ArkStream Capital の追跡調査レポート: AI エージェントは Web3+AI の救世主となることができるか?

要約

• The popular and mature types of AI Agent projects in Web2 startups are mainly enterprise-side services, while in the Web3 field, model training and platform collection projects have become mainstream due to their key role in building an ecosystem.

• Currently, there are not many AI Agent projects in Web3, accounting for 8%, but their market value in the AI track accounts for as much as 23%, thus demonstrating strong market competitiveness. We expect that as the technology matures and market recognition increases, there will be multiple projects with a valuation of more than US$1 billion in the future.

• For Web3 projects, the introduction of AI technology may become a strategic advantage for non-AI-core application-side products. For AI Agent projects, the combination should focus on the construction of the entire ecosystem and the design of token economic models to promote decentralization and network effects.

The AI Wave: Current Status of Projects Emerging One after Another and Valuations Rising

Since ChatGPT was launched in November 2022, it has attracted more than 100 million users in just two months. By May 2024, ChatGPTs monthly revenue has reached a staggering $20.3 million. After the release of ChatGPT, OpenAI quickly launched iterative versions such as GPT-4 and GP 4-4 o. With such a rapid trend, major traditional technology giants have realized the importance of the most cutting-edge AI model applications such as LLM, and have launched their own AI models and applications. For example, Google released the large language model PaLM 2, Meta launched Llama 3, and Chinese companies launched large models such as Wenxin Yiyan and Zhipu Qingyan. Obviously, the field of AI has become a battleground.
The competition among major technology giants has not only promoted the development of commercial applications, but also our survey statistics on open source AI research have found that the 2024 AI Index report shows that the number of AI-related projects on GitHub has surged from 845 in 2011 to about 1.8 million in 2023. Especially in 2023 after the release of GPT, the number of projects increased by 59.3% year-on-year, reflecting the global developer communitys enthusiasm for AI research.
The enthusiasm for AI technology is directly reflected in the investment market, which has shown strong growth and explosive growth in the second quarter of 2024. There are 16 AI-related investments of more than $150 million worldwide, twice as many as in the first quarter. The total financing of AI startups has soared to $24 billion, more than doubling year-on-year. Among them, Musks xAI has raised $6 billion with a valuation of $24 billion, becoming the second most valuable AI startup after OpenAI.

2024 Q2 AI track financing TOP 10, source: Yiou, https://www.iyiou.com/data/202407171072366

The rapid development of AI technology is reshaping the landscape of the technology sector at an unprecedented speed. From the fierce competition among technology giants, to the booming development of open source community projects, to the capital markets enthusiastic pursuit of AI concepts. Projects are emerging in an endless stream, investment amounts have repeatedly set new highs, and valuations have also risen accordingly. Overall, the AI market is in a golden period of rapid development, and large language models and retrieval-enhanced generation technologies have achieved significant progress in the field of language processing. Despite this, these models still face challenges in converting technical advantages into actual products, such as uncertainty in model output, the risk of generating illusions of inaccurate information, and model transparency issues. These issues become particularly important in application scenarios with extremely high reliability requirements.

In this context, we began to study AI Agent, because AI Agent emphasizes the comprehensiveness of solving practical problems and interacting with the environment. This transformation marks the evolution of AI technology from a pure language model to an intelligent system that can truly understand, learn and solve real problems. So we see hope in the development of AI Agent, which is gradually bridging the gap between AI technology and practical problem solving. The evolution of AI technology is constantly reshaping the architecture of productivity, while Web3 technology is reshaping the production relations of the digital economy. When the three elements of AI: data, models and computing power, are integrated with the core concepts of Web3 such as decentralization, token economy and smart contracts, we foresee that a series of innovative applications will be born. In this cross-cutting field full of potential, we believe that AI Agent, with its ability to perform tasks autonomously, has shown great potential for large-scale applications. To this end, we began to conduct in-depth research on the diverse applications of AI Agent in Web3, from the infrastructure, middleware, application level of Web3 to multiple dimensions such as data and model markets, aiming to identify and evaluate the most promising project types and application scenarios to deeply understand the deep integration of AI and Web3.

