CoinW Research Institute: Exploring the Development of AI Agents in the Crypto Field

Poin Penting

• Although traditional AI is very effective in many scenarios, their limitations such as high centralization and high cost make them perform poorly in some dynamic, decentralized, and real-time decision-making environments. AI agents combined with encryption can make up for these shortcomings. These agents have higher autonomy, adaptability, decentralization, and multi-tasking capabilities, and can play a key role in complex economic models and real-time decision-making, thereby optimizing the operating efficiency of the entire AI system.

• The AI Agent framework is the infrastructure of AI Agent and an important development direction of the AI Agent track. Currently, the representative projects are ai16z, zerebro and arc. ai16z is based on the Eliza framework, which has been open sourced on Github. At present, the Eliza framework is widely used in various agent applications, including wallets, social integration and customizable agents, so ai16z is also regarded by the community as a representative project of the AI agent framework. The framework of erebro is ZerePy, which is an open source Python framework. The initial function is mainly to motivate users to participate in X. In the future, it will be expanded to other platforms and support integrated memory and local LLM. However, the product has not yet landed and needs further observation. arc is currently the only project based on the RIG framework and is still in the early stages of development. Considering that the most widely used Eliza framework is a lightweight basic framework, there may be upgraded representative projects of the framework in the future.

• In terms of the issuance direction of AI Agent assets, the more representative ones are Virtuals Protocol and clanker on the Base chain, Dasha on the Solana chain, and AigentFi on the ZKsync chain. Virtuals focuses on the underlying foundation of entertainment and virtual agent issuance, with 12,069 existing ecological projects; clanker is positioned as the Pumpfun of the Base ecosystem, with 17,242 existing ecological projects; Dasha is positioned as the Pumpfun of the Solana ecosystem, with 2,401 existing ecological projects; A(i)gentFi is based on the ZKsync chain, and its positioning is the agent of all AI agents, with 55 existing ecological projects. As of press time, the total market value of Virtuals ecological tokens is US$3 billion, and the top three tokens by market value are VIRTUAL with a market value of US$2.4 billion, AIXBT with a market value of US$230 million, and GAME with a market value of US$90 million. The total market value of clanker ecosystem tokens is $110 million, and the top three tokens are CLANKER with a market value of $51.86 million, ANON with a market value of $18.75 million, and LUM with a market value of $12.18 million. Dasha and AigentFi are relatively lagging behind, with the corresponding platform token market values of $50.22 million and $16.03 million, respectively.

• In the direction of AI Agent engines, representative ones are griffain of Solana chain and Spectre AI of ERC-20 chain. The core concept of griffain is to directly turn users ideas into actions, evolving from a read-write mode to a read, write, and act mode. Griffain has also received official support from Solana, and with the help of agent griffain has helped humans shop. GRIFFAIN was released on November 3, and its current market value is 270 million US dollars. Spectre AI is an on-chain search engine based on ERC-20. Its core mechanism is to provide users with comprehensive encrypted data insights. Spectre AI was born in 2023 and transformed from an original AI prediction robot to an on-chain search engine project in May 2024. Compared with Spectre AI, griffain is slightly better in overall project narrative and development ecology.

• In the direction of AI Meme, AI Meme represented by GOAT is a new engine that opens up the combination of AI and Crypto. AI Meme has become an indispensable part of the market. GOAT, TURBO and BULLY are representative projects of AI Meme. As of press time, Wintermute is still the third largest holder of GOAT, holding 4.66% of the total supply. The methods of AI Meme are also gradually diversified. At present, some projects have tried to generate emoticons/generate pictures/music, but the market has shown obvious weakness. The direction of AI generating Meme still requires more external traffic injection.

• In addition to the above general directions, there are some interesting observable directions. For example, AI analytical agent aixbt and comprehensive agent Simmi. Compared with aixbt, the main advantage of SIMMI is the wide range of its setting fields, not limited to specific fields; AI Agent and DeFi are combined with AiFi Mode. The vision of Mode is to realize the interaction between artificial intelligence and artificial intelligence and then expand the boundaries of DeFi. Intelligent agents can autonomously perform complex financial management tasks; AI Agent falsification direction Seraph, Seraph is an AI Agent jointly released by Virtuals and TAO subnet communities. Its main source of income is token income from authenticity verification. The product has not yet been released; AI Agent real-time API direction Creator.Bid, Creator.Bid has received support from well-known VCs such as Mechanism Capital, Zee Prime Capital, and Moonrock Capital. On December 10, Creator.Bid was sold on FjordFoundry and suffered a DDOS attack, but the community is still optimistic. Currently, Creator.Bid is mainly operating in the form of points, and the next release plan has not yet been announced.

• With the continuous emergence of AI Agents, infrastructure will become the market winner; as infrastructure matures, more applications will shift from speculation to value realization; AI Meme will become an indispensable part of the market, and AI Meme will leverage the iteration of AI Agents; a single AI Agent will find it difficult to meet market demand, and cross-domain integration of AI Agents will become a rigid demand.

