Personal thoughts and judgments on the AI Agent track.
Written by: cryptoHowe.eth
Recently, ArkStream Capital published a research report on the AI Agent track. After reading it, I found it quite reasonable, and I agree with many of the viewpoints. I will also extend some points mentioned in the article to share my thoughts and perspectives, and I welcome everyone to engage in discussion.
Statement: This article carries a strong personal subjective consciousness, and the viewpoints mentioned do not constitute any investment advice, but are for communication and sharing purposes. This article is based on my existing knowledge and current data for inference, and will be updated at any time in the future.
Why the AI Agent track can occupy such a large market share
From the report, we can see that AI Agents occupy nearly a quarter of the entire AI track's market share. I believe there are two main reasons for this large share:
1. The applicability of Agents is broad, the threshold is relatively low, and the product cycle is short.
Currently, the AI track mainly consists of five aspects: data, storage, computing, algorithms, and communication. In terms of data, sufficient resource accumulation is required and is easily affected by geopolitical factors. Storage and computing are currently very resource-intensive fields, while algorithms and communication have high technical thresholds.
The Agent track is in a "just right" position; it does not require massive data, storage, and computing power like general large models, nor does it need significant improvements in algorithms and communication like squeezing toothpaste. It only needs to meet the development needs of the product itself. Therefore, Agent products do not have a high threshold compared to other AI products, have a wide range of use cases, and the overall development cycle is relatively short, allowing projects to gain momentum quickly, commonly referred to as "small but beautiful."
2. Agents are closer to the needs of the general public, better grounded, and fulfill the narrative of Mass Adoption.
The narrative of Agent products is actually very similar to that of the chain abstraction track, both aiming to allow users to focus solely on their own needs without considering the implementation paths and various participants in between. Agents play a significant role in helping Web2 users transition to Web3; users no longer need to learn basic knowledge such as wallets and signatures from scratch, but can directly express their needs through natural language, allowing for automatic execution of related operations. For example, if a user wants to exchange all their BTC for ETH, the Agent will automatically plan the relevant interaction process and execute cross-chain transactions, and the user only needs to wait for the Agent to complete the operation to get the desired result. Therefore, Agents can be classified as one of the directions that can truly achieve Mass Adoption.
The survival dilemma of content generation Agents
The report also categorizes Agent products into infrastructure and content generation types, with most current products belonging to the infrastructure category. Why is the development of content generation types relatively slow? In other words, what is the survival dilemma of content generation Agents? I believe there are two main points:
1. Content generation is more a way to satisfy emotional needs, and emotional needs are difficult to price.
In short, the business model is hard to close the loop. For infrastructure products, it is sufficient to provide relevant services or resources, such as providing computing power or model services for AI developers. In this process, the product's price is easy to measure; for instance, the user uses a certain model of GPU computing power for a certain duration, and the relevant price can be derived through basic arithmetic, with small price fluctuations.
For content generation products, meeting users' emotional needs sustainably is a very challenging task. Users' emotional needs are inherently unstable; they may be very happy today and suddenly feel down tomorrow. Users' willingness to engage with the product also varies, and different users have different emotional needs at different stages, leading to varying willingness and levels of payment, resulting in large price fluctuations.
2. Determining whether the content of the Agent meets user needs is a major challenge.
In content generation products, human subjective consciousness occupies a significant portion of the product. For example, whether the generated image is satisfactory does not have a standard for measurement; it relies more on the user's personal feelings, unlike the computing power market, which has clear standards and market prices. Therefore, user retention and conversion for these products will be more difficult.
Some personal judgments on AI Agents
Regarding the future development of the AI Agent track, I believe there are four points to pay attention to:
1. Pure Agent narratives will struggle to gain a competitive advantage in the market and need differentiated competition. In the current environment, more and more AI projects will incorporate Agents into their narratives, making it difficult for pure Agent projects to stand out. Imagine, among hundreds or thousands of AI projects, simply discussing the Agent narrative is unlikely to attract user attention; after all, good wine fears no alley.
2. AI Agents will gradually transition from being independent to interconnected AgentFi. Currently, products in the Agent track are independent, and their data or services do not interconnect. Users need to provide relevant personal data from scratch when using different Agent products. If there could be a reasonable way to connect different products, allowing users to use a trained Agent from Product A in Product B, the imaginative space and user experience would be significantly improved.
3. Projects that follow the "selling water" logic will emerge first and capture a large share of the market. In simple terms, while everyone is developing Agent products, I will create a tool that can efficiently develop Agents. Such products are likely to be profitable gold shovel projects.
4. The revenue of Agent products mainly comes from B2B, while B2C is more of a strategy for building reputation. This should be a common phenomenon in the AI track; the willingness and ability of C-end users to pay are far less than those of B-end users. Therefore, for a product to truly profit, it largely depends on the quality of B-end partners. However, the promotional capabilities of C-end users should not be underestimated; having enough users to use the product can significantly aid in subsequent promotion and marketing.
Finally, here is a nice summary diagram of the AI Agent framework that I recently came across.
免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。