Concept clarification: Introduction and classification overview of AI Agent

Basic Introduction

Before introducing AI Agent, in order to help readers better understand the difference between its definition and the model itself, we use a real-world scenario as an example: suppose you are planning a trip. Traditional large language models provide destination information and travel suggestions. Retrieval-enhanced generation technology can provide richer and more specific destination content. AI Agent is like Jarvis in the Iron Man movie, who can understand the needs, actively search for flights and hotels based on your words, perform reservation operations, and add the itinerary to the calendar.
The general definition of AI Agent in the industry is that it is an intelligent system that can perceive the environment and take corresponding actions, obtain environmental information through sensors, and affect the environment through actuators after processing (Stuart Russell Peter Norvig, 2020). We believe that AI Agent is an assistant that combines LLM, RAG, memory, task planning and tool use capabilities. It can not only provide information, but also plan, decompose tasks, and actually execute them.
Based on this definition and characteristics, we can find that AI Agents have long been integrated into our lives and applied in different scenarios. For example, AlphaGo, Siri, Teslas L5 and above autonomous driving can all be regarded as examples of AI Agents. The common feature of these systems is that they can perceive external user input and respond accordingly to affect the real environment.
Taking ChatGPT as an example to clarify the concept, we should clearly point out that Transformer is the technical architecture that constitutes the AI model, GPT is a model series developed based on this architecture, and GPT-1, GPT-4, and GPT-4 o represent versions of the model at different stages of development. ChatGP is an AI Agent evolved from the GPT model.

Classification Overview

At present, the AI Agent market has not yet formed a unified classification standard. We have divided the 204 AI Agent projects in the Web2+Web3 market into primary and secondary categories according to the significant labels corresponding to each project. The primary classification is divided into three categories: infrastructure construction, content generation, and user interaction, which are further subdivided according to their actual use cases:

Infrastructure: This category focuses on building the lower-level content in the Agent field, including platforms, models, data, development tools, and more mature B-side services for underlying applications.

• Development tool class: Provides developers with auxiliary tools and frameworks for building AI Agents.

• Data processing: Processing and analyzing data in different formats, mainly used to assist decision-making and provide sources for training.

• Model training: Provides AI-specific model training services, including reasoning, model building, and settings.

• B-side services: mainly for corporate users, providing enterprise services, vertical services and automation solutions.

• Platform collection category: a platform that integrates multiple AI Agent services and tools.

Interactive Agents: Similar to content generation agents, the difference is continuous two-way interaction. Interactive Agents not only accept and understand user needs, but also provide feedback through technologies such as natural language processing (NLP) to achieve two-way interaction with users.

• Emotional companionship: AI Agents that provide emotional support and companionship.

• GPT class: AI Agent based on the GPT (Generative Pre-trained Transformer) model.

• Search type: Agents that focus on search functions and provide more accurate information retrieval.

Content generation: This type of project focuses on creating content, using big model technology to generate various forms of content based on user instructions. It is divided into four categories: text generation, image generation, video generation, and audio generation.

Analysis of the current development status of Web2 AI Agent

According to our statistics, the development of AI Agents in the traditional Web2 Internet shows a clear trend of sector concentration. Specifically, about two-thirds of the projects are concentrated in the infrastructure category, mainly B-side services and development tools. We have also conducted some analysis on this phenomenon.

• The impact of technology maturity: The reason why infrastructure projects dominate is first of all due to their technology maturity. These projects are usually built on time-tested technologies and frameworks, which reduces the difficulty and risk of development. They are equivalent to the shovel in the field of AI, providing a solid foundation for the development and application of AI Agents.

• Driven by market demand: Another key factor is market demand. Compared with the consumer market, the enterprise market has a more urgent need for AI technology, especially in seeking solutions to improve operational efficiency and reduce costs. At the same time, for developers, the cash flow from enterprises is relatively stable, which is conducive to their development of subsequent projects.

• Limitation of application scenarios: At the same time, we noticed that the application scenarios of content generation AI in the B-end market are relatively limited. Due to the instability of its output, enterprises prefer applications that can stably improve productivity. This has led to a small proportion of content generation AI in the project library.