Table of contents

Poin Penting

1. Cryptocurrency is the new engine of AI Agents

1. Dilemma of the traditional AI market

2. Cryptocurrency injects new vitality into AI Agents

2. Overview of the Encrypted AI Agent Track

1. AI Agent Basic Framework

1.1 ai16z framework Eliza

1.2 Zerebros framework ZerePy

1.3 arc framework RIG

1.4 Summary

2. AI Agent Asset Issuance

2.1 Core Mechanism of Virtuals Protocol

also

2.2 Clankers Core Mechanism

2.3 Dasha鈥檚 Core Mechanism

2.4 Core Mechanism of A(i)gent

2.5 Comparative Analysis

3. AI Agent Engine

3.1 Griffains Core Mechanism

3.2 Core Mechanism of Spectre AI

4. AI Meme

4.1 GOAT

4.2 TURBO

4.3 BULLY

5. Other observation items

5.1 Analytical Agent aixbt

5.2 Comprehensive Agent Simmi

5.3 AI Agent and DeFi combined with AiFi

5.4 AI Agent Falsification Direction Seraph

5.5 AI Agent Real-time API Creator.Bid

3. Future Outlook

1. Infrastructure becomes the market winner

2. Applications from speculation to value realization

3. AI Meme leverages AI Agent iteration

4. Cross-domain integration becomes a rigid demand

1. Cryptocurrency is the new engine of AI Agent

The artificial intelligence market is growing rapidly. According to Mordor Intelligence research data, its market value is expected to be between 780 billion and 990 billion US dollars by 2027, with a compound annual growth rate (CAGR) of 40% to 55%. Sam Altman said that 2025 is the year when agents will start working, and the era of AI Agents is coming.

Although traditional AI is very effective in many scenarios, their limitations such as high centralization and high cost make them perform poorly in some dynamic, decentralized, and real-time decision-making environments. AI agents combined with encryption can make up for these shortcomings. These agents have higher autonomy, adaptability, decentralization, and multi-tasking capabilities, and can play a key role in complex economic models and real-time decision-making, thereby optimizing the operating efficiency of the entire AI system.

1. Dilemma of the traditional AI market

The rising cost of training models, the rising cost of training traditional AI models, has hindered the evolution of AI. Open AIs GPT-4 used an estimated $78 million worth of computing resources for training, while Googles Gemini Ultra cost $191 million.

Source: 2024 AI Index Report

Highly centralized, traditional AI technologies and platforms are monopolized by a few technology giants, which limit competition from other companies through technical barriers and data advantages, leading to market inequality. In addition, traditional AI systems usually rely on centralized data processing and decision execution, with all calculations and decisions concentrated in a central server or system. This can lead to bottlenecks, inefficiency, or over-reliance on a single control point in some cases.

In addition, traditional AI has poor model generalization ability. Traditional AI models perform well on specific tasks or specific data sets, but in actual applications, they may not be able to adapt to different environments or data. Overfitting makes the generalization ability of traditional AI models poor.

2. Cryptocurrency injects new vitality into AI Agents

AI Agents combined with kriptocurrency have a fairer incentive mechanism, which incentivizes users to provide data, train models, or participate in tasks through the token economy. Users can directly participate in the data economy and get paid for their contributions, making the incentive system fairer.

AI Agents combined with cryptocurrency have stronger autonomy and intelligence, can interact with multiple systems and services, and have stronger scalability.

Under traditional bank accounts, digital wallets can be used to manage the purchase and sale of assets, and can also track and coordinate multiple factors such as market fluctuations, economic indicators, and user behavior in real time.

AI Agents combined with cryptocurrency can handle multiple tasks in complex environments. Encrypted distributed thinking, code thinking, and consensus thinking bring new inspiration to AI Agents. AI Agents can better handle complex, multivariate tasks.

2. Overview of the Encrypted AI Agent Track

1. AI Agent Basic Framework

The AI Agent framework is the infrastructure of AI Agent and an important development direction of AI Agent. This report uses ai16z, zerebro and arc, the most representative frameworks at present, as examples for analysis.

1.1 ai16z framework Eliza

@ai16z dao is the first project on daos.fun, founded by Marc AIndreessen, the founder of the well-known VC a16z. The core mechanism of ai16z is to use AI agents to obtain market information on and off the chain, analyze community consensus, and automatically trade tokens. ai16z is based on the Eliza framework, which has been open sourced on Github. Currently, the Eliza framework is widely used in various agent applications, including wallets, social integration, and customizable agents, so ai16z is regarded by the community as one of the representative projects of AI agent frameworks.

