather narrative judgments: whichever concept is hot, they pursue it. And then they stumble. Many teams only realize when they're actually pushing their projects forward that while concepts can be quickly built, products are difficult to implement. Where are the users? What are the specific scenarios? How will they generate revenue? Can they attract investment? These questions often only emerge after the project has been running for a while. By the time the hype dies down, what remains on the market are often a bunch of unworked projects. Some products remain in the demo stage, some barely launch but can't find users, and some simply disappear along with the narrative. In the short term, it seems like a new track has opened up, but looking back after a while, not much has truly remained. Therefore, the dilemma is whether to continue focusing on Crypto or shift to AI. Choosing the former means facing a poor market and uncertain returns on investment; choosing the latter leaves one uncertain. The technical barriers, talent structure, and competitive environment of AI differ from Web3. Many teams' technical stacks, product experience, and community resources accumulated over the past few years are actually built within the Crypto ecosystem. A complete shift to AI is tantamount to re-entering a completely unfamiliar track. From model capabilities and data resources to engineering teams, almost everything needs to be rebuilt. More realistically, the AI track itself is already extremely crowded. Large modeling companies, traditional internet enterprises, and numerous startups have all invested heavily in this field. For a startup team originally working on Web3, entering this market simply because of a narrative shift can easily lead to finding that they lack both technical advantages and industry resources. In fact, for many Web3 startups, there's another path they can explore. They don't necessarily have to transition to AI; instead, they can continue along their existing Web3 path while considering what capabilities Crypto can contribute to the AI ecosystem. If you look closely at the current wave of AI development, you'll find that many key aspects haven't been fully resolved. The most typical example is data. Models are becoming increasingly powerful, but where does the training data come from? Is the data trustworthy and compliant? Especially, how can AI agents achieve 1v1 customization? These questions still lack a good mechanism. For AI that relies on large-scale data training, this is a long-standing fundamental problem. Another example is identity and collaboration. When AI agents begin to participate in task execution, automated transactions, and even operational decision-making, they themselves need identity, permissions, and collaboration rules. Who can call a particular agent? How are tasks divided among agents? How is settlement handled after task execution? These issues essentially all relate to identity and value distribution in open networks. There's also the payment problem. Once AI agents begin autonomously calling services, acquiring data, or performing tasks within the network, it means they need a micro-payment system that can automatically settle payments. In the traditional internet ecosystem, such a payment structure is difficult to implement. These all seem like AI problems, but many solutions already exist within the Crypto technology framework. Whether it's data-incentivized networks, on-chain identity systems, or open payment networks, these are directions that Web3 has been exploring for the past few years. If a Web3 startup team truly intends to try these directions, there are several things they must first consider. First, the team's technical capabilities must be assessed. Different Web3 projects have vastly different levels of technical expertise. Some teams excel at on-chain protocols, some have long focused on data networks, and others are more application-layer focused. If a team has been building data-related infrastructure for the past few years, such as data collection, data extraction, or data marketplaces, then extending to the data layer around AI will be relatively natural. This could include data contribution networks, verifiable data sources, or providing incentivized data marketplaces for models. If the team is more focused on on-chain protocols or infrastructure, they can consider building around the operating environment of AI agents, such as agent on-chain identity, permission management, task execution protocols, or providing agents with automated settlement and payment capabilities. For teams already developing application-layer products, such as trading tools, content platforms, community products, or consumer applications, AI is more suitable as a capability layer embedded in their existing product system. For example, AI can be used to improve data analysis capabilities, automate operational processes, or use agents to complete functions that previously required manual processing. Secondly, it's crucial to consider whether there are real-world business scenarios. Many AI projects disappear quickly not because the technology is flawed, but because they lack a clear use case from the outset. The concept can be very popular, but questions like where the people who actually need the product are, why they want to use it, and why they are willing to pay for it are often not seriously answered. Some concepts are widely discussed in the industry, such as "AI + Web3," "Agent Economic System," and "AI Trader," which sound very ambitious. However, if you delve deeper, the truly stable user base is actually quite small. Conversely, some seemingly less "sexy" needs, such as data processing, automated operations, information filtering, or task execution, have long existed in real-world business. Therefore, when deciding whether to enter a particular AI field, it's better to look at the scenario itself rather than whether the concept is popular: Is this scenario a long-standing business problem? Are people already paying for it? And can AI truly improve efficiency in this area? If these conditions are met, then this direction is more likely to move from narrative to product. Further down the line, it's necessary to examine whether the Web3 startup team has the resources to truly enter these areas. The data, identity, and payment directions mentioned earlier are not essentially simple technical issues, but rather issues of network resources. For example, in data networks, if a team lacks a stable data source and a user base capable of consistently contributing data, even if the technology is developed, it's difficult to create a true network effect. Similarly, building an identity system or collaborative network for AI agents requires the participation of real developers, applications, or agents; otherwise, the protocol itself struggles to form an ecosystem. Payment and settlement systems follow a similar logic. Once AI agents begin calling services, acquiring data, or performing tasks within the network, small payments become very frequent. However, such a payment network is only meaningful when a large number of agents and services exist simultaneously; otherwise, it remains merely a technical module. Therefore, for many Web3 teams, the real challenge isn't assessing "whether there's technological potential in this direction," but rather whether they can become part of the network. Whether the team already possesses data sources, a developer ecosystem, or application scenarios often determines whether a project can truly enter the AI infrastructure layer, rather than remaining merely a concept.