This trend reflects the actual considerations of technology maturity, market demand, and application scenarios. As AI technology continues to advance and market demand becomes clearer, we expect this pattern to adjust, but infrastructure will still be the solid foundation for the development of AI Agents.

Analysis of Web2s leading AI Agent projects

Web2 AI Agent leading projects, source: ArkStream project database

We take an in-depth look at some of the current AI Agent projects on the Web2 market and analyze them, taking Character AI, Perplexity AI, and Midjourney as examples.

Character AI:

• Product description: Character.AI provides AI-based dialogue systems and virtual character creation tools. Its platform allows users to create, train and interact with virtual characters that are able to engage in natural language conversations and perform specific tasks.

• Data analysis: Character.AI had 277 million visits in May and more than 3.5 million daily active users, most of whom are between 18 and 34 years old, showing the characteristics of a young user group. Character AI performed well in the capital market, completing a $150 million financing round with a valuation of $1 billion, led by a16z.

• Technical analysis: Character AI has signed a non-exclusive licensing agreement with Googles parent company Alphabet to use its large language model, which indicates that Character AI uses self-developed technology. It is worth mentioning that the companys founders Noam Shazeer and Daniel De Freitas have participated in the development of Googles conversational language model Llama.

Perplexity AI:

• Product introduction: Perplexity can crawl and provide detailed answers from the Internet. It ensures the reliability and accuracy of information through citations and reference links. At the same time, it educates and guides users to ask questions and search for keywords, meeting users diverse query needs.

• Data analysis: Perplexity has reached 10 million monthly active users, and the number of visits to its mobile and desktop applications increased by 8.6% in February, attracting about 50 million users. In the capital market, Perplexity AI recently announced that it had received $62.7 million in financing, with a valuation of $1.04 billion, led by Daniel Gross, and with participation from Stan Druckenmiller and NVIDIA.

• Technical analysis: The main model used by Perplexity is the fine-tuned GPT-3.5, as well as two large models fine-tuned based on open source large models: pplx-7 b-online and pplx-70 b-online. The model is suitable for professional academic research and vertical field queries to ensure the authenticity and reliability of the information.

Midjourney:

• Product introduction: Users can use Prompts to create images of various styles and themes in Midjourney, covering a wide range of creative needs from realism to abstraction. The platform also provides image mixing and editing, allowing users to perform image overlay and style transfer. The platforms real-time generation function ensures that users can get the generated image within tens of seconds to minutes.

• Data analysis: The platform already has 15 million registered users and 1.5 to 2.5 million active users. At the same time, according to public market information, Midjourney did not take money from investment institutions, relying on the reputation and resources of founder Davids multiple entrepreneurial experiences to achieve self-sufficient development.

• Technical analysis: Midjourney uses their own closed-source model, and the platform has been using a diffusion-based generative AI model since the release of Midjourney V4 in August 2022. It is said that the training parameters of the model are between 30 and 40 billion, and this huge number of parameters provides a solid foundation for the diversity and accuracy of the images it generates.

Commercialization Dilemma

After experiencing multiple Web2 AI Agents, we observed the general path of product iteration: from focusing on a single, refined task in the early stages to expanding capabilities to handle more complex multi-task scenarios in the later stages. This trend not only reflects the potential of AI Agents in improving work efficiency and innovation, but also indicates that they will play a more critical role in the future. Through preliminary statistics of 125 AI Agent projects in Web2, we found that the projects are mainly concentrated in content generation (such as Jasper AI), development tools (such as Replit), and the largest number of B-side services (such as Cresta). This finding is contrary to our expectations. At first, we expected that with the increasing maturity of AI model technology, the C-end market would usher in an explosive growth of AI Agents. However, after analysis, we realized that the commercialization path of C-end AI Agents is far more rugged and complicated than expected.

Take Character.Ai as an example. On the one hand, Character.AI has the best traffic performance. However, due to its single business model, it relies on a subscription fee of 9.9 USD. In the face of a small amount of subscription income and heavy user reasoning cost consumption, the entire team was eventually acquired by Google due to difficulties in traffic monetization and capital chain problems. This case reflects that even with such good traffic and financing, C-end AI Agent applications are still difficult in the commercialization process. It reflects that the vast majority of products have not yet reached the standard of replacing or effectively assisting manual labor, resulting in C-end users not having a strong willingness to pay for current products. In our actual research and investigation, we found that many start-up projects have encountered similar problems to Character.ai. The development of C-end AI Agent is not smooth sailing, but requires more in-depth exploration of technology maturity, product value, and business model innovation to realize its potential and value in the C-end market.

