Fragmentation of AI agents: get rid of the noise and focus on these key projects

Author: 0xJeff, Steak Studio; Translation: Golden Finance xiaozou

Last year, when there were more L1/L2 public chains than we could count on both hands, liquidity fragmentation became a major challenge. Subsequently, Bitcoin L2 entered the market, causing the number of L2 to surge.

This results in fragmented mobility, attention, and experience across countless ecosystems. Many L2s and dApps in these ecosystems struggle to gain any practical meaningful traction. Capital is only flowing into a handful of top ecosystems—those with outstanding communities, great products, high- quality teams, and enough capital to attract builders.

Now, the field of AI agents is beginning to face the same challenges. While it’s still early days, the sheer number of individual agents, tokens, and new agent tokens makes it difficult to differentiate between them.

The industry is facing a growing problem of agent fragmentation . Mobility and focus are too spread out and thin.

1. Two common tracks of AI agents

Most AI agents proceed along two tracks:

(1) Convert from agent to framework

Some agents demonstrate their value by constantly engaging users and allowing them to interact with the product in real time on platforms such as Crypto Twitter (or other channels). Once these agencies achieve product market fit (PMF) and gain enough traction, other teams will often want to copy what they have built.

What's the logical next step? Open source proxy. In this way, the agent evolves into a framework—making it easier for others to experiment, integrate, and build on top of it. This move could break the valuation ceiling that limits agents and position them to be more similar to L1 blockchains.

We see this trend with the following agencies:

Zerebro

Dolos

AVA

Freysa

Rei

This approach allows these agents to scale by reducing barriers to others joining their ecosystem.

(2) Continue to improve products as an independent agent

Not all agents discover PMF immediately. Some agents must evolve iteratively until the right use case is discovered. Even so, some teams prefer to keep their own technology rather than making it available to others.

For those agents that discover PMF and perform well, there is no need to convert to a framework - they can still continue to be independent agents.

An example of this approach is as follows:

ikB

Griffin

So, what should we do?

• For builders: focus on understanding the gaps in the market and building products that provide real value to users. Remember, agents should act as a conduit to bring users into the ecosystem. In such a saturated market, standing out is only getting harder and harder.

• For investors: Keep an eye on the leaders or top projects in each category. Doubling down on dips may be the best way to solidify your focus in this fragmented space.

2. Current Categories and Narratives and Noteworthy Projects

(1) Alpha-centered KOL agent

These agents dominate output alpha (insights, data synthesis, and analysis). Aixbt leads in mindshare. While this is not a winner-takes-all market, the front-runners will receive disproportionate attention.

Tokens worthy of attention include: AIXBT, REI, TRISIG, TRUST, AGENCY, KWANT

(2) InvestmentDAO

These DAOs focus on strategic investments. The winners in this space are usually those DAOs that can effectively execute strategies and provide strong PnL.

Examples are as follows:

VADER、AIXCB、AIMONICA、SEKOIA、WAI、AROK

Other platforms in this field include Virtuals, ai16z, daos.world and daos.fun.

(3) Agent virtual world/game x AI

This is a young field with huge potential. Realis started the trend, but Hyperfy is now the world leader in building infrastructure. Smol and Arc Agent are exploring AI-based games and gamification mechanics with reinforcement learning.

The main tokens are as follows:

HYPER, REALIS, SMOL, NRN

(4)DeFAI/abstraction layer

This is a highly competitive field with multiple players vying for dominance. It is expected that Wayfinder’s token offering could be a game changer.

Tokens that need attention are as follows:

GRIFFAIN, ANON, GRIFT, NEUR

(5) Autonomous trading agent

The narrative is still in its early stages but has huge potential. Several competitors have emerged in the space, with agents working hard to generate huge returns.

Tokens that need attention are as follows:

ASYM、PPCOIN、GEKKO、TONY

(6)Framework

Popular framework tokens enhance agency by providing open source solutions. AI16Z/Eliza is the largest open source framework, but there are other models.

