“Have you raised lobsters?” Recently, Web3ers greeting each other are probably mostly asking this question.
At the start of 2026, after China’s Spring Festival Gala robot stole the show, a new generation of AI Agents represented by OpenClaw has become the new toy among tech enthusiasts. Some use AI for customer service, others write code with AI, and some even begin to experiment with Agents to simulate a complete “digital workforce.” The concept of a “one-person company,” which has been frequently mentioned across various internet platforms lately, refers to a single individual running a workflow powered by AI to handle tasks that previously required a small team.
Of course, Web3 isn’t sitting still either. Recently, if you pay attention to industry media, you’ll notice many projects starting to focus on AI Agents. Some are researching how Agents can directly interact with on-chain assets or smart contracts, others are developing payment, identity, or financial infrastructure for Agents, and some are discussing an “Agent economy,” enabling AI to participate in networks like users do. There are even new calls for a “Web4.0” era.
Seeing all this, a familiar feeling arises.
They say fashion is cyclical, and it turns out the tech world (or crypto space) is the same. Remember the bear market starting in 2022? ChatGPT exploded overnight, and AI suddenly became everyone’s hot topic. The Web3 community didn’t lag behind; quickly, a bunch of new concepts emerged—AI Agents, AI traders, automation strategies—anything related to AI could be spun into a new story. But this excitement didn’t last long. When the crypto market rebounded, everyone’s attention shifted back to crypto itself.
Now, in the second half of 2025, with the crypto market showing signs of a bear trend again, Web3 is searching for new concepts to latch onto.
However, from Portal Labs’ perspective, the problem lies precisely here. When a narrative becomes popular, many Web3 startups aren’t actually making technical or business judgments—they’re making narrative judgments: whatever concept is hot, they jump on it. And then they often stumble—
Many teams only realize after pushing their projects that while concepts can be quickly assembled, products are hard to implement. Where are the users? What are the specific use cases? How to sustain revenue? Can they attract investment? These questions often only surface after the project has been underway for some time.
When the hype fades, what remains are often unfinished projects. Some products are stuck at the demo stage, some barely launched without users, and others simply vanish along with the narrative. It may seem like a new track has opened up in a short time, but looking back after a while, there aren’t many truly lasting outcomes.
Therefore, the dilemma is whether to continue deepening in crypto or switch to AI. Choosing the former means facing a tough market with uncertain returns; choosing the latter means lacking a solid foundation. AI’s technical barriers, talent structure, and competitive environment differ significantly from Web3. Many teams’ accumulated tech stacks, product experiences, and community resources over the past few years are built within the crypto ecosystem. Fully shifting to AI means entering a completely unfamiliar track—requiring rebuilding from model capabilities, data resources, to engineering teams.
A more pragmatic point is that the AI track itself is already highly crowded. Large model companies, traditional internet giants, and numerous startups are pouring huge resources into this field. For a Web3 startup, simply pivoting into AI due to narrative reasons often reveals a lack of technical advantage and industry resources.
In fact, many Web3 startups still have a viable path. They don’t necessarily need to pivot into AI; instead, they can continue on their Web3 journey while exploring how crypto can complement AI systems.
If you look closely at the current wave of AI development, you’ll find many key issues still unresolved.
The most typical is data. Models are becoming more powerful, but where does training data come from? Is the data trustworthy and compliant? How to achieve 1v1 customization for AI Agents? These questions lack good mechanisms. For AI relying on large-scale data training, this remains a fundamental, long-term issue.
Another is identity and collaboration. As AI Agents begin participating in task execution, automated trading, or operational decision-making, they also need identities, permissions, and collaboration rules. Who can call a particular Agent? How do Agents coordinate? How are tasks settled after completion? These questions fundamentally involve identity and value distribution in an open network.
Then there’s payment. Once AI Agents start autonomously calling services, fetching data, or executing tasks online, they require a small, automated settlement system. In traditional internet systems, such payment structures are hard to realize.
While these seem like AI problems, many solutions already exist within the crypto tech ecosystem. Incentive networks for data, on-chain identity systems, open payment networks—all have been explored by Web3 over the past few years.
If Web3 startups truly want to explore these directions, several things must be considered first.
First, the team’s own technical capability. Different Web3 projects have varying levels of technical expertise. Some excel at on-chain protocols, some focus on data networks, others on application-layer products. If a team has been working on data infrastructure—like data collection, extraction, or markets—then extending into AI data layers—such as data contribution networks, verifiable data sources, or incentivized data markets—feels more natural. Conversely, teams rooted in on-chain protocols or infrastructure might focus on AI Agent environments—like on-chain identity, permission management, task execution protocols, or automated settlement and payment systems. Teams primarily engaged in application-layer products—like trading tools, content platforms, or community apps—can embed AI as capabilities within their existing products, such as enhancing data analysis, automating operations, or enabling Agent-driven functions that previously required manual effort.
Second, the existence of real business scenarios. Many AI projects fade quickly not because of technical failure, but because they lack clear use cases from the start. Concepts like “AI + Web3,” “Agent economy,” or “AI traders” sound grand, but when digging deeper, the actual, stable user base is often small. Conversely, more “unsexy” needs—like data processing, automation, information filtering, or task execution—are long-standing in real business. When evaluating whether to pursue an AI direction, it’s better to look at the scenario itself: Is it a long-term business problem? Are there paying users? Can AI genuinely improve efficiency in this area? If these conditions are met, the idea is more likely to transition from narrative to product.
Third, the resources the team has to access these areas. Data, identity, and payment are not just technical issues—they’re network resource issues.
For example, a data network requires a stable data source and a community willing to contribute data; without these, even the best technology can’t generate network effects. Similarly, building an identity or collaboration network for AI Agents needs real developers, applications, or Agents participating; otherwise, protocols can’t form an ecosystem. Payment and settlement systems also depend on a large number of Agents and services operating simultaneously; otherwise, small-scale microtransactions are impractical.
For many Web3 teams, the real question isn’t whether there’s technical space in this direction, but whether they can become part of this network. Do they already have data sources, developer ecosystems, or application scenarios? These factors often determine whether a project can truly enter the infrastructure layer of AI, rather than just stay at the conceptual level.