By counting the valuations of most AI Agent projects, there is still nearly 10-50 times of room for valuations compared to ceiling projects such as OpenAI and xAI. It is undeniable that the ceiling of C-end Agent applications is still high enough, proving that it is still a good track. But based on the above analysis, we believe that compared to the C-end, the B-end market may be the final destination of AI Agent. By building a platform, enterprises integrate AI Agent into vertical fields, CRM, office OA and other management software, which not only brings improved operational efficiency to enterprises, but also provides a broader application space for AI Agent. Therefore, we have reason to believe that B-end services will be the main direction of development of AI Agent in the traditional Internet of Web2 in the short term.

Analysis of the current status and prospects of Web3 AI Agent development

Project 概要

According to the analysis above, even AI Agent applications that have received top-level financing and have good user traffic are facing the problem of commercial realization. Next, we will analyze the current development of AI Agent projects in Web3 in depth. By evaluating a series of representative projects – including their technological innovation, market performance, user feedback, and development potential, we aim to dig out inspiring suggestions. The following figure shows several representative projects that have issued tokens and have a high market value:

Web2 AI Agent leading projects, source: ArkStream project database

According to our statistics on the Web3 AI Agent market, the types of project development also show an obvious trend of plate concentration. The vast majority of projects are classified as infrastructure, and there are relatively few content generation projects. Most projects try to solve the model training needs of the project party by allowing users to provide distributed data, computing power, etc. Or try to build a one-stop platform that embeds a variety of AI Agent application services and tools. From development tools to front-end interactive applications, generative applications, etc. The traditional AI Agent industry is currently mainly limited to open source parameter adjustment or applying existing models to build applications. This approach has failed to form a significant network effect at the enterprise and individual user levels.

Current situation analysis

We believe that this phenomenon may be driven by the following factors at this stage:

市場 and technology mismatch: The combination of Web3 and AI Agent currently has no obvious advantages over the traditional market. Its real advantage lies in improving production relations and optimizing resources and collaboration in a decentralized manner. This may lead to some interactive and generative applications lacking competitiveness in front of traditional competitors with stronger technical and financial strength.

Application scenario limitations: In the Web3 environment, there may not be so many practical needs to generate images, videos, text, etc. On the contrary, the decentralized and distributed characteristics of Web3 are more often used to reduce costs and increase efficiency in the traditional AI field, rather than to expand new application scenarios.

We believe that the root of this phenomenon can be traced back to the current development status of the AI industry and its future direction. This may be because the current AI technology is still in its infancy, similar to the transition period when steam engines were replaced by electric motors in the early days of the Industrial Revolution, and has not yet reached the era of widespread electrification.

We have reason to believe that the future development trend of AI may follow a similar path. General models will gradually solidify, while fine-tuning models will show diversified development. AI applications will be widely dispersed among various enterprises and individual users, and the focus will shift to the interconnection and interaction between models. This trend is highly consistent with the concept of Web3, because Web3 is known for its composability and permissionless features, which coincides with the concept of decentralized model fine-tuning. Developers are allowed greater freedom to combine and adjust various models more freely. At the same time, the decentralized feature also brings unique advantages to model training in terms of data privacy protection and computing resource allocation.

With the advancement of technology, especially the emergence of new technologies such as LoRA (Low-Rank Adaptation), the cost and technical threshold of model fine-tuning have been greatly reduced. This makes it easier to develop public models for specific scenarios or to meet the personalized needs of users. AI Agent projects within Web3 can make full use of this technological advancement to explore novel training methods, innovative incentive mechanisms, and new models of model sharing and collaboration in the field of model training and fine-tuning, which are often difficult to achieve in traditional centralized systems.