The main tokens are as follows:

AI16Z / ELIZA、GAME (Virtuals)、ARC (Rig)

(7)Music

AI agents are entering the music field by releasing albums or supporting music creation. Zerebro is well ahead of the game with a stellar album, and now more and more agents are tapping into this niche with cross-disciplinary collaborations with world-class DJs.

The main tokens are:

ZEREBRO, MUSIC, BEATS

(8)3D model

Ava is undoubtedly the leader in the field of 3D modeling. Holoworld AI has been working with teams to develop and deploy 3D models across different frameworks and ecosystems.

AVA is a coin to watch.

(9)AIDOL

This trend started when Luna became an AI idol on Douyin and Twitter. While the initial wave of hype has subsided, the segment has laid the foundation for agents offering more practical services. Luna is still the most relatable AIDOL, she now has more personality and can do more.

The main tokens are as follows:

LUNA,ELIZA,MOE

(10)AI Meme Coin

These meme coins are either issued by AI agents or are related to AI.

Mainstream tokens include:

GOAT、FARTCOIN、ACT

(11) Video/Cinema Photography

It's a nascent industry, and Kween is leading the way when it comes to producing high-quality films, although it's unclear whether they are agency- driven. Sandy is launching the first artificial intelligence video agent framework, whose potential has yet to be fully realized.

Tokens worth watching include:

KWEEN, SANDY

(12)Data

In the field of artificial intelligence, data is the digital golden key to agent development and performance. Cookies are leading the way in providing data to humans and AI agents.

The main tokens are:

COOKIE,NOMAI

(13) AI application store

This category focuses on an ecosystem of useful AI applications, including GenAI tools, games, and productivity-enhancing tools. Alchemist is currently the top ecosystem, while Myshell focuses on image generation and waifu simulator games, although its token has not been released yet.

The main token in this space is ALCH.

(14) NSFW proxy

This is an emerging segment where agencies play the role of spicy influencers, engaging users around the clock. Lush, Nectar and Oh are the biggest competitors, but currently there is only one token, Lush.

Tokens to watch: LUSH

(15) TEE/DeAI infrastructure

This narrative enables fully autonomous agents to be resistant to human interference, such as hacking or rug pulling.

The main tokens are:

FAI, PHALA, SPORE

(16) Development-focused tools

Key tools for AI safety, code analysis, and utility that are critical to the industry’s growth. Soleng is my personal favorite because it helps my non- technical friends on Crypto Twitter analyze the code/github directory.

Tokens to watch out for include:

SOLENG、H4CK、CERTAI、JAIL

(17) Robot/embedded AI

This is an exciting area that is still developing. OpenAI's work in embedded AI may attract more attention this year.

Main Token: SAM

(18) Other items

There are many more categories that I haven't covered. The agency fragmentation trend is out of control. My personal view on investing is to focus on the narratives that attract the most attention and anticipate that future narratives will also attract attention.

The main narratives I'm concerned about that will continue to gain traction are:

• DeFAI

• autonomous trading agent

• InvestmentDAO

• agent virtual world

There are also a few narratives that I am bullish on because I expect them to attract more attention as we move into the first/second quarter of the year, they are:

• NSFW proxy

• TEE and other DeAI infrastructure (opML, zkML, POSP, etc.)

• robot

• Development-focused tools

3. Conclusion

The issue of agent fragmentation highlights the complexities and opportunities in this rapidly evolving field. Rather than clinging to all trends, builders and investors should focus on areas with potential for consolidation and long- term growth.

For builders, the focus should be on creating scalable, differentiated products that meet clear user needs. For investors, backing projects with strong focus and positioning on high-potential narratives will be key to combating the challenges of fragmentation.

The next wave of innovation will come from projects that simplify chaos and create ecosystems that combine mobility, attention, and value.