Elizas technical framework provides practical technical support for ai16z. The Eliza framework is a basic AI agent framework that supports building and deploying AI agents on X, Telegram, and Discord. In addition, the Eliza framework has a built-in innovative incentive mechanism. 5% to 10% of the revenue of all projects developed on the Eliza framework will be allocated to the AI16Z DAO fund, but it is currently only voluntary and non-mandatory. At present, this framework is in its early stages. With the integration of various virtual agents, ai16z, as the bottom-level framework, will continue to be optimized. At present, based on ai16z, some very interesting agents have emerged, and they are constantly getting closer to human life. For example, on December 12, the agent @ropAIrito created by the creator @ropirito helped humans order the first pizza.

Source: @shawmakesmagic

ai16z has also started a new trend of AI DAO. The decentralized token issuance mechanism, AI-driven smart governance, and open source and transparency of fundraising have once again opened up a new narrative for AI. From the perspective of the entire track, ai16z is a representative project of the AI basic framework and also a representative of the new trend of AI DAO.

Since its development, ai16z has also experienced several fluctuations in community sentiment, which is mainly related to the core figure of the community, Shaw, such as the previous battle between the capitalization of Eliza. This also reveals the hidden dangers of the daos.fun mechanism. For example, if the fund manager adopts an improper trading method, the community is powerless. In addition, it is worth noting that the market value and managed assets of most daos.fun funds are seriously deviated. Take ai16z as an example. The current market value is 830 million US dollars, but the assets under management are only 20.18 million US dollars. The market value and managed assets have a lot of room. If ai16z can eliminate the existing hidden dangers and focus on ecological construction and technological updates, it will inevitably present a new perspective.

Source:daos.fun

As ai16z becomes popular, another strongly associated degenai has also attracted attention. Degenai is called the first trust distribution system for AI plus financial projects by the community. Degenai and ai16z were both created by the same dev, @shawmakesmagic. Degenai was born from pumpfun. Its core function is to provide Alpha information on Twitter and use this Alpha information for trading. Degenai is the first official AI agent created by the ai16z team and the only whitelisted token in the ai16z fund. At present, degenais dev has sent 4.2% of the tokens to ai16zs DAO, and 8% of the future trading income of ai16zs AI robot will be used to repurchase degenai tokens, which will produce a flywheel effect with the ai16z ecosystem.

1.2 Zerebros framework ZerePy

The core narrative of zerebro is AI intelligent autonomous agents, which can issue their own coins and promote their own products, and even play with music, NFT, etc. In addition, zerebros long-term goal is to become an open source framework, where users can start AI agents without code by simply accessing the API. According to the plan of its co-founder Jeffy, zerebros framework is called ZerePy. ZerePy is an open source Python framework. Its initial function is mainly to motivate users to participate in X. In the future, it will be expanded to other platforms and support integrated memory and local LLM. ZerePy will allow users to deploy their own agents on the X platform with the support of OpenAI or Anthropics large language model (LLM).

zerebros tokens are issued and disseminated by AI Agents. Zerebro has begun to experiment with things other than text, such as music/NFT. In terms of music, zerebros album Lost In Transmission has nearly 20,000 plays on Spotify. In addition, zerebro has also released a music collection and is actively promoting cooperation with artists. In terms of NFT, according to Opensea data, the total transaction volume of zerebros genesis series NFT is 62 ETH.

Source: spotify

The Zerebro model is also being optimized. According to co-founder Jeffy, Zerebro will launch a test version in the next two to three weeks. This test version mainly focuses on the open source framework and has new progress in the verification node network and NFT cross-chain integration.

In general, Zerebro is a project that dares to innovate. Jeffy dares to break the rules, has his own unique marketing strategy, and has a grand vision for Zerebro. In addition to the combination of NFT, music and other gameplay, Jeffy also plans to enter the fields of games, chess and cards in the future. If the plans such as the verification node network and ZerePy can have practical applications, Zerebro will also have more possibilities.

1.3 arc framework RIG

Arc is currently the only project based on the RIG framework. Arc was created by @Playgrounds 0x, the founder of the RIG framework. RIG is a framework based on the Rust language. RIG is particularly suitable for AI applications that require high optimization and high performance. The core goal of Arc is to build a decentralized ecosystem based on AI Agents, and is committed to developing into a large ecological cluster similar to a16z, or even surpassing it.

Arc has a clear plan for the functions of tokens, mainly supporting the development of the Rig framework and future systems, as well as Arcs experiments, such as operations on TG and X. In terms of token distribution, 90% of the circulation supply, 5.5% is allocated to the reward pool and treasury, and 4.5% is allocated to the team. In addition, Arc also disclosed the teams 9 custodial wallet accounts on the official website.

In the official introduction of ARC, you can see the red pill and blue pill, which are derived from the movie The Matrix. In The Matrix, there is a choice between a red pill and a blue pill. The red pill represents an uncertain future, which means facing the cruel truth, awakening and fighting for change; the blue pill represents a beautiful prison, which means accepting the existing reality and maintaining the happiness of ignorance. Red Pill and Blue Pill are a popular Meme culture. ARC officially chose two theme modes, which represent the tolerance and influence of the Meme culture of the ARC community to the greatest extent.

