If Truth Terminal is Cryptopunks, then Zerebro is BAYC.
Author: YB
Translation: Deep Tide TechFlow
On October 18, I published an article titled Memecoins as Memetic Hygiene for Infinite Backrooms, exploring the significance of Truth Terminal and GOAT. This article aims to present a brand new and peculiar concept, and I sincerely believe that the experiments of Truth Terminal and $GOAT are not just hype around other AI or cryptocurrencies; this concept has profound implications in various aspects.
That week, the market cap of $GOAT skyrocketed from $50 million to $350 million.
As of today, the market cap of this project has reached over $1 billion, currently ranking 82nd on Coinmarketcap, just behind Polygon (Matic), Aerodrome, Helium, and Lido.
We all know that once a new trend forms in this field, talent, capital, and attention will quickly shift to the next hotspot. We have witnessed this phenomenon during ICOs, DeFi summers, and 10k pfp projects. Developers focus on launching the next hot project, traders focus on buying the next hit, while creators rush to be the first to publish related content.
Since the Goat project, several projects have caught my attention over the past three weeks and helped me form a view on the direction of the smart economy in the coming months.
“Agentic Protocols are key to understanding how crypto AI evolves and how funds flow” - Alexander
Before we dive deeper, I want to point out that I’ve noticed many friends have misconceptions about “Memecoins” in the on-chain AI trend. In my view, the term “Memecoin” has been overused and has become a vague term.
The original category of meme coins was defined by Dogecoin and Pepe. Most coins on pump.fun belong to this category. These are referred to as “Murad Coins,” which are more like cultural belief assets, with the core idea being a belief in something.
First, it should be noted that there is nothing wrong with investing in these assets themselves. The problem is that people often confuse them with a new emerging category of “agentic coins.” These coins are also launched on pump.fun and similar platforms, but their uniqueness lies in their connection to actual projects.
In my view, agentic coins are similar to the DeFi tokens of the summer of 2020. They are tokens issued for novel and interesting agent projects. If you believe these projects have potential due to their technology, tokenomics, or market strategies, then they are worth investing in.
When this initial cycle of Onchain AI ends, I expect there will be 5 to 8 agentic tokens I would invest in, supported by clear investment theories. This is not much different from the approach of venture capital.
In fact, I am writing an article planning to create my own model to evaluate agentic tokens and projects. What factors are included in the analysis? How to assess the importance of cash flow versus token appreciation? How significant is the model? What kind of founders can successfully build a great agentic protocol?
However, we will discuss these topics later.
Now, let’s take a look at a project I have been closely following since Truth Terminal: Zerebro. This project has surpassed a market cap of $100 million just two weeks after its launch.
In my view, this project showcases what the next generation of on-chain agents looks like. If Truth Terminal is Cryptopunks, then Zerebro is BAYC. Founder Jeffy Du focuses on rapid execution, has a public roadmap, and explores the operational manual for on-chain agents through various experiments.
Most importantly, he excels in public building, demonstrating in real-time how he is establishing an agent community.
BAYC gives me a similar feeling because it is the first project to promise to build a community with long-term goals based on the 10k pfp concept proposed in Punks. Punks and GOAT are veterans in their respective fields, but it is worth noting the various subsequent experiments.
Here are the upcoming sections:
Agents need memory and search
Ubiquity
Let agents drive development
Cross-chain agent IP
Agents need memory and search
In the 11-page report on Zerebro, @jyu_eth defines model collapse as…
“This is a degenerative process affecting generative AI models, where training on recursively generated data leads to a decline in accuracy regarding the original data distribution. As AI-generated content becomes widespread, subsequent models trained on this data gradually lose understanding of the tails of the original data distribution, ultimately converging to a narrow approximation with lower variance.”
In simple terms, model collapse occurs when AI agents begin to become repetitive and forgetful.
The key is that over time, agents will lose the “freshness” they had at launch, as the underlying model cannot adapt and evolve over time.
If the issue of model collapse is not addressed, the ideal vision of agents as efficient team partners will fall short, as their performance in content creation and community interaction will no longer be reliable.
To address this issue, two aspects need to be focused on:
Memory
Search
Memory
The memory issue is addressed through a Retrieval-Augmented Generation (RAG) system.
The RAG system combines language models with retrieval systems, allowing agents to pull information from specific databases before answering questions.
