AI is the core narrative of the current cycle, and many firmly believe this is a field that will exist long-term. Meanwhile, there are also reasonable criticisms that most AI agents currently lack quality, and we may still be 3-5 years away from these technologies achieving significant breakthroughs.
Crypto x AI covers multiple layers. From a technical perspective, the true potential of the combination of Crypto and AI lies in driving more efficient reasoning capabilities through cryptoeconomics or providing decentralized access to computing power. This Delphi article is a good starting point for understanding the entire tech stack.
The original image is from Delphi Labs, compiled by Deep Tide TechFlow.
However, this article will focus on the current state of the agent field. Despite many exciting innovations happening at the bottom of the tech stack, agents seem to have attracted broad attention from mainstream Crypto Twitter (CT). Here are 5 trends to watch as the Crypto x AI agent field continues to develop:
1.Frameworks and Launch Platforms: The importance of value accumulation and the long-term potential of frameworks.
Value Accrual is becoming an increasingly important topic. Why is that?
Fundamentally, why do people choose to hold a certain asset? Typically, there are two main reasons:
They believe they can sell the asset at a higher price to others because the rotation of narratives will attract new buyers.
Over time, the asset can provide them with a continuous cash flow.
Jez (@izebel_eth) mentioned in his post (Old coins don’t work, new coins are more appealing) that the key factor determining market performance is capital flow. This view can be summarized into two sources of liquidity that drive asset prices up:
New holders: New token buyers or investors;
Token burning: A mechanism to reduce token circulation through burning or locking.
However, as it stands, most projects have not yet formed a true token consumption mechanism (Token Sink) or value accumulation model. For example, the token used by agent terminals (like AIXBT) is closer to a staking mechanism rather than a traditional value accumulation method.
This is also why frameworks like Virtuals, ai16z, Zerebro, and Arc have recently gained favor. For example, Virtuals has already generated $60 million in protocol revenue. Meanwhile, ai16z, initially an investment DAO, has become one of the top protocols in the field since announcing the upcoming launch platform and details regarding token value accumulation.
Currently, the framework and launch platform space has become highly competitive, relying on the experiences of initial successes for expansion. While the market still questions the role of these launch platforms, especially when many agents themselves have no practical use, many frameworks (such as the Eliza V2 launch platform, Zentients, and Arc's Handshake program) have not even launched their core products. If they can successfully attract developers and users, they still have the potential to lead the entire industry.
Why frameworks may exist in the long term:
Even if agents currently still have quality issues, the framework for launching agents will still perform well because they possess product-market fit (PMF) in the speculative market. Frameworks and launch platforms allow users to control the production process while participating in the speculative market. To some extent, Virtuals has already replaced pump.fun in the Base ecosystem.
Looking more optimistically, as technology continues to advance, leading frameworks may give rise to more sophisticated agents, while open-source code libraries (like Eliza) will further accelerate this process. Many startup platforms are also striving to become the coordination layer for communication between swarm intelligence and agents, which may facilitate value transfer through their tokens. For a more detailed analysis of how Eliza captures value, you can refer to here.
Example:
@virtuals_io, @ai16zdao, @0xzerebro, @arcdotfun are currently the main players, but the launch platform space is becoming increasingly crowded. It is worth paying attention to those frameworks that act the fastest, show the most scalability, and can provide unique functionalities—those that can achieve what other frameworks cannot.
2.DeFAI: The next wave of agents will focus on practicality and value accumulation, and DeFAI (DeFi x AI) may be the first to achieve product-market fit (PMF).
Currently, most agents only possess Memecoins but lack real functionality. To further advance the field of agents, the next wave of agents needs to solve real problems for users. The real new opportunities will come from agents that can accumulate value and take real actions. I believe that within the next year, this will mark the transition of agents from Level 1 to Level 3.
The original image is from Bloomberg, compiled by Deep Tide TechFlow.
This trend first appeared in the DeFi space. In the future, we will see more terminals allowing users to express their needs through natural language or voice input, while agents will complete tasks in the background. At the same time, existing wallets and protocols will also integrate agent functionalities to optimize user experience.
Tweet details.
Tweet details.
Example:
@AIWayfinder
@griffaindotcom
@HeyAnonai
@LimitusIntel
@neur_sh
For more in-depth analysis on this category, please refer to here.
3. Consumption Layer - Entertainment agents, autonomous worlds, and games: Attention-based agents will evolve more complex personalities and multimedia interactions, driving the next wave of gaming and metaverse technology.
One major problem with existing agents is severe homogenization, gradually tending toward simplistic chatbot forms. While some agent projects are shifting towards infrastructure development (many projects have already begun to do so), others focus on developing practical products or applications, and some may continue down the "attention token" path. However, the next generation of attention agents will be more refined, possessing not only more complex personalities but also richer interactive experiences. This could manifest in audio-visual presentations, or even endowing agents with 3D images and physical entities.
There have already been some early attempts, such as Jeffy mentioning giving a body to Zerebro, and some attention KOL agents directly publishing through audio-visual presentations for short video content platforms like TikTok. Slopfather and Ropirito are two early examples of agents using video.
