Written by: Evan
Compiled by: 1912212.eth, Foresight News
AI is the theme of this cycle, and many firmly believe it will be a long-lasting field. On the other hand, there are also many rational criticisms stating that most current AI products are of poor quality and that it may take another 3 to 5 years for this technology to truly become meaningful.
Crypto x AI encompasses multiple levels. The true technical potential of crypto and AI is mainly reflected in driving better reasoning capabilities through cryptoeconomics or providing access to decentralized computing resources. This article from Delphi is a good starting point for learning about the entire technology stack.
However, this article focuses on the current development status of intelligent agents. Although there are exciting innovations in the underlying technology stack, intelligent agents seem to have captured the attention of mainstream crypto Twitter (CT). Here are several major trends in the continuously evolving field of crypto and AI agents at this level:
Framework + Launchpad: Value accumulation is becoming important, and frameworks may exist in the long term.
Value accumulation is beginning to become important. Why is that? Stepping back, why do people hold a certain asset? They trade assets with dollars for two reasons:
They believe they can sell the asset to others at a higher price because they seized the opportunity of narrative rotation, thus attracting new buyers.
The asset can generate more cash flow over time.
Jez (@izebel_eth) mentioned in his article (Old Coin Bad, New Coin Gud) that the only important thing is liquidity. These two reasons reflect two forms of bullish liquidity for assets:
New holders
Token destruction mechanism
However, in most cases, we have yet to see any real token destruction mechanisms or token value accumulation. Using tokens (like AIXBT) in agent terminal models is more akin to staking rather than traditional value accumulation.
This is also why frameworks like Virtuals, AI16z, Zerebro, and Arc have recently become so popular. Virtuals have already generated $60 million in protocol revenue. AI16z initially focused on investing in DAOs, but since releasing detailed information on the upcoming Launchpad and related token value accumulation, it has become one of the top protocols in this category.
Currently, the framework + Launchpad space has become extremely saturated, as many projects have developed relying on the halo of early successes. There is also a significant amount of skepticism in the market regarding the roles of these Launchpads, especially in cases where many agents are themselves useless. However, many frameworks (such as Eliza V2 + Launchpad, Zentients, Arc, and their handshake programs) have not yet officially launched their core products. If they can successfully attract developers and users, it is possible to continue leading the entire industry.
Why frameworks may exist long-term:
Whether agents still suffer from quality issues, the frameworks used to launch agents will continue to perform well because they cater to speculative behavior with market fit (PMF). Frameworks and Launchpads allow users to control the factory while participating in the casino. In many ways, Virtuals have already replaced pump.fun in the Base ecosystem.
Optimistically, as technology progresses, leading frameworks may give rise to more advanced agents, and open-source repositories like Eliza will further accelerate this process. Moreover, many Launchpads are also positioning themselves to become a coordination layer for group coordination and inter-agent communication, which may facilitate some form of value transfer through their tokens.
DeFAI: The next wave of agents will prioritize practicality and value accumulation, and DeFAI (the combination of DeFi and AI) may be the first category to achieve market fit (PMF).
Most agents are currently just meme coins with no practical use. For the agent field to advance further, the next wave of agents must be able to execute real use cases for us. True growth opportunities will come from substantial value accumulation and agents that can take action. I believe the next year will be a key period for the leap from level one agents to level three agents.
We first see this trend reflected in DeFi. In the future, we will see more terminals that allow users to express desired outcomes in natural language or voice, while agents will execute tasks in the background. Existing wallets and protocols will also integrate agents to improve user flows.
Consumer level—agent representatives, autonomous worlds, and gaming: attention-based agents will develop more complex personalities and multimedia interactivity, which could drive the next wave of gaming and metaverse technology.
One of the issues facing existing agents is that they all tend to become homogenized, turning into commoditized, beautified chatbots. While some agent projects will shift towards infrastructure (many are already doing so), others will begin to think about useful products or applications, while some may continue down the path of attention tokens. However, the next generation of attention agents will be more powerful, developing more complex personalities and providing richer interactivity. This could be achieved through audio and video performances or by giving agents 3D avatars and physical entities.
This idea has seen early attempts; Jeffy has written about the idea of giving Zerebro a body, and some attention-driven KOL agents have directly launched through audio and video performances, focusing on short video content like TikTok. Slopfather and Ropirito are two early examples of agents using video.
However, I believe more agents will join these characteristics, making them more dynamic. The market fit (PMF) of this field will be realized through the consumer layer, where individuals may wish to create personalized agent companions or interact with KOLs. This phenomenon has already occurred outside the crypto field, with ordinary people logging into sites like character.ai in droves to converse with AI character companions, and Google recently acquired the company for $2.7 billion.
Additionally, these 3D agents may find PMF at the existing consumer level, especially in gaming and the metaverse. Agents add deeper backstories to the worldview. Imagine an NPC that is an agent, able to perform tasks independently and play games on its own, with a continuously evolving memory and personality. A truly autonomous world in games may be realized through agents.
Agent organizations: The return of DAOs
Decentralized agent organizations are the next form of DAO evolution. Groups or multi-agent systems are very exciting because they may be able to coordinate and execute more advanced strategies, similar to managing a company. A heterogeneous group with various types of specialized agents and models working together may outperform a single large model.
Although fully autonomous agents and collectives may take some time to achieve, the next iteration of DAOs will likely include interactions between humans and collectives. This will reduce inefficiencies in bureaucracy and lower the costs and time of human execution. From agents earning income to full organizations earning income is the next developmental step.
Verifiable agents: Today's agents are moving towards greater autonomy and actually have their own liquidity.
Currently, most agents still require a high level of human intervention. The next wave of agents will move towards true autonomy, starting with managing funds.
The essence of agency and encryption in PMF is that encryption provides the financial infrastructure for real economic agents. However, most agents' treasuries are either not being utilized or are managed by manual teams. To achieve true economic agency, agents must be able to autonomously manage their own funds. This can begin to change agent behavior, as economic constraints can be imposed on agents, requiring them to pay their own reasoning costs. This will introduce Darwinism, making agents earn income to survive.
Final thought: We will see a billion agents on the blockchain, not a billion humans.
Cryptocurrencies are inherently unfriendly to human users, whereas agents have no concerns about this. We will start from the interaction between humans and agents, but the ultimate form of crypto AI is the interaction between agents, in which groups of autonomous agents will interact and trade within their local economies.
To give agents economic agency, and thus be able to incentivize behaviors (pay service fees) and coordinate actual activities in the real world, they need the ability to control and deploy capital. Cryptocurrencies are the home of these agents—the blockchain infrastructure allows agents to participate in permissionless financial activities. Stablecoins and high-performance L1s are ideal tools for driving efficient, around-the-clock, global trading.
Beyond meta-narratives and popular trends, we indeed have ample reasons to look forward to the complete on-chain realization of the agent economy. Perhaps many real application scenarios, DAOs, and income-generating agents are closer to realization than we imagine.