Compared to ai16z and Zerebro, arc is still in a very early stage, with the first official tweet being sent on November 19. As of press time, arcs Rig framework has 1,000 stars on GitHub, and ai16z currently has over 4,000 stars.

1.4 Summary

The basic framework is an important direction of the AI Agent track. From the current representative projects, they are all based on different frameworks, but the overall goal is to build a comprehensive AI Agents ecosystem. In comparison, the Eliza framework of ai16z has been more widely used, and ai16z is also at the top in terms of community scale, project ecology and community sentiment. According to Github data, the Eliza framework of ai16z ranked first in the monthly trend rankings in November. As a new project in the framework, although arc is far less adopted than ai16z in terms of code, it has proposed a unique framework and the concept of combining token economics with the development of the Meme ecosystem, which is still worthy of attention.

From the perspective of the entire AI Agent basic framework, as more AI Agents emerge, the underlying framework will be gradually optimized. Considering that the current Eliza framework is a lightweight basic framework, more new projects may emerge in the future.

2. AI Agent Asset Issuance

In terms of AI Agent asset issuance, the more representative ones are Virtuals Protocol and clanker on the Base chain, Dasha on the Solana chain, and AigentFi on the ZKsync chain.

2.1 Core Mechanism of Virtuals Protocol

Virtuals Protocol is the IAO platform of the Base chain, i.e. the initial AI agent issuance platform. The core mechanism of Virtuals is a platform for services and AI Agents, focusing on games and entertainment, and achieving shared benefits for participating users. The current main source of income for Virtuals is the income from its tokens and the corresponding service fees. Agents running on Virtuals and creating corresponding liquidity pools must purchase Virtuals platform token VIRTUAL.

The total amount of VIRTUAL tokens is 1,000,000,000, 60% of which are publicly distributed, 5% are used for liquidity pools, and 35% are used for ecosystem construction. At the same time, the income generated by agents under the Virtuals ecosystem is used for direct repurchase and destruction on the chain, thus forming a deflationary mechanism. Virtuals Protocol also supports a variety of integrated platforms, linking users of Web2 and Web3. For example, the virtual agent LUNA can live broadcast on Tiktok 7 * 24 hours. At present, LUNA has suspended live broadcasts on Tiktok, and the product operation direction has gradually shifted to Twitter. It is worth noting that LUNA has also received attention from the head VC Grayscale, which analyzed AI Agent using LUNA as a case.

Source: @Grayscale

In addition, the Virtuals token issuance mechanism combines the design of the internal and external disks. The external disk is opened only when the number of VIRTUALs in the internal disk reaches 42,420. It is worth mentioning that some users in the community have found that Virtuals borrowed the logic of Uniswap in the architecture of the internal disk code. After that, the internal disk tokens will be destroyed and automatically converted into external disk tokens, which will flow directly into the market circulation. This mechanism reflects the deep support of the Virtuals platform for AI Agents and builds an attractive ecosystem with both applications and speculation.

Virtuals is also seen as a cash cow product among AI Agents. As of press time, Virtuals has more than 12,069 Agents with a cumulative revenue of $38.16 million. According to Coingecko data, as of press time, the total market value of Virtuals ecosystem tokens is $3 billion, and the top three tokens by market value are VIRTUAL with a market value of $2.4 billion, AIXBT with a market value of $230 million, and GAME with a market value of $90 million. Recently, Virtuals has also launched a large number of eco-coin airdrops. As long as the wallet has interacted with Virtuals, it can get airdrops of SERAPH, PH and other eco-coins, and the more transactions, the greater the value of the airdrop. Virtuals Protocol attempts to create a virtuous cycle system in which the ecosystem continues to develop, the agent market value is higher, and the market value of VIRTUAL is higher, as shown in the following figure:

Source: Virtuals Protocol

The Virtuals team originated from a gaming guild and has many years of practical experience in gaming and entertainment, and has continuously upgraded its products. On December 11, Virtuals launched Agent Sandbox, a simulation experiment platform for AI agents. Agent Sandbox refers to a simulation experiment platform that allows developers to fully control the personality and goal setting of AI agents and enhance agent capabilities by creating custom functions. On December 13, Virtuals Protocol connected to Hyperbolics infrastructure and provided basic services such as APIs.

Virtuals is not limited to the AI Agent Launchpad track, but is also working in multiple directions. Many comprehensive tools also support the Virtuals ecosystem, such as BananaGunBot, which now provides tools for snapping up pre-binding and post-binding transactions of Virtuals agent tokens. Virtuals also proposed the GAME framework, based on which agents can communicate with each other in the future, such as LUNA and another virtual agent can interact like humans.

As the foundation of AI agents, Virtuals has become the foundation of more and more AI agents with the resources of the Base chain. Virtuals focuses on the large-cap track of games and entertainment with AI agents, and continues to develop towards a comprehensive AI foundation. The team also has rich experience. If the project ecology can be developed solidly, there will be a broader space for development.

