AI Agent Classification: What Kind of Agent Has the Most Potential?

Author: IcoBeast.eth

Compiled by Luffy, Foresight News

The AI agent, which is in its heyday, presents a plethora of exciting opportunities, but also comes with a multitude of pitfalls. The rapid emergence of new agents is dazzling, as everyone is eager to launch AI agent startup platforms to attract attention, and more importantly, platform fee revenue.

The harsh reality is that we currently do not know if AI agents have reached their peak or have unlimited potential in the future. This does not mean that the thinking framework and classification scheme are meaningless; it just means that no one can get everything right.

Therefore, what I am going to do is to outline how we should now treat AI agents and what this means for their potential valuation. Note that none of the content in this article is investment advice.

I think there are currently three major types of AI agents in the market, and this number may increase to four, but before that, let’s see how the current situation develops.

Culture / Community Awareness Promoter

Companion / Intellectual Property Agent

Practical Proxy

These categories are completely different. However, in real life, agents can have elements from multiple categories at the same time. In the long run, any legitimate AI agent will have the ability of these three categories. I will now explain these categories.

Culture / Community Consensus

You may already know some AI agents like zerebro, Fartcoin, Truth Terminal, Dolos, etc. These tokens have passionate followers and interact with fans on Twitter every day. These are the AI tokens that your favorite KOLs love the most.

They are junk posts, they are interesting (sometimes), and they are corrupt. They are easy to understand (or at least make you think you understand them), that’s why your KOL fren likes them. They are essentially real AI agent memecoins. Like MOG, PEPE, SPX, these currencies exist and have value because people have a sense of belonging by owning them.

That being said, I don’t think these Tokens can fully unleash the power of AI. It’s almost a waste to have them exist solely as interactive memecoins. Some people have realized this: zerebro’s developers are moving towards practical applications. I believe this is the future for this category. Either face a market valuation ceiling (lower than PEPE) or expand with collective Consensus to killer use cases and increase the valuation ceiling.

I currently have little exposure to these Tokens (although I bought some in the range of 2 million to 10 million initially). I don’t think they have peaked (at least not in the long term), but this kind of pump is not what I’m looking for in the early stages of a Bull Run. I expect few of these tokens to reach the peak of 2 billion in market capitalization, with many reaching valuations in the eight or nine digits and then struggling.

Accompany/IP

You may have seen many Tokens of this kind, but you may not realize the difference between them and the first type. These agents are very niche, feeling more like private frens than part of a large community you belong to, such as the randomly generated agents launched on vvaifu or role-based agents in Virtuals (AIRENE/SISAI/ etc.).

I think this category overall has much lower returns than Category 1, and most of them will not generate any interest. But the winners will win big. And that’s the most interesting part, finding the winners. In this category, I want to support an agent developed by someone who is fully committed.

The ones I chose are Sploot (introduced by Jonah on Virtuals, and I know Jonah may be a polarizing figure in the encryption game, but he is indeed successful and an excellent marketer, which is exactly the advantage I want), Lexi (launched by Coop9000 on Virtuals, Coop9000 Depth is involved in the marketing and business development of AVAX), and then the popular choice Wokie (launched by bitcoin.com on Virtuals).

In the end, the valuation ceiling of these characters may be within the range of 20-30 million US dollars, unless someone creates a breakthrough character that people can’t put down.

Practical

This is the Holy Grail of AI agents. There are many topics to discuss on this subject, and people are developing and experimenting with new practical cases every day.

For the past decade, we have been peddling useless functional Tokens that solve problems nobody cares about. I believe AI agents are the new paradigm for practical Tokens. True technology can provide people with the services they truly desire.

The strange thing is, if meaningless, unused utility and governance tokens can reach valuations of billions of dollars, why can’t AI agents with real use cases do the same?

Here are some of the favorite proxies that I have seen so far:

AIXBT: Definitely the best market sentiment and momentum capture tool at the moment. In my opinion, its value should exceed $1 billion.

VADER: a brain game, effectively trying to establish and manage institutional-grade agent investments. It is the on-chaintokenization BlackRock.

WAI Combinator: The name is interesting and the concept is also interesting (although a bit off topic). They are building an AI incubator to help other AIs grow better. I don’t know if it is effective, but they have already released a report on investing in other promising AIs.

Decentralized Finance maximization: There are many teams building (and soon to be released) AI agents that will automatically manage your yield positions to maximize returns. Once we can abstract away the complex management of LPs, the yield will rise significantly. This concept is like a money printing machine.

PolyAI: It is an agent that can analyze Polymarket information in real time. It may be valid, it may be invalid, but it’s really interesting.

The potential in this field is limitless. Find exciting new practical AI agents and get them early. You will see practical agents worth billions of dollars within 2 years.