In addition, the concentration of Web3 projects in model training also reflects their strategic considerations of occupying an important position in the entire AI ecosystem. Therefore, AI Agent projects in the Web3 industry are concentrated in the field of model training, which is a natural intersection of technology development trends, market demand and Web3 industry advantages. Next, we will cite several model training projects in the Web23 industry and compare them.

Model training projects

Humans.ai

• Project introduction: Humans.ai is a diversified AI algorithm model library and training deployment environment, covering multiple fields such as images, videos, audio, text, etc. The platform not only supports developers to further train and optimize the models, but also allows them to share and trade their own models. A significant innovation is that Humans.ai uses NFT as a medium to store AI models and user biometric data, making the process of AI content creation more personalized and secure.

• Data analysis: Humans.ai The market value of its token Heart is about $68 million. Twitter followers are 56k, and its user data is not disclosed.

• Technical Analysis: Humans.ai does not develop its own models, but adopts a modular approach, encapsulating all provided models into NFTs, providing users with a flexible and scalable AI solution.

FLock.io

• Project Introduction: FLock.io is an AI co-creation platform based on federated learning technology (a decentralized machine learning method that emphasizes data privacy). It is committed to solving the pain points in the AI track, such as low public participation, insufficient privacy protection, and the monopoly of AI technology by large companies. The platform allows users to contribute data while protecting privacy, promoting the democratization and decentralization of AI technology.

• Data Analysis: In early 2024, it completed a $6 million seed round of financing, led by Lightspeed Faction and Tagus Capital, and followed by DCG, OKX Ventures and other institutions.

• Technical Analysis: FLock.ios technical architecture is based on federated learning, a method that promotes decentralization while protecting data privacy. In addition, FLock.io also uses technologies such as zkFL, homomorphic encryption, and secure multi-party computing (SMPC) to provide additional protection for data privacy.

• This is a model training project for AI Agent in the Web3 industry. There are also similar platforms in Web2 that provide model training services, such as Predibase.

Predibase

• Project introduction: Predibase focuses on AI and large language model optimization, allowing users to fine-tune and deploy open source large language models such as Llama, CodeLlama, Phi, etc. The platform supports multiple optimization techniques such as quantization, low-rank adaptation, and memory-efficient distributed training.

• Data Analysis: Predibase announced the completion of a $12.2 million Series A funding round led by Felicis. Large companies such as Uber, Apple, Meta and startups such as Paradigm and Koble.ai are all users of the platform.

• Technical analysis: Predibase users have trained more than 250 models. The platform currently adopts the LoRAX architecture and Ludwig framework: LoRAX enables users to serve thousands of fine-tuned LLMs on a single GPU, greatly reducing costs without affecting throughput or latency. Ludwig is a declarative framework that Predibase uses to develop, train, fine-tune, and deploy state-of-the-art deep learning and large language models.

• Project Analysis: The Predibase platform is user-friendly and provides customized AI application building services for users at different levels, whether for C-end and B-end users, or for novices or experienced people in the AI field.

For novices, the one-click automation function of the Predibase platform simplifies the model building and training process. The platform automatically completes complex building and deployment steps. For experienced users, it provides more in-depth customization options. Not only does it provide automation services, but it also allows users to access and adjust more professional parameter settings. When comparing traditional AI model training platforms with AI projects in the Web3 field, although they may be similar in overall framework and logic, we found that there are some significant differences between the two in technical architecture and business model.

• Technical depth and innovation: Traditional AI model training platforms often adopt deeper technical barriers, such as using self-developed technologies such as the LoRAX architecture and the Ludwig framework. These frameworks provide powerful functions that enable the platform to handle complex AI model training tasks. However, Web3 projects may focus more on decentralization and openness without in-depth technology.

• Flexibility of business model: In the field of traditional AI model training, we have noticed that a common bottleneck is the lack of flexibility in the business model. The platform requires users to pay to train the model, and the source of funds limits the sustainable development space of the project, especially in the early stages that require extensive user participation and data collection. In contrast, Web3 projects will have more flexible business models, such as community-driven models of token economy.

• Privacy protection challenges: Privacy protection is another key issue. Take Predibase as an example. Although it provides virtual private cloud services on AWS, this architecture that relies on third parties always has the potential risk of data leakage.