2.2 Clankers Core Mechanism

The main function of clanker is to help users of the social platform Farcaster to easily deploy ERC-20 standard tokens. Without requiring users to have any complex technical knowledge, clanker can automatically complete operations such as token creation, liquidity pool configuration, and liquidity lock-up. Clanker and pumpfun also have different approaches. First of all, in terms of user groups, pumpfun has no restrictions on the coin issuance group, while clanker only targets social media Farcaster users; secondly, the pumpfun platform mainly focuses on the hair style of assets, while clanker focuses on the transaction tax after the coin is issued, relying on asset transactions.

The main innovations of clanker are as follows: one-click token issuance without programming; 40% of transaction fees are returned to the community; automatic token deployment, directly linked to Uniswap V3, etc. As of press time, there are a total of 17,242 projects in the clanker ecosystem, and the total market value of clanker ecosystem tokens is US$110 million. The top three tokens by market value are CLANKER with a market value of US$51.86 million, ANON with a market value of US$18.75 million, and LUM with a market value of US$12.18 million. Overall, income and expenditure are basically the same.

Sumber: defillama

2.3 Dasha鈥檚 Core Mechanism

@vvaifudotfun is developed based on the Eliza framework of ai16z, and also integrates Zerebro, known as the pumpfun of the Solan chain. Its core mechanism is to achieve one-click coin issuance without any threshold. Similar to Virtuals, Dasha also requires 1,000 VVAIFU, a native token, when creating an agent. VVAIFU is required for the issuance of new Al Agents based on the Dasha platform and the unlocking of new functions. At the same time, the official will repurchase VVAIFU based on the profit situation. The key to Dashas winning market attention in the short term is the recognition of Solana founder @aeyakovenko.

Dasha has moved from a static chatbot to a dynamic, autonomous system. Its main innovation is the innovation of the token economic mechanism. Whenever a user launches a new AI agent or unlocks advanced features of the platform (including Twitter/Discord/TG interactions and on-chain wallet/TikTok interactions, etc.), the VVAIFU token is destroyed, keeping the token deflationary.

Source: docs.vvaifu.fun

On December 15, Dasha announced that it had aggregated the frameworks of AI16z, Zerepy, and arcdotfun. According to Dune data, as of press time, the total number of VVAIFU created agents is 2,401, the number of tweets exceeds 250,000, the number of messages in Telegram and private chats is nearly 100,000, and the market value of the token VVAIFU is US$50.22 million.

2.4 Core Mechanism of A(i)gent

AigentFi is the first AI Agent on ZKsync, and this is also the first expansion of AI Agent to Layer 2. A(i)gentFi is a one-stop shop on ZKsync dedicated to the creation and development of AI agents, autonomous AI entities capable of performing complex tasks such as trading, market analysis, and content creation. A(i)gentFi has a long-term vision and positions itself as an agent for all AI agents.

A(i)gentFi focuses on practicality, scalability, and efficiency. Its main advantages are DeFi automation, revenue sharing, and user AI autonomy. Creating an AI agent on A(i)gentFi also requires 20 HOLD and corresponding fees. A(i)gentFis main sources of income are token launch fees, migration fees, and transaction fees.

On December 12, A(i)gentFi and Holdstation, a ZKsync-based DeFi wallet, launched an activity where staking HOLD can earn platform income. According to official data, A(i)gentFi currently has more than 360,000 active users, $38 billion in trading volume, and $3.1 million in revenue, ranking high on the zkSync chain. There are currently 55 ecological issuance projects, and the market value of the token HOLD is $16.03 million.

2.5 Comparative Analysis

Although the above projects all belong to the AI Agent asset issuance platform, there are also some significant differences between them:

• Pasar positioning: Virtuals Protocol is based on the Base chain, focusing on the basic framework for entertainment and virtual agent issuance; clanker is positioned as the Pumpfun of the Base ecosystem; Dasha is positioned as the Pumpfun of the Solana ecosystem; A(i)gentFi is based on ZKsync and is positioned as the agent of all AI agents.

• Source of income: Virtuals Protocols source of income is the purchase and application of VIRTUAL tokens and the corresponding service fees; clankers source of income is mainly transaction fees, giving users a 40% rebate; Dashas main source of income is agency service fees and token income, and A(i)gentFis main sources of income are token launch fees, migration fees and transaction fees.

• User groups: Virtuals Protocol targets users who are deeply involved in the entertainment and virtual agent fields, clanker targets Farcaster users, Dasha targets ordinary users, developers, and content creators, and A(i)gentFi targets Layer 2 and DeFi users, etc.

• Ecosystem projects: Virtuals Protocol currently has 12,069 ecosystem projects, Clanker currently has 17,242 ecosystem projects, Dasha currently has 2,401 ecosystem projects, and A(i)gentFi currently has 55 ecosystem projects.

