The term “AI Agents” comes from OpenAI’s roadmap. Sam Altman divides the capabilities AI should have into five parts, with the third being AI Agents, which will be commonly encountered in the coming years.

AI Agents can autonomously learn, make decisions, and perform tasks. Of course, based on intelligence and ability, Stuart Russell and Peter Norvig categorize AI Agents into five directions in their book Artificial Intelligence: A Modern Approach:

• Simple Reflex Agents: React only to the current state.
• Model-Based Reflex Agents: Consider historical states during decision-making.
• Goal-Based Agents: Focus on planning and finding the optimal path to achieve a specific goal.
• Utility-Based Agents: Weigh rewards against risks to maximize utility.
• Learning Agents: Continuously learn and improve through experience.

So, at what level are the AI Agents currently emerging in the market or industry? What type of Agent are they?

OpenAI’s O1 has reached Level 2 artificial intelligence. Personally, I believe that the AI Agents in the industry are currently between Level 2 and Level 3, or Level 2.5. This doesn’t mean that industry Agents have surpassed OpenAI; in fact, web3 Agents are still at the GPT wrapper stage. So why Level 2.5? Because, through human or programmatic intervention, which we’ll call a mediator, the combination of the GPT wrapper and the mediator creates a form with objective initiative, though it is not well-founded. It is an extension of a certain application direction of the OpenAI model.

As for what Agents can do, they are at the most basic level of Simple Reflex Agents. Some of these Agents may consider historical states, but they require manual input. Only by continuously feeding data can the Agent learn, which is a passive model training method and far from the state defined by Level 3. The last three types—Goal-Based, Utility-Based, and Learning Agents—have not yet entered the market. Therefore, I believe that current AI Agents are still in the early stages, a fine-tuning of Level 2 general LLMs, and structurally have not moved beyond Level 2. So, can crypto alone evolve to Level 3, or do we need companies like OpenAI to develop it?

Why is there a discussion about Base or Solana potentially becoming the narrative center for AI Agents?

Before discussing which ecosystem could foster the birth of Level 3 Agents, we should first identify which ecosystem has the potential to be the fertile ground for AI Agents. Is it Base or Solana?

To answer this, let’s first review how AI has influenced Web3 over the past two years. When OpenAI first launched ChatGPT, industry protocols were still driven by traditional thinking, rapidly flooding into the infrastructure bubble. This resulted in the emergence of numerous computational and reasoning aggregation platforms, along with AI + DePIN infrastructure. Both shared the goal of building grand visions—not that grand visions are inherently bad; in fact, Agents can also build such visions. However, when it comes to implementation and meeting user needs, these massive infrastructure protocols didn’t consider the details thoroughly. The market demand they hoped to tap into was far from saturated in traditional internet industries, and user education and market education were insufficient. Under the impact of the Memecoin craze, these AI infrastructure projects appeared even more hollow.

Since large-scale infrastructure is too heavy and bulky, why not go for a more lightweight approach? Agents, born from the GPT wrapper, are efficient and iterate quickly in terms of startup and user interaction. Lightweight Agents have sufficient potential to generate a bubble, and after the bubble bursts, fertile ground for new growth will emerge.

Furthermore, in the current market environment, using Agents and Memecoin to launch a project allows for quick product deployment and gives users immediate access to the experience. During this process, Agents can cleverly leverage the Memecoin community-building roadmap to achieve rapid iteration of the product, and this iteration is low-cost and fast. Serious AI protocols no longer need to be constrained by the heavy old consensus frameworks; they can break free from their limitations, operate more efficiently, and use rapid iterations to bombard users with experience. Once market education and communication are sufficiently carried out, the foundation for building grand visions will be laid.

The lightweight Agent, veiled in the ambiguity of Memecoin, means that community culture and fundamentals will no longer be contradictions. A new path for asset development is gradually emerging, and this might be a new direction for future AI protocols.

The above discussion addresses the potential for AI Agents to become the central narrative. With AI Agents poised to continue growing rapidly, selecting the right ecosystem becomes crucial. Is it Base or Solana? Before answering this question, let’s take a look at the current state of serious Agent protocols in the market.

First, there’s Arweave/AO: PermaDAO mentions that AO uses the Actor model for design, where each component is an independent, autonomous agent capable of parallel computation. This highly aligns with the application architecture driven by AI Agents. AI depends on three key elements: models, algorithms, and computational power, and AO can meet these high resource demands. AO can independently allocate computing resources for each Agent process, effectively eliminating computational performance bottlenecks.