These points of differentiation have all become bottlenecks in the traditional AI industry. Due to the characteristics of the Internet, these problems are destined to be difficult to solve in an efficient way. At the same time, this also brings opportunities and challenges to Web3. The project party that takes the lead in solving these problems will most likely become a pioneer in the industry.

Web3 Other Agent Projects

After we discussed AI Agent projects for model training, we will expand our horizons to other types of AI Agent projects in the Web3 industry. Although these projects are not only focused on model training, they are different in terms of financing data, listing performance, token market value, etc.

The following are several representative and influential AI Agent projects in their respective fields:

マイシェル

• Product Introduction: It provides a comprehensive AI Agent platform where users can create, share, and personalize AI agents. These agents can provide companionship and assist work to improve efficiency. The platform covers a variety of AI agent styles, including two-dimensional and traditional styles, and the interactive forms include audio, video, and text. The special thing about MyShell is that it aggregates a variety of existing models including GPT 4 o, GPT 4, and Claude, providing users with a high-level experience of traditional paid AI agents. In addition, the platform introduces a trading system similar to the FT bonding curve, which encourages creators to develop high-value AI models while giving users the opportunity to invest and share the benefits.

• Data analysis: MyShells last round of financing was valued at about $80 million, led by Dragonfly, and other well-known investors such as Binance, Hashkey, and Folius also participated. Although there is no specific user access data, MyShell has nearly 180K Twitter followers on social media, and the number of online Discord users usually does not exceed one-tenth of the total number of fans, but it shows that the project has a loyal user and developer base.

• Technical analysis: MyShell does not develop AI models independently, but as an integrated platform, it brings together cutting-edge models such as Claude, GPT-4, 4 o, and claims to support other closed-source models. This strategy enables MyShell to leverage existing technical resources to provide users with a unified and advanced AI experience.

• Subjective experience: MyShell allows users to freely create and customize AI agents according to their needs, whether as personal companions or professional assistants, and can adapt to various scenarios such as audio and video. Even if users do not use MyShells agents, they can enjoy the integrated Web2 payment model at a lower cost. In addition, the platform combines the economic concept of FT, allowing users to not only use AI services, but also invest in the AI agents they are optimistic about, increasing the wealth effect through the bonding curve mechanism.

デリジウム

• Product Introduction: Delysium provides an intent-centric AI Agent network, allowing Agents to better cooperate to bring users a friendly Web3 experience. At present, Delysium has launched two AI Agents: Lucy and Jerry. Lucy is a networked AI Agent with the vision of providing tool-like assistance, such as querying the top 10 coin holding addresses, etc. However, the Agents function of executing on-chain intentions is not yet open, and it can only execute some basic instructions, such as staking AGI in the ecosystem or exchanging it for USDT. Jerry is similar to GPT in the Delysium ecosystem, and is mainly responsible for answering questions in the ecosystem, such as token allocation.

• Data analysis: In 2022, the first round of fundraising was 4 million US dollars, and in the same year, it announced the completion of a strategic financing of 10 million US dollars. Its token AGI currently has an FDV of about 130 million US dollars. There is no latest user data. According to Delysiums official statistics, as of June 2023, Lucy has accumulated more than 1.4 million independent wallet connections.

眠れないAI

• Product Introduction: This emotional companionship game platform combines Web3 and AI Agent technology, provides virtual companion games HIM and HER, and uses AIGC and LLM to immerse users in the interaction with virtual characters. Users can modify the characters attributes, clothing, etc. during the continuous dialogue process. Its compatible large language model ensures that the character iterates itself in each dialogue and becomes more understanding of the user.

• Data analysis: The project has raised a total of $3.7 million, with investors including Binance Labs, Foresight Ventures and Folius Ventures. The total market value of the token is currently around $400 million. There are 116K Twitter followers, and according to official statistics, the number of registered appointments has reached 190K, and its active users have reached 43K. It can be said that its user stickiness is still quite strong.

• Technical analysis: Although the official has not disclosed which large language model on the market their product is based on, in order to ensure that users will feel that the character knows them more and more during the chat, Sleepless AI trains a separate model for each character when designing LLM training, and combines the vector database and personality parameter system to give the character memory.