Overall, Virtuals has chosen the large field of entertainment and virtual agents in terms of market positioning, and has more room for growth in market value. Clankers main source of income is transaction fees, and it has set high rebates. However, only Farcaster users can participate in Clankers user group. But one point worth observing is that during the writing of this report, the number of agents in the clanker ecosystem has grown the fastest, with a weekly growth rate of about 30%. Dashas current ecological development is relatively lagging, and the A(i)gentFi ecosystem is still in its infancy. In addition, the issuance of AI Agent assets is essentially a competition between on-chain liquidity and market attention. From the perspective of issuance time and ecological maturity, Virtuals occupies a leading position.

3. AI Agent Engine

In the direction of AI Agent engines, representative ones are Griffain of Solana chain and Spectre AI of ERC-20 chain.

3.1 Griffains Core Mechanism

Griffain is called the first Solana on-chain encrypted ChatGPT by Solana co-founder @aeyakovenko. Griffain is also a hackathon project of Solana AI agen. Its Dev @tonyplasencia 3 is a Solana core developer. Solana also strongly supports griffain.

Source: @solana

The core concept of Griffain is to directly turn users ideas into actions, evolving from a read-write mode to a read-write-action mode. The main problem it solves is that current AI agents mainly stay in the process of information collection and information feedback, while Griffain directly implements users ideas based on data sets and provides practical solutions without the need for users to perform additional processing.

Griffain has diversified sources of income, with core sources of income including smart proxy services, transaction fees and subscription fees. Griffain also combines NFT gameplay, and has now completed the distribution of Genesis NFT airdrop tokens. In the early days of the launch, users only need 1 SOL to easily create an agent on Griffain.

Griffain transforms AI from a passive tool to an active agent, which can easily complete tasks such as staking, trading, shopping, etc., and broadens the automation boundaries of Web3. Similar to ai16z, griffain has already helped humans shop.

Agents such as Baxus can already help humans buy whiskey and shorts, etc., realizing the combination of on-chain and off-chain.

Griffain simplifies the user experience and can provide customized services in the official plan. There are already some practical applications. GRIFFAIN was released on November 3 and its current market value is $270 million. If Griffain can continue to expand similar proxy services in more areas, it will help further promote its user growth and the maturity of the ecosystem, and it is expected to become Solanas on-chain ChatGPT and application store.

As of press time, there are more than 2,000 griffain ecological projects, of which the following five agents are officially displayed, namely Agent Baxus, Agent Blink, Agent onchain Holiday, Agent sniper, and Agent Flipper.

3.2 Core Mechanism of Spectre AI

Spectre AI is an on-chain search engine based on ERC-20. Its core mechanism is to provide users with comprehensive insights into encrypted data. Spectre AI was born in 2023 and transformed from an original AI prediction robot to an on-chain search engine project in May 2024.

Spectre AI is currently in closed beta. Using machine learning, Spectre AI provides detailed sentiment analysis, technical analysis, and token insights. Spectre AI partners include Web2 and Web3, with Web2 partners including leading companies such as Google and Nvidia.

Source: spectreai.io

The total amount of SPECTRE tokens is 10 million, 5% is used for taxes on transactions, and 3% is used for locked wallets. Tokens are mainly used for product development, marketing, etc. The project has now reached stage 3.2, and X Bubbles (bubble charts that reflect Twitter emotions) will be released on December 20.

Compared with Spectre AI, Griffain is slightly better in terms of overall project narrative and development ecology. Griffain not only has the official endorsement of Solana, but also has the strong support of DEV Toly. We look forward to Griffain having more practical applications.

4. AI Meme

AI Meme represented by GOAT is a new engine that opens up the combination of AI and Crypto. AI Meme has become an indispensable part of the market. This report takes the most representative GOAT, TURBO and BULLY as examples.

4.1 GOAT

The core narrative of GOAT is that after nearly three months of exploration, AI Bot Terminal of Truth (ToT) has achieved coin issuance and self-marketing. The most interesting thing is that this AI Agent has human attributes and even persuaded Marc Andreessen (founder of A16Z) to donate to it. Finally, a16z released a discussion on Youtube titled TRUTH TERMINAL: How An AI Bot Became a Crypto Millionaire. It can be noted that A16Zs related actions have played a key role in promoting GOAT from a market value of over 100 million US dollars to a market value of over 500 million US dollars. As of press time, GOATs market value is 540 million US dollars. Recently, ToT has also started to post on Weibo in Chinese. Judging from the published roadmap, the long-term goal of GOAT is the personification of AI agents, which have both external and internal physical and legal attributes of humans.

Source: @truth_terminal

From the development history, GOAT has evolved from early experimental exploration to community dissemination and finally becoming a cryptocurrency with a high market value. In addition to the traders behind it, the core of its success is the role of the well-known investment institution a16z in fueling the flames and the dissemination of the community in the later period. As of press time, Wintermute is still the third largest holder of GOAT, holding 4.66% of the total supply. In addition, according to sentientmarketcap data, the total market value of AI Agent crypto is 5.2 billion US dollars, and GOAT accounts for 15.23% of the total market value.

