Have you raised lobsters lately?" These days, when Web3ers greet each other, this is likely what they’ll say eight or nine times out of ten.
At the start of 2026, following the robotic performance that stole the show at China’s Spring Festival Gala, a new generation of AI agents, exemplified by OpenClaw, has become the latest obsession among tech enthusiasts. Some use AI for customer service, others for coding, and some are even experimenting with agents to simulate an entire team of “digital employees.” Lately, the concept of a “one-person company”—where a single individual can run tasks that once required a small team, all through an AI-powered workflow—has been frequently mentioned across various internet platforms.
Web3 hasn’t been standing still either. Lately, if you pay attention to industry media, you’ll notice many projects beginning to focus on AI agents. Some are exploring how agents can directly interact with on-chain assets or smart contracts, others are building payment, identity, or financial infrastructure for agents, while some are discussing an “agent economy” where AI can participate in networks just like users—and others are even reviving the slogan of “Web4.0.”
By now, you might feel a sense of familiarity.
They say fashion is cyclical, but few expected the tech world—or the crypto world—to be the same. Remember how, during the bear market that began in 2022, ChatGPT suddenly went viral and AI became the topic on everyone’s lips? The Web3 community didn’t sit idle either; soon, a flood of new concepts emerged—AI agents, AI traders, automated strategies—as if merely tying something to AI was enough to craft a compelling narrative. But this excitement didn’t last long. Once the crypto market rebounded, attention quickly shifted back to Crypto itself.
This second half of 2025, as the crypto market shows signs of a bearish trend, Web3 is seeking new concepts to take the baton.
But, according to Portal Labs, the problem lies precisely here. When a narrative starts to gain popularity, many Web3 startup teams aren’t making technical or business decisions—they’re making narrative decisions: whichever concept is trending, they pursue it. And then they stumble.
Many teams only realize after actually moving forward with their project that while the concept can be quickly assembled, bringing the product to life is much harder. Where are the users? What are the specific use cases? How will you sustainably generate revenue? Can you secure investment? These questions often only emerge after the project has been underway for some time.
Once the hype fades, what remains on the market are often numerous projects that never fully materialized. Some products stall at the demo stage, others barely launch without finding users, and some simply vanish along with the narrative. In the short term, it may seem like a new赛道 has been opened, but looking back after some time, very little truly endures.
As a result, whether to continue focusing on Crypto or shift to AI has become a difficult decision. Choosing the former is challenging because the market is weak, and investment may not yield returns; choosing the latter is uncertain, as there’s no solid foundation. AI’s technical barriers, talent structure, and competitive landscape differ significantly from Web3. Over the past few years, many teams have built their technology stacks, product experience, and community resources within the Crypto ecosystem—fully transitioning to AI would mean entering an entirely unfamiliar field. From model capabilities and data resources to engineering teams, almost everything would need to be rebuilt from scratch.
More realistically, the AI sector is already extremely crowded. Major large-model companies, traditional internet firms, and countless startup teams have all invested substantial resources in this space. For a startup team originally focused on Web3, entering this market solely due to a shift in narrative could easily leave them without technical advantages or industry resources.
In fact, for many Web3 startup teams, there is another practical path: instead of necessarily transitioning into AI, they can continue along their Web3 journey while considering what capabilities crypto can bring to the AI ecosystem.
If you look closely at this wave of AI development, you'll find that many key aspects have not yet been fully resolved.
The most typical example is data. As models grow increasingly powerful, questions remain unanswered about where training data comes from, whether the data is trustworthy and compliant, and especially how AI agents can achieve one-to-one customization. For AI systems that rely on large-scale data training, this is a long-standing foundational issue.
For example, consider identity and collaboration. When AI agents begin participating in task execution, automated trading, or operational decisions, they themselves require identity, permissions, and collaboration rules. Who can invoke a specific agent? How do agents divide responsibilities? How are settlements handled after task completion? These questions fundamentally involve identity and value distribution within an open network.
There are also payment issues. Once AI agents begin autonomously invoking services, retrieving data, or performing tasks over the network, they require a micropayment system capable of automated settlement. However, such a payment structure is difficult to implement within the traditional internet framework.
These may seem like AI problems, but many of the solutions already exist within Crypto’s technological framework—whether it’s data incentive networks, on-chain identity systems, or open payment networks, all of which Web3 has been exploring over the past few years.
If a Web3 startup team truly intends to explore these directions, there are several things they must first clarify.
First, examine the team’s own technical capabilities. Different Web3 projects vary significantly in their technical expertise. Some teams specialize in on-chain protocols, others have long focused on data networks, and some are more oriented toward application-layer products. If a team has spent the past several years building data-related infrastructure—such as data collection, data extraction, or data marketplaces—extending into AI-driven data layers will come naturally, for example, through data contribution networks, verifiable data sources, or incentive-driven data markets for models. If the team’s background is primarily in on-chain protocols or infrastructure, consider building around the execution environment for AI agents—such as on-chain identities for agents, permission management, task execution protocols, or automated settlement and payment capabilities for agents. For teams already focused on application-layer products—such as trading tools, content platforms, community products, or consumer apps—AI is best integrated as a capability layer within their existing product ecosystem, for instance, by enhancing data analysis, automating operational workflows, or using agents to handle functions previously requiring manual intervention.
Next, consider whether there is a real business use case. Many AI projects disappear quickly not because of poor technology, but because they lacked a clear use case from the start. Concepts may sound exciting, but the fundamental questions—Who actually needs this product? Why would they use it? And why would they pay for it?—are often left unanswered. Some ideas, such as “AI + Web3,” “Agent economy,” or “AI traders,” sound impressive on the surface, but upon closer examination, there are few stable, existing user bases. In contrast, less “sexy” needs—like data processing, automated operations, information filtering, or task execution—continue to exist as persistent real-world business challenges. That’s why, when evaluating whether to enter an AI direction, it’s more important to examine the use case itself than to chase trending concepts: Is this a long-standing business problem? Are people already paying to solve it? And can AI genuinely improve efficiency in this specific area? If these conditions hold true, then this direction has a much higher chance of evolving from a narrative into a viable product.
Further down, we need to assess whether the Web3 startup team has access to resources that can truly enter these stages.
The previously mentioned areas—data, identity, and payments—are fundamentally not purely technical issues, but rather issues of network resources.
For example, in a data network, if the team lacks a stable source of data and a user base that can consistently contribute data, even if the technology is developed, it will be difficult to generate real network effects. Similarly, if you aim to build an identity system or collaboration network for AI agents, you need real developers, applications, or agents to participate; otherwise, the protocol itself cannot form an ecosystem. The same logic applies to payment and settlement systems. Once AI agents begin invoking services, accessing data, or executing tasks within a network, microtransactions will become extremely frequent. However, such a payment network only holds meaning when a large number of agents and services coexist; otherwise, it remains merely a technical module.
For many Web3 teams, what truly needs to be evaluated isn’t whether there’s technical room in this direction, but whether they can become part of this network. Whether the team already has access to data sources, a developer ecosystem, or real-world use cases often determines if a project can truly enter the infrastructure layer of AI, rather than remaining at the conceptual stage.