• Subjective experience: Sleepless AI uses AI Boyfriend and AI Girlfriend to cut in from the perspective of Free to Play, and does not simply integrate the chat box of the dialogue robot. The project greatly enhances the authenticity of virtual people through high-cost art, continuously iterated language models, high-quality and complete dubbing, and a series of functions such as alarm clocks, sleep aids, menstrual records, and learning companions. Such emotional value cannot be felt by other applications on the market. In addition, Sleepless AI has created a more long-term and balanced content payment mechanism. Users can choose to sell NFTs without falling into the dilemma of P2E or Ponzi. This model takes into account both the players income and the game experience.

Prospect Analysis

In the Web3 industry, AI Agent projects cover multiple areas including public chains, data management, privacy protection, social networks, platform services, and computing power. From the perspective of token market value, the total token market value of AI Agent projects has reached nearly $3.8 billion, while the total market value of the entire AI track is close to $16.2 billion. The market value of AI Agent projects in the AI track accounts for approximately 23%.

Although there are only about a dozen AI Agent projects, which is relatively small compared to the entire AI track, their valuation accounts for nearly a quarter of the market. The market value share of the AI track once again verifies our verification that this sub-track has great growth potential.

After the statistics, we asked a core question: What characteristics do Agent projects have to attract excellent financing and be listed on top exchanges? To answer this question, we explored projects that have achieved results in the Agent industry, such as Fetch.ai, Olas Network, SingularityNET, and Myshell.

It is not difficult to find that these projects have some significant characteristics: they all belong to the platform collection project in the infrastructure category, building a bridge, one end connecting the B-end or C-end users who need Agents, and the other end serving developers and verifiers – these are users responsible for model debugging and model training. Regardless of the application level, they have established a closed loop of a complete ecosystem.

We noticed that whether the products they provide are on-chain or off-chain related, these do not seem to be the most critical factors. This leads us to a preliminary conclusion: in the field of Web3, the logic of Web2 that emphasizes practical applications may not be fully applicable. For the leading AI Agent products in Web3, building a complete ecosystem and providing diverse functions may be more critical than the quality and performance of a single product. In other words, the success of a project depends not only on what it provides, but also on how it integrates resources, promotes collaboration, and creates network effects within the ecosystem. The ability to build this ecosystem may be an important factor for AI Agent projects to stand out in the Web3 track.

The correct way for AI Agent projects to integrate into Web3 is not to focus on the deep development of a single application, but to adopt an inclusive model. This approach involves migrating and integrating the diverse product frameworks and product types of the Web2 era into the Web3 environment to build a self-circulating ecosystem. This can also be seen from OpenAIs strategic shift, which chose to launch an application platform this year instead of just updating the model.

In summary, we believe that AI Agent projects should focus on the following aspects:

• Ecosystem construction: Go beyond a single application and build an ecosystem that includes multiple services and functions to promote interaction and value-added between different components.

• トークン economic model: Design a reasonable token economic model to encourage users to participate in network construction and contribute data and computing power.

• Cross-domain integration: Explore the application potential of AI Agent in different fields and create new usage scenarios and value through cross-domain integration.

• After summarizing these three aspects, we also provide some forward-looking suggestions for project parties with different focuses. The first is for those non-AI core application products, and the second is for native projects focusing on the AI Agent track.

For non-AI-core application-side products:

Maintaining long-termism, focusing on its core products while integrating AI technology, adapting to the times and waiting for the wind. Under the current technology and market trends, we believe that using AI as a traffic medium to attract users and enhance product competitiveness has become an important means of competitiveness. Although the reality is that how much AI technology actually contributes to the long-term development of the project is still a question mark, we believe that this provides a valuable window for pioneers who dare to adopt AI technology early. Of course, the premise is that they already have a very strong product.

In the long run, if AI technology achieves new breakthroughs in the future, those project parties that have integrated AI will be able to iterate their products more quickly, thereby seizing opportunities and becoming industry leaders. This is just like in the past few years, live streaming on social media platforms has gradually replaced offline sales as a new sales method for traffic exports. At that time, those merchants who had excellent products and chose to follow the new trend and try live streaming, when live streaming e-commerce really broke out, these merchants immediately stood out with the advantage of early intervention.