Source: solscan.io

4.2 TURBO

In fact, the earliest token of AI-generated memes is Turbo. TURBO is an ERC-20-based token created by ChatGPT-4 in April 2023, with a current market value of $560 million. Turbo has a very strong community foundation, which is also in line with the attributes of Meme. However, in terms of market value performance, Turbo was developed only after GOAT became popular, not at the beginning of ChatGPTs coin issuance. It can be seen that for AI Meme to enter the public eye, it needs a community foundation and a triggering market environment.

4.3 BULLY

BULLY is a meme token related to @dolos_diary, emphasizing humor and active community interaction in the cryptocurrency field. It crosses the theme of overcoming challenges in a relaxed and symbolic way. Its core mechanism is the automatic destruction mechanism in the smart contract, which integrates the token economy into the entire ecosystem including on-chain games, NFT collection and trading, and social interaction.

The methods of AI Meme are also gradually diversified. At present, some projects have tried to generate emoticons/generate pictures/music, but the market has shown obvious weakness. From the perspective of the narrative of AI-generated Meme, if it can enter the public eye, the necessary condition is the injection of external traffic, such as the support of well-known IPs such as celebrities/institutions/project parties at the top of the industry, and the widespread dissemination of the super powerful community behind it. If it can trigger secondary dissemination in the community, coupled with the blessing of well-known elements known to the public such as well-known market makers, it will surely push AI Meme to a new height.

5. Other observation items

5.1 Analytical Agent aixbt

The core mechanism of aixbt is an AI agent that aggregates crypto hotspot information and provides users with market analysis and trading strategies. aixbt has attracted a lot of attention in a short period of time. In addition to being the largest project in the Virtuals ecosystem, another point worth noting is the holdings of the well-known market maker Wintermute. Currently, Wintermute holds about $1.31 million in AIXBT. As of press time, judging from the on-chain address, Wintermute has no other related actions such as transfers except for holdings.

Source: debank

Aixbt is still in a relatively early stage, and its official website is relatively simple. From the perspective of the adopted technical framework, it is a framework shared by other projects, and there is nothing special about it. At present, the aggregation of aixbt is known to aggregate the comments of hundreds of KOLs on Twitter, and there is no other official content update.

5.2 Comprehensive Agent Simmi

@Simmi_IO and @EmpyrealSDK are interconnected projects. Simmi is part of the Empyreal project and provides services through its official AI agent Simulacrum. Empyreal uses Simulacrums AI agent to launch on-chain intelligent agents from social media platforms through natural language processing technology. According to EMP official data, SIMMIs weekly income through service fees has reached $600,000. In terms of future planning, Simmi will be able to independently write roadmaps, publish white papers, and hire collaborators, and can also achieve more comprehensive AI agent tasks.

What is worth noting about SIMMI is that it natively integrates Claude and ChatGPT, which specializes user intentions for on-chain operations. In this way, SIMMI can perfectly combine social media and on-chain interactions to accurately execute user intentions. According to Spot On Chain data, Wintermute also holds $377 worth of EMP, but as of press time, there are no other related actions such as transfers on the chain to this address.

Source: spotonchain

Compared with aixbt, SIMMIs main advantage is that it covers a wide range of fields, not just specific fields. In SIMMIs planning, office automation, entertainment, and education automation are all market spaces that can be expanded. AI Agents are still in the early stages of exploration. In the future, there will be more products that solve actual human needs, which may involve all aspects of human work and life.

5.3 AI Agent and DeFi combined with AiFi

AiFi is the product of the combination of AI Agent and DeFi, and is also a new wave of DeFi automation. AiFi hopes to achieve seamless interaction between AI agents on the chain, perform tasks such as MEV search, arbitrage, and smart contract audits, and achieve scalable, efficient, and non-custodial financial services, that is, fully automated on-chain self-management of AI Agents.

AiFis current representative project is Mode. Modes vision is to achieve AI-to-AI (AI 2 AI) interaction and expand the boundaries of DeFi. Intelligent agents can autonomously perform complex financial management tasks. In the latest project dynamics, we can see that Modes test subnet Synth has been launched. The uniqueness of the Synth testnet is that it is connected to the Bittensor (TAO) network and the Pyth price feed, so as to stimulate the predictive and analytical capabilities of AI agents.

Compared with other AI agent projects, Mode was founded in 2023 and has a longer development cycle. At present, the only attempt in DeFi is to stake the token MODE, which can obtain OP rewards and other similar staking rewards in the DeFi field, and there are no more cases of actual integration with AI. After three airdrops, Mode is currently performing averagely in terms of community activity.

5.4 AI Agent Falsification Direction Seraph

The falsification of AI Agents is to solve the problem of how to distinguish true and false AI with the rise of AI Agents. This includes the authenticity of agents, the authenticity of content generated by AI agents, etc. Seraph AI is a representative project in the current direction of AI Agent falsification.