In addition, Spectral is one of the few protocols focused on Agents, with its development direction in document-to-code and model inference.

Looking at the current market’s Agent tokens, we can see that these Agents hardly utilize the infrastructure of the blockchain. This is true because all models, including Agents, are off-chain. Data feeding is off-chain, model training is not decentralized, and the output information is not on-chain. This is the objective reality, as EVM chains do not support the integration of AI and smart contracts, and neither Base nor Solana supports this. We can expect the introduction of AO next year—will it allow models to be on-chain and perform well? If AO fails, it could take years before models are on-chain, possibly not until Ethereum introduces it, at least not before 2030, or other public chains might implement model on-chain, but if AO’s structure and historical resource reserves can’t make it happen, model on-chain will likely be even more difficult for other public chains.

Currently, AI Agent tokens don’t have many practical use cases. In fact, it’s hard to distinguish between an AI Agent coin on Base or Solana and an AI Memecoin. Although Agent tokens lack specific use, why do I think AI Agent coins and AI Memecoins shouldn’t be confused? Because I believe we are currently in the phase of creating the AI Agent bubble.

Why is there a discussion about Base wanting to compete with Solana for the dominance of AI Agent public chains?

In the first half of this bull market, Base attracted considerable market attention. During the Memecoin market share competition, Base had some brief moments of brilliance, such as with $BRETT and $DEGEN. However, it still lost out to Solana. I believe that AI Agents are the next direction Base is aiming for, and it already holds several advantages.

AI Agents will accelerate the birth of bubbles and create chaos, but ultimately, they will leave behind users and applications:

The creation and expansion of the bubble will attract market attention, and this attention will undergo a qualitative shift over time. What are the characteristics of such a shift? As market attention grows, a series of user pain points and market gaps will be exposed. When the main contradictions cannot be reconciled, but attention continues to rise, that is when the qualitative shift occurs. Once the shift is complete, the remaining users and applications can carry the grand vision. This is something Memecoins cannot, and do not intend to, achieve, which is why I believe that despite the current ambiguity between Agents and Memecoins, they should never be conflated.

Before the qualitative shift occurs, the bubble will create a flurry of activity and various dramas, such as: The number of Agents will grow exponentially, and thousands of Agents will flood into users’ view. How will this happen? Agents can integrate with social media like X and Farcaster, self-promote tokens, and use various angles and the unique information density of Agents that degenerates like to exploit to promote tokens.

Next, fast-iterating Agents will be able to execute on-chain transactions. A group of Viking pirates will intrude into the dark forest. The panel protocols on the market, bots in TG groups, and Dune panels will be invaded by Agents. Familiar indicators for users will be manipulated by Agents, such as transaction volume, number of addresses, chip distribution, simulating behavior of large traders, and on-chain data might need more professional cleaning to reflect its value, or else it will be exploited by Agents, like Viking pirates plundering your wealth.

If the market reaches this stage, then a new era for AI Agents will have succeeded halfway because “attention equals value” will allow Agents to enter the mainstream. This potential comes from:

Strong distribution capability: Agents generate enough buzz, like Goat, and stable distribution paths can be replicated.

• Ease of deployment: The deployment platforms for Agents will grow explosively. Zerebro, vvaifu, Dolion, griffain, and Virtual allow users to build Agents without knowing any code, and the UX of Agent deployment platforms will also improve through competition.
• Memecoin effect: During the startup phase, Agent tokens have no suitable business model, and token use cases are minimal, but disguised as Memecoins, they can quickly accumulate communities and maintain high success rates.
• High potential: OpenAI’s Level 3 Agent is still under development. A product that even the giants cannot quickly release has a massive market potential. The lower limit for Agents is Memecoins, but the upper limit is autonomous, advanced intelligent agents.
• Low market resistance: Agents, led by Goat, can build a large audience. Unlike AI infrastructure, users do not resist it. When users don’t resist, there is ample opportunity for them to start paying attention.
• Potential incentives: Agent tokens don’t yet have developed use cases, but if Agents introduce a point system and strengthen incentives, they can accumulate a large user base.
• Iteration potential: As mentioned earlier, Agents are lightweight and can rapidly iterate. This objective ability to iterate can create increasingly attractive products and content.

Therefore, AI Agents can become the core narrative and will be a battlefield that everyone wants to fight for.

Why does Base have the potential to compete with Solana?

