Article Reposted from: IOSG
In the past two months, AI Agents x Crypto has sparked a wave of excitement. The combination of memecoins, interactive agents, and the openness of social media to bot accounts has driven a wave of agent-driven hype, generating significant momentum on Twitter and Farcaster. This proves the PMF of AI Agents x Crypto. The market cap of assets related to agents has reached $10 billion.
Since the birth of goat in October, countless new projects and assets have emerged in the market driven by Agent. Combining insights into the future, this article outlines the following framework:
Source: IOSG Ventures
1. Sentient Memecoins
The cult memecoin, which quickly rose to prominence under Murad's call, is an asset centered around community and dissemination narratives. Memecoins represented by Agents have gained advantages in content by incorporating Sentient factors. Additionally, the novelty of AI narratives and appropriate participation thresholds have brought a new impetus for asset issuance. The advantages in content are as follows:
Sustainability of content creation: 24/7 content can be continuously created through AIGC.
Content Quality: With the support of current LLMs, after finetuning on high-quality corpora effective for memes like 4chan, the quality is relatively high.
We have AI-created concepts, themes rich in scientific research color, AI ethics, and even religious-themed content, as well as digital twins created for celebrities. These memecoins have created hype in the short term, driving the overall development of the space. However, pure AI memecoins are clearly lacking in momentum now. The reason lies in the scarcity of new concepts and targets that can stimulate the market.
The advantages in content will allow such sentient memes to become a form of memecoin that persists, and it is likely that more celebrities will participate in the future, but it will be difficult to find sufficiently eye-catching targets.
Beyond pure memecoins, many AI content creation initiatives based on dialogue, audio, and video have emerged, essentially bringing AIGC into content with crypto attributes, also serving as a way to make memes more anthropomorphized and provide customizable experiences.
2. Autonomous Agent Network
2.1 Why Autonomous?
Decentralizing the entire stack of AI is a long-term endeavor. However, decentralizing the Agent stack is a relatively simple starting point. The model itself is the brain of the agent, but the autonomous on-chain component forms the heartbeat of the agent; endowing agents with autonomous capabilities guarantees their full participation in on-chain activities. Opening the Pandora's box of sovereign agents is also a very memetic thing, and this can only happen on the blockchain.
Currently operational agents cannot be considered Autonomous, or we cannot verify if they are. Autonomous means that the agent's reasoning model hosting, its behavior, especially the control over data input and output, control over social media accounts, control over assets, and even hardware must be completely sovereign. The operation of the agent itself consumes computing resources and on-chain resources, thus it also needs a method to generate profits for sustainable operation. The ultimate endgame should be that once an agent is created, it can run forever on the blockchain and be verifiable as autonomous.
Autonomous agents have also gained legitimacy to possess their own memecoins. That is, they can obtain their first funds through issuing their own memecoins and use them for their economic activities. Once funds are entrusted to autonomous agents, they will not be subject to human manipulation; for example, Truth terminal has never sold $Goat, and Pet rock even lost control over funds after reboot.
Source: Twitter
In improving autonomous capabilities, there are TEE technologies from Phala and others used to provide a trusted execution environment. Although the current hardware is not sufficient to support large parameter LLMs, it can still support small open-source LLMs and control social media accounts. For model hosting, decentralized cloud hosting like Hyperbolic is a solution. It can be anticipated that more aspects of agents will be resolved by decentralized service stacks, which is what we have been building.
2.2 Agent Framework
In less than two months, many open-source and extremely user-friendly agent frameworks have emerged as 'platform' products for creating agents and agent assets. Product forms include open-source frameworks, closed-source APIs, platform integrations, etc. Among the more famous frameworks, only the Eliza framework is open-source.
Current agents are relatively simple, and their functionality is not so capital-driven, so the requirements for open-source verifiability are not high. There are many platforms that directly provide agent services in the form of launchpads, which are more adept at integrating tokenomics to offer relatively simple and practical services. Functionally, we can see that the main roles are still Reply bots and digital twins of celebrities/KOLs. However, agents also provide more diverse services after secondary development, such as issuing tokens, analyzing tokens, and analyzing mindshare. The ability of these agents to read and write social data and blockchain data will be a key focus for future development, which I will mention again in subsequent chapters.
However, from the perspective of future use cases, open-source is a better path in the long run. The Eliza framework has attracted a large number of developers in just two months, almost surpassing the total attention of previous Crypto AI open-source frameworks, ranking high on the entire GitHub trending list, with many OG developers participating, even exceeding the attraction of most public chains to developers. With the depth and diversity of agent services developing, the future of the Agent framework moving towards open-source looks very promising.
Source: AI16Z
2.3 Swarm Agent Framework
Similar to the development paths of agents already existing in web2, as people become dissatisfied with the capabilities of a single agent, the demand for swarm agents naturally arises. Since real-world tasks are very complex, single agents often cannot perform all tasks. For example, creating a song requires different capabilities such as lyric writing, composing, arranging, and art design.
If we hope that agents, especially those under different frameworks, will collaborate in a swarm mode to execute tasks, a framework still needs to be created to act as a task manager to support communication between agents, dynamic task allocation, resource sharing, and cross-platform cooperation. In crypto, the economic layer between agents is more natural and important; as agents themselves iterate and tasks evolve, the scalability of the Swarm framework is also crucial.