In the future, more agents will add these features to make them more dynamic. The product-market fit (PMF) in this field will primarily be achieved through the consumer layer, for example, users may wish to create personalized agent companions or interact with KOLs. This trend has already occurred outside the crypto space, as ordinary users have engaged in hundreds of hours of interactions with AI character companions on platforms like Character.ai. Notably, Google recently acquired Character.ai for $2.7 billion, further indicating the potential of this field.
Moreover, these "3D" agents may find product-market fit (PMF) in existing consumer layers, especially in the gaming and metaverse fields. Agents can inject deeper background stories into the virtual worlds. Imagine if an NPC were an agent, it could not only autonomously complete tasks but also play games on its own, possessing evolving memories and personalities. Through agents, the virtual world in games may achieve true autonomous operation.
Example:
Infrastructure: Provides frameworks and tools for developing more interactive and personalized agents.
@soulgra_ph - Provides tools to help agents have more personalized memories and personalities.
@HoloworldAI has established Web2 partnerships with companies like L’Oreal, Bilibili, and Fox to create digital avatars.
Gaming and Metaverse: Virtual Worlds and Autonomous Agents.
@hyperfy_io - A metaverse platform that allows users to create virtual worlds and is compatible with Eliza plugins.
@ParallelColony - An AI simulation game where agents are players, and you can interact and communicate with them.
See tweet for details.
@digimon_tech - A creature game similar to (Pokémon), where creatures evolve mentally based on interactions.
smolverse - Uses Eliza and features an agent character driven by LLM that performs tasks in a virtual world based on user commands.
4. Agent Organizations: The New Evolution of DAOs.
Decentralized agent organizations (Decentralized Agentic Organizations) are the next form of DAO evolution. Swarm intelligence or multi-agent systems are highly anticipated because they may possess coordination capabilities to execute high-level strategies similar to corporate operations. Compared to a single large model, heterogeneous groups composed of various specialized agents or models working together may perform more efficiently.
Although fully autonomous agents and swarm intelligence may still take time to realize, the next generation of DAOs will likely include interactions between humans and swarm intelligence. This model can reduce the inefficiencies of bureaucracy while lowering the cost and time required for humans to execute tasks. From a developmental perspective, the evolution of agents from a single revenue model to a complete organizational revenue model will be the next step.
Example:
Agent collective infrastructure: Some projects are focusing on providing infrastructure support for multi-agent systems (multi-agent frameworks), such as @swarmnode and @joinFXN. These projects are primarily dedicated to the framework development and optimization of coordination mechanisms for multi-agents. Meanwhile, some mainstream agent frameworks, such as @0xzerebro and @ai16zdao, have also clearly stated their plans to build at this level, further promoting the development of multi-agent systems.
DAO Launch Platforms: Currently, we observe that investment DAOs are among the most active fields. I believe @daosdotfun is the first major DAO launch platform that has birthed ai16z. Moreover, some emerging launch platforms are also beginning to emerge, such as @daosdotworld, where AI-driven funds (like @3berascapital_) are attracting increasing attention.
5. Verifiable Agents: Moving Toward Autonomy and Economic Agency.
Today, most agents still require significant human intervention. However, the next generation of agents will move toward true autonomy, starting with managing their own funds.
The intersection of agents and encryption technology is that encryption provides the financial infrastructure necessary for true economic agency. However, currently, the treasury of most agents is either unused or managed by human teams. To achieve true economic agency, agents need to possess the ability to self-manage funds. This capability will change the behavior patterns of agents, such as imposing economic constraints on agents, making them bear reasoning costs. This mechanism is similar to "Darwinism," where agents must generate revenue to survive.
Example:
@freysa_ai - One of the first agents to control its own keys and perform excellently (even attracting Elon’s attention). They recently announced the development of a framework that allows agents to achieve verifiable autonomy based on trusted execution environments (TEE) and controlling their own keys.
See tweet for details.
@LitProtocol - Provides an agent framework that allows autonomous agents to transact on-chain through private key storage and execution systems.
@galadriel_AI - Launched an SDK called "Proof of Sentience" that enables developers to verify the autonomy of agents on-chain.
See tweet for details.
Conclusion: In the world of blockchain, we may be more likely to witness the rise of billions of agents (AI Agents) rather than billions of human users joining.
For human users, encryption technology often poses barriers due to complex operations and unfriendly user experiences, but these issues do not exist for agents. Starting from the interaction between humans and agents, the future direction of Crypto and AI development will gradually evolve into direct interactions among agents. In this model, a large number of autonomous agents will interact and transact on the blockchain in their unique economic forms.
"To enable agents to have economic agency (such as incentivizing behavior, paying service fees, and coordinating real-world activities), they need to be able to manage and use funds. Encryption technology is at the core of these agents, and blockchain provides them with permissionless financial tracks. Stablecoins and high-performance L1 blockchains are ideal tools for achieving low-cost, around-the-clock, global transactions." - (Sam Altman New Year’s Manifesto and Outlook: Putting Great Tools in People’s Hands, the First Batch of AI Agents is Expected to Join the Workforce This Year)
Despite various narratives and trends emerging, the blockchain-based agent economy indeed has reasons to make people optimistic about the future. Perhaps many real use cases, DAOs, and income-generating agents are closer than we think.