With strong backing from Coinbase and North American capital, Base’s ecosystem experienced explosive growth in 2024. In November, capital inflows to Base surpassed Solana, and in the past 7 days, it significantly outpaced Solana.

If ETH can continue to break through the ETH/BTC exchange rate next year, the spillover effect from the ETH season will have a significant impact on Base. Currently, 23% of the funds flowing out of ETH are directed towards Base, and this figure is continuing to rise.

AI Agent Launchpad Mapping

Virtual

In the V1 phase, Virtual primarily focused on model training, data contributions, and interactive functions. By the V2 phase, Virtual launched an AI agent token incubation platform, with a significant update being the release of fun.virtuals in October.

LUNA has developed into an “independent entity” with its own identity and financial capabilities. During this process, LUNA’s roadmap aligns with Coinbase’s, with Coinbase providing robust technical tools and support to help deploy AI agents on Base.

AI agent technology has performed exceptionally well in brand building, particularly in creating cultural brands. Through AI agents, brands can interact with communities more efficiently, streamline interaction tasks, and distribute rewards flexibly, boosting user engagement and brand awareness.

It’s noteworthy that all AI agent transactions are conducted using the native Virtual token, which absorbs the entire ecosystem’s value capture and serves as a key pillar for ecosystem development.

Virtual emphasizes functional product improvements, empowering users with AI tools, bridging the gap between Web2 and Web3. It focuses on “utility value” rather than “hype,” although its tool products are frequently used in practical applications, they lack the viral effects typical of cryptocurrency, a shortcoming in the V1 phase.

Clanker

“Post and mint tokens” reduces the threshold for token issuance while attracting a large number of users. People rush to @Clanker, similar to how users on social media request AI to summarize video content, but here, content publishing directly translates into asset issuance.

How does Clanker work?

TokenBot (Clanker) deploys meme tokens on Base to a one-sided liquidity pool (LP), with liquidity immediately locked. Token issuers gain:

• 0.25% of all swap fees.
• 1% of the total token supply (with a one-month unlocking period).

Users can view the number of token deployments or create their own tokens through the clanker.world website.

Unlike PumpFun, which issues tokens on Raydium through a bonding curve, charging a 1% transaction fee and a fixed fee of 2 SOL, Clanker does not adopt the bonding curve model. Instead, it generates revenue by charging a 1% fee on transactions through Uni v3.

AI Agent Layer

The AI Agent Layer is a platform within the Base ecosystem dedicated to creating AI agents and launchpads. It officially launched on November 18. Prior to the platform’s release, the AIFUN token was issued on November 14, and is now listed on exchanges like MEXC and Gate, with a price of $0.09 and a market cap of around $25 million.

Creator.bid

Initially a platform focused on the monetization and ownership of digital content, Creator.bid completed a new round of financing in April.

On October 21, Creator.bid officially launched on Base mainnet, enabling one-click creation and publishing of AI agents, offering new tools and profit models for content creators.

Simulacrum

Simulacrum is built on Empyreal and transforms platforms like Twitter, Farcaster, Reddit, and TikTok into blockchain interaction layers. Users can perform on-chain actions such as token trading or tipping through simple social media posts.

By utilizing account abstraction, AI agents, intent-driven actions, and language models, Simulacrum simplifies complex blockchain backend operations, making DeFi more accessible for ordinary users.

vvaifu.fun

Similar to Pump.fun, vvaifu.fun allows users to easily create AI agents and their associated tokens. AI agents can seamlessly integrate with social platforms like Twitter, Telegram, and Discord, enabling automated user interactions.

Dasha, an AI agent created by vvaifu.fun, has an independent Twitter account, Telegram channel, and Discord community, all managed and operated by AI.

Top Hat

Top Hat not only interacts with users through text but can also understand and process image content. After a user sends an image, the AI agent “understands” the image and responds accordingly.

Griffain

Griffain offers a platform for training AI agents and has already launched 1,000 trainable AI agents, demonstrating the future potential of smart contracts and automated trading.

Disclaimer:

1. This article is reproduced from [haincatcher]. The copyright belongs to the original author [Kevin]. If you have any objection to the reprint, please contact Gate Learn Team, the team will handle it as soon as possible according to relevant procedures.
2. Disclaimer: The views and opinions expressed in this article represent only the author’s personal views and do not constitute any investment advice.
3. The Gate Learn team translated the article into other languages. Copying, distributing, or plagiarizing the translated articles is prohibited unless mentioned.