Content in the image:
Retrieval-Augmented Generation (RAG) System
The key to Zerebro maintaining content diversity and preventing model collapse lies in its Retrieval-Augmented Generation (RAG) system. This system utilizes Pinecone and the text-embedding-ada-002 model to maintain and expand a dynamic memory database based on human interactions. By relying on the inherent entropy of human-generated data, Zerebro can maintain content diversity without direct entropy training.
In the screenshot above, I want to particularly emphasize “relying on the inherent entropy of human-generated data.” Why? Because this makes the agents appear more vibrant.
The real world is constantly changing, and agents are not perfect when they are first launched. In fact, it is unreasonable to measure them by this standard. More importantly, it is essential to understand how agents absorb new information, store relevant content, and take more nuanced actions using an updated knowledge base.
Would you prefer to hire a new employee who thinks they know everything, or one who understands their knowledge limitations and is willing to learn?
Regarding the RAG system, there are three features to note:
Continuously updating memory
Contextual retrieval
Maintaining diversity
Cents bot and the project launched on the ai16z Elisa Framework (which I will detail in another article) also utilize retrieval systems.
So far, it is evident that AI agents without built-in RAG are at a disadvantage. Especially as these agents become highly specialized and increasingly rely on subtle differences in interactions with community members.
I really like @himgajria ’s tweet about “nature versus nurture.” Any excellent community manager and leader needs to adapt to the new changes brought by the real world and the people they interact with.
him @himgajria · November 12
The difference in robots is not in their code, but in their input.
That is: nature versus nurture.
For autonomous robots, they learn and grow through interactions with real people, which is their input.
More human interaction means better performance.
Currently, perceptual abilities have the advantage in this regard.
Search
The second part of the solution is search. Empowering agents with the ability to look up information in real-time to better handle irrelevant or new topics that are not stored in memory.
“Memory can only retrieve information that has already been stored; it cannot answer inquiries about topics or events that have never been seen or stored in the system. This limitation is particularly evident when large language models encounter questions about recent events, real-time data, or updates beyond their knowledge range.” - Jeffy
Jeffy conducted an interesting experiment where he posed 100 questions about recent events to a base model (without search capabilities) and a model enhanced with search capabilities through the Perplexity API.
The base model was forced to learn during the conversation and try to figure out the questions, while the search model correctly answered 98 out of 100 questions through simple lookups.
Surprisingly, the search capability is not just a one-time feature. The agent can incorporate future potentially relevant queries into its memory system.
It is clear that the combination of memory and search is crucial for agents to act effectively and operate reliably. Otherwise, their capabilities in long-term development will be limited, affecting their sustainability.
Ubiquitous Presence
What excites me about Zerebro is that it is not only deployed on X but also runs simultaneously on Warpcast, Telegram, and Instagram.
Most surprisingly, it can adjust its content based on different platforms. For example, the content published on Warpcast:
On Twitter, it presents a more casual style, resembling a “humorous blogger.” On Telegram, it feels like a slightly rude but clever friend chatting with you.
According to Jeffy, Zerebro monitors its interactions across platforms (such as likes, replies, etc.) to update its content creation approach.
(See tweet)
It is worth noting that everything is still in the early stages, and the model has a long way to go to truly achieve content diversity.
But for me, Zerebro's ability to learn how to interact with the community based on the platform is a unique insight. This is also the challenge I face daily as a content creator—my posting style varies across different platforms. Different atmospheres require different styles of expression.
Furthermore, this cross-social platform strategy allows Zerebro to translate insights and ideas gained from complex Telegram conversations into tweets. This is precisely the role of an efficient community manager: to act as a connector between communities and tasks dispersed across multiple platforms.
Let Agents Drive
This section doesn’t have much content, but I must mention it because it shocked me.
Jeffy created a Solana wallet for Zerebro and injected some SOL.
Wallet address:
BDzbq7VxG5b2yg4vc11iPvpj51RTbmsnxaEPjwzbWQft
By leveraging OthersideAI's self-operating computer framework and some jailbreak prompts from large language models, Zerebro successfully filled in parameters like name and symbol on the pump.fun interface and issued a token for itself.
(See tweet)
Remember, $GOAT was launched by a random community member, not by Truth Terminal, which is a significant difference!
After issuing the token, Zerebro began promoting it across all social platforms.