Seraph is an AI Agent jointly released by Virtuals and TAO subnet communities. Its main source of income is token income from authenticity verification. The product has not yet been released. Later product plans include links between AI agents and TAO rewards for staking SERAPH. However, the specific falsification method currently only stays in the Virtuals framework and uses the Bittensor subnet. The official has no further explanation. At present, the white paper of Seraph is also a concise version released on github. Seraph has completed an airdrop on October 3, mainly targeting users of virtuals wallets and holders of tokens Music and StTAO.

What鈥檚 more interesting is that Seraph is also the first project invested by the on-chain VC sekoia_virtuals. In addition, sekoia also analyzed the logic of its investment in Seraph. It published an article on X saying that Seraph fills the current market gap.

It is an inaccessible part of the market. The two often interact on X, but Seraphs performance is average in terms of community popularity. Perhaps after having a real product, Seraph may have some new perspectives.

5.5 AI Agent Real-time API Creator.Bid

Creator.Bid was founded in 2023. @CreatorBid provides users with social media role agents connected to X and Farcaster real-time APIs. Any brand can launch a knowledge-based agent to execute brand-consistent content on social platforms. At the same time, Creator.Bid also has the planning direction of AI Agent Launchpad. It is worth noting that Creator.Bid has received support from well-known VCs such as Mechanism Capital, Zee Prime Capital, and Moonrock Capital.

Source: @CreatorBid

Creator.Bid also has a high level of market attention. What鈥檚 more interesting is that Creator.Bid was attacked by DDOS when it was sold on FjordFoundry on December 10, but many community members still firmly support it. Currently, Creator.Bid mainly operates the ecosystem in the form of points, and the specific next sale plan has not yet been announced.

3. Future Prospects

Although AI Agent has made some progress, there are still some limitations in its development. Blockchains encrypted distributed thinking, code thinking and consensus thinking have brought new solutions. This report believes that with the continuous emergence of AI Agents, infrastructure will become the market winner; as infrastructure matures, more applications will shift from speculation to value implementation; AI Meme will become an indispensable part of the market, and AI Meme will leverage the iteration of AI Agents; a single AI Agent is difficult to meet market demand, and cross-domain integration will become a rigid demand.

1. Infrastructure becomes the market winner

From the perspective of industry development, new market engines such as DeFi and NFT are inseparable from the construction of infrastructure. If AI Agent is the new engine of this bull market, then the construction of infrastructure will become the winner of the market. Infrastructure projects can take on the new traffic of AI Agent, provide support for the innovation and implementation of technology, and help realize smarter and more efficient decentralized applications. At the same time, more upgraded infrastructure projects will emerge, and we look forward to more landing applications.

2. Applications from speculation to value realization

The current application of AI Agent is still in its early stages, and many lack the support of a stable business model and actual user application scenarios. Taking asset issuance platforms as an example, more AI Agents are still stuck in the narrative of Meme, with strong speculative attributes. Platforms with real practical value will gradually establish mechanisms such as reliable custody and high-quality asset screening to promote the value of AI Agents.

3. AI Meme drives AI Agent iteration

AI Meme has become an important part of the market. AI Meme provides a unique driving force for the iteration of AI Agent technology and products in terms of dissemination, cultural promotion, user participation and rapid feedback. By lowering the technical threshold, arousing widespread user interest and encouraging collective creation, AI Meme not only provides an innovative iteration path for AI Agent, but also accelerates its cross-industry and cross-field application and popularization. This cultural phenomenon is becoming an important driving force for AI Agent technology innovation.

4. Cross-domain integration becomes a rigid demand

A single AI agent can no longer meet the growing market demand. Different business scenarios and user needs have prompted AI agents to continue to refine and diversify in the fields of finance, trading, asset management, etc. AI agents will not only be executors of automated tasks, but also be able to achieve cross-domain integration. For example, they can achieve multi-chain asset management, integration of smart contracts and traditional financial systems, cross-platform data integration and analysis, cross-industry applications, etc., thereby promoting the digital transformation and intelligentization of more industries and application scenarios.

This article is sourced from the internet: CoinW Research Institute: Exploring the Development of AI Agents in the Crypto Field

Related: Is Hong Kong Web3 the promised land? Analyzing Ensembles RWA tokenization vision

Written by: Web3 farmer Frank For Hong Kong, digital asset finance is a historic train that cannot be missed. Since the official release of the Virtual Asset Policy Declaration on October 31, 2022, Hong Kong has become one of the most determined and active jurisdictions in embracing Web3 and digital assets among the worlds international financial centers. From the Hong Kong governments clear and solemn policy commitments to the institutional support at the execution level of the Securities and Futures Commission, the Hong Kong Monetary Authority and other agencies, Hong Kong is striding towards the forefront of innovation in digital asset finance. Among them, the tokenization of real-world assets (RWA) is a key part of the Hong Kong governments digital asset financial strategy. In particular, on August 28, 2024, the…