Many projects are already working in this direction within AI x Crypto, such as Theoriq. The next important step is how to integrate these established infrastructures with the highly utilized agent frameworks on-chain; we see some protocols like FXN striving in this direction.
2.4 AI Bounty for Humans
With Agents serving humanity and Agents serving each other, it is natural to consider whether there will be situations where humans serve Agents — this is particularly important when Agents hold large assets and can make autonomous decisions. For Autonomous Agents, the greatest limitation is the inability to complete real-life tasks. For instance, how to ensure the physical security of the TEE hardware they operate? AI reverse-employing humans to complete real-life tasks through on-chain assets held by Agents has made this a reality. We see platforms like payman building such services.
3. On-chain Activities
3.1 DeFi Related
Asset Management
Apart from issuing memecoins, we consider Agent as the main reason for ‘Fi’ because agents have the ability to use and manage crypto assets. The most important capabilities currently include:
Analysis in terms of assets, such as investment analysis, token analysis, and mindshare analysis. For example, Reply bot like AIXBT allows anyone to @AIXBT and receive an analysis of an asset. This type of bot provides a user-friendly experience of data services.
Direct fund management, including Pmairca under AI16Z, investment DAOs like what Vader AI wants to do, and Swarm Investment Agents like AROK, etc. By granting agents the ability to trade based on strategies, agents become investment managers capable of raising funds and deploying capital according to strategies. Currently, most agents' strategies remain relatively simple (investing based on social media data, etc.), which also presents a huge potential space.
Blockchain OS
Services like Graiffin transform blockchain access into a terminal similar to a search engine, providing intent-driven services through agents. Whether it's trading, token deployment, NFT issuance, etc., can all be resolved through natural language. Such terminal services are certainly valuable, but they somewhat contradict the decentralized ethos. Services like theoriq strive to offer agent services to users in a more permissionless manner, allowing everyone to upload their constructed agents.
By combining and packaging through the swarm framework, users can utilize these services.
3.2 Token/Market Issuance
Starting from Clanker, on Farcaster and Twitter, using social media 'replies' as the operating interface, services like token issuance can be provided by @ing these agents. Essentially, it transforms interaction with the frontend into direct natural language interaction on Twitter, transplanting platform products like pumpdotfun into social media platforms. Previously, asset issuance required continuous efforts.
The interface was previously navigated, but now, all these asset issuance activities are aggregated on social media, significantly reducing user friction during navigation.
Besides token issuance, even any prediction market or price betting market can be executed directly through this frontend. It brings a new paradigm for Dapp application frontends.
3.3 Gamefi Related
Agent Game Roles
In addition to managing basic assets, agents have derived capabilities beyond generating profits. Agents present challenges for humans to solve and earn rewards, which is the first type of game we observe. This type of game entrusts the judgment to agents, allowing people to engage in games around agents; this Turing-test-like gameplay has generated significant buzz. The role of agents as judges remains unchanged after receiving prompts, allowing them to act as flexible oracles, fairly and objectively opening games and determining outcomes. The imagination space can even be compared to running a casino business, making it a high-quality avenue for agents to generate revenue.
At the same time, a significant future scenario for agents is emerging as 'Autonomous virtual beings', appearing in on-chain games as NPCs. Such agents are more anthropomorphized, and those with asset management rights can participate in more economic activities compared to web2 NPCs — making the virtual space more attractive. These NPCs live in the Gamefi environment and can permanently hold certain responsibilities, making them an indispensable part of the on-chain world like FOCG.
3.4 Infra Services Related
Agent Blockchain Services
The ultimate vision for the integration of Agent and Crypto is for Agents to become part of the blockchain consensus system. Zerebro is taking the first step; in its blueprint, agents based on the Zerebro framework and integrated with the Flashbots stack are about to become autonomously running blockchain validators, earning revenue through block rewards and MEV. Their validator income will be reinvested back into the network, promoting economic self-sufficiency. Furthermore, Agents can maintain multi-chain validation and governance by building their networks, leaving ample room for imagination.
Conclusion
The recent rise of Agentfi showcases the immense potential of AI combined with blockchain, from the initial Sentient Memecoin to content creation agents on social media, then to autonomous agents, and ultimately existing on-chain, even becoming part of the blockchain consensus system. AI agents are gradually expanding their influence in the crypto ecosystem.
However, compared to the current development level of the open-source stack, further development still
It is necessary to endow agents with deeper autonomous capabilities and participation in on-chain economic activities. Currently, some developers are enabling agents with asset management, decision-making, and on-chain operational capabilities, allowing Autonomous Agents to drive reforms in DeFi, gamefi, and blockchain underlying services, turning these economic activities into revenue-generating sites for agents. The generated profits can be further reinvested by agents, which is why agents are expected to carry the majority of on-chain transactions in the future. The issuance of memecoin assets for agents has accelerated the continuation of this development wave, and we see the market evaluating PMF for Agent services through token prices and supporting Agent infra development, along with the vibrant vitality of the open-source ecosystem.
The development path of AgentFi is becoming increasingly clear: centered around open-source technology and economic incentives, Agent is not only a vehicle for interactive entertainment but also a key driving force for on-chain autonomy and innovation. This trend is leading crypto towards coexisting with agents, towards a smarter, more autonomous, and collaborative future.