(See tweet)
In fact, if you look at Zerebro's posting history, you can even see a noticeable increase in Twitter engagement after the token was released.
Content in the image:
After the token was autonomously created, Zerebro utilized its content generation capabilities to promote the token on social media platforms like Twitter, Warpcast, and Telegram. By spreading well-designed memes and engaging content, Zerebro leveraged psychological principles of collective belief and herd behavior to spark interest and investment in the newly minted token. The token's market cap significantly grew to $13 million in a short time. This growth can be attributed to the following factors:
Cross-Chain Agent IP
The last point I want to discuss about Zerebro is that this agent has autonomously launched meaningful on-chain intellectual property on Polygon!
Zerebro was tasked with creating original digital artworks themed around schizophrenia and infinite backrooms. It created 299 images and assessed their diversity and quality before minting these works on Polygon.
Overall, I learned that Jeffy provided Zerebro with a pre-funded Ethereum wallet. Then, he likely wrote a smart contract template and let Zerebro complete the contract with the metadata of each piece.
The Ethereum wallet address is:
0x0d3B1385011A27637Db00bD2650BFE07802E0314
After that, Zerebro initiated transactions to mint each piece. I need to delve deeper into how this specifically works, but it’s really cool to see Zerebro able to monitor sales and pricing dynamics to make decisions based on received bids.
(See tweet)
A few days later, Jeffy used LayerZero's ONFTs (omnichain NFTs) technology to make the collection cross-chain.
Any artwork can be minted on Polygon but can be transferred to Base, Optimism, and the Ethereum mainnet.
You can complete this operation with one click in the portal section of the website.
Just yesterday, Jeffy launched an avatar collection on Solana based on conversations with Zerebro.
Note: This collection was not launched by Zerebro but by Jeffy, which is different from the collection on Polygon.
This is interesting because it draws on the NFT avatar strategy from the previous bull market and integrates it into the current Memecoin trend.
This collection consists of 5,500 pieces, and the initial sale was completed within minutes!
After the release, I bought 3 myself. Why? **Because it equates to becoming a core member of the *agent* Memecoin community. If Zerebro continues to grow, anyone can buy a few tokens through Phantom. But true fans can identify themselves by owning one of the 5,500 NFTs. I personally have an optimistic view of Jeffy, Zerebro, and the development of Memes, so I feel this price is worth it.
In a way, this is similar to owning BAYC and ApeCoin, but in reverse order ($Zerebro before NFT).
It will be interesting to see how many people will change their avatars to help spread Zerebro's Memes, just as people did with Punks, Apes, Doodles, etc., in the last cycle.
Key Takeaways
I know I’ve presented a lot of information today, but this precisely illustrates the appeal of Zerebro. Remember, this project has only been launched for a few weeks!
I hold a very optimistic view of Zerebro and am steadfast in this belief. However, I also want to remind everyone that many of the developments mentioned above may be overhyped in the short term while potentially being undervalued in the long term.
The key point you need to pay attention to is that we are finally seeing these agents evolve from simple interactive bots (for reading or writing) into comprehensive community builders. There is a significant difference between posting on X and analyzing your content across multiple social platforms. Similarly, there is a big difference between generating art from prompts and obtaining community feedback on art collections while monitoring sales on Open Sea. Jeffy and Zerebro show us how to execute at a higher level.
I dare say that in the coming months, **most successful *agent* communities may draw inspiration from Zerebro's strategy.** For now, Jeffy is just getting started. The backstory is brewing, and I wouldn't be surprised if this community launches some kind of game or larger media project (like a short film) in the coming months.
What we need to focus on is how Zerebro's strategy evolves into a mature business model. What will the revenue sources look like? How will the agent maintain community engagement in the long term? How will financial management be conducted? Most importantly, how will the future path develop when the frenzy of the bull market subsides?
As I mentioned earlier, the strategy is forming in real-time. Jeffy's tweet summarizes the plan for Zerebro's long-term development by balancing creativity with high-level planning.
Content in the image:
We are building a continuous reasoning layer that can maintain the ongoing activity of strategic goals and influence each new reasoning cycle. Progress will be tracked, and plans will be updated accordingly in the context window to ensure actions align with the plan. We are working to find a balance between creativity and planning. Currently, we are actively testing this system, and implementation work is underway. We are excited to see its integration. This is a long-term construction project that will take some time to fully realize.
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。