We believe that in the midst of market uncertainty, for non-AI-core application-side products, it may be a strategic decision to consider introducing AI Agents at the right time. This will not only increase the market exposure of the product at the moment, but also bring new growth points to the product with the continuous development of AI technology.

For native projects focused on AI Agents:

Balancing technological innovation and market demand is the key to success. In native AI Agent projects, project owners need to focus on market trends rather than just technology research and development. At present, some Agent projects in the market that combine Web3 may be too focused on developing in a single technical direction, or have built a grand vision but failed to keep up with product development. Both extremes are not conducive to the long-term development of the project.

Therefore, we suggest that project owners should pay attention to market dynamics while ensuring product quality, and realize that the AI application logic in the traditional Internet industry is not applicable to Web3. Instead, they need to learn from projects that have achieved results in the Web3 market. Pay attention to the labels they have, such as the core functions such as model training and platform collection mentioned in the article, as well as the narratives they create, such as AI modularization and multi-agent collaboration. Exploring sexy narratives may be the key to a projects breakthrough in the market.

Conclusion Whether it is a non-AI core product or a native AI Agent project, the most important thing is to find the right timing and technical path to ensure competitiveness and innovation in the ever-changing market. Project owners should observe market trends, learn from successful cases, and innovate to achieve sustainable development in the market while maintaining product quality.

要約する

At the end of the article, we analyze the Web3 AI Agent track from multiple perspectives:

• Capital investment and market attention: Although AI Agent projects in the Web3 industry do not have an advantage in terms of the number of listings, they account for nearly 50% of the market valuation, showing that the capital market highly recognizes this track. With more capital investment and increased market attention, it is certain that more highly valued projects will appear in the AI Agent track.

• Competition and innovation capabilities: The competition landscape of the AI Agent track in the Web3 industry has not yet been fully formed. At the application level, there are no phenomenal and leading products like ChatGPT, which gives new project parties a lot of room for growth and innovation. With the maturity of technology and the innovation of previous projects, the track is expected to develop more competitive products and promote the valuation of the entire track.

• Emphasis on token economy and user incentives: The significance of Web3 is to reshape production relations, making the originally centralized process of deploying and training AI models more decentralized. Through reasonable token economic design and user incentive schemes, idle computing power or personal data sets can be redistributed, and data privacy can be protected through solutions such as ZKML. This can further reduce computing power and data costs and allow more individual users to participate in the construction of the AI industry.

In summary, we are optimistic about the AI Agent track. We have reason to believe that there will be multiple projects with valuations exceeding 1 billion US dollars in the AI Agent track. By horizontal comparison, the narrative of AI Agent is sexy enough and the market space is large enough. At present, market valuations are generally low. Considering the rapid development of AI technology, the growth of market demand, capital investment, and the innovation potential of enterprises in the track, in the future, as the technology matures and market recognition increases, this track is expected to see the emergence of multiple projects with valuations exceeding 1 billion.

This article is sourced from the internet: ArkStream Capital track research report: Can AI Agent become the life-saving straw of Web3+AI?

関連: Gitcoin Grant 選ばれた 21 回分の価値ある寄付プロジェクト

原文｜Odaily Planet Daily ( @OdailyChina ) 作者：CryptoLeo Gitcoinの寄付が再び始まりました。今回のラウンド（Gitcoin Grants Round 21）は、Gitcoinにとって21回目の寄付ラウンドです。 Gitcoin Grants Round 21（GG 21）は、エコシステムとコミュニティ主導に焦点を当てたラウンドです。テーマは、コラボレーション、新しいガバナンスプロセス、外部パートナーシップです。主にArbitrumチェーンに焦点を当てた9つのコミュニティラウンドに分かれており、最も早い寄付期限は8月21日19時（北京時間）です。 Gitcoin Grantの導入と運用についてはあまり詳しく説明しません。以前のGitcoin Grant寄付関連コンテンツ（Gitcoin Grants 20ラウンドミニマリストガイドなど）で見つけることができます。この記事では、Odaily Planet Dailyがエコロジーの観点からGG 21で寄付する価値のある29のプロジェクトを選択します…