With the launch of $ACT and #Binance , in addition to waves of messages about getting rich quickly on Twitter, AI Agent+Crytpo also sparked discussion.

It just so happens that Binance Research’s latest report has expanded on the origins of AI Agents, market cases, and other aspects. I have also simplified and refined some important parts of the original report and extended them to a certain extent.

I hope that through this article, you can have a basic understanding of this hot track.

There are some premises that the report does not mention that need to be popularized first:

(1) AI Agent is not a native Web3 concept. As the name suggests, it is related to AI. Our main discussion is about what existing or potential gameplays have been created through the fusion of AI Agent and encryption.

(2) I don’t know why everyone defaults to calling it AI Agent. After all, I am a creator who mainly uses Chinese, so I will translate it as [AI Agent] in this article.

1. What is AI agent❓

A more written description: refers to an AI-driven intelligent agent that can perform specific tasks in a specific environment.

  • Let me understand it as an AI-driven clone, or an outsourced worker.

It is different from the AI ​​model we know. Since my understanding of AI is relatively broad, when I was looking up relevant information, I found a tweet that I think explains it very clearly. You can also check it out. It can help you understand: The relationship between AI agents, encryption, and large prediction models. 丨https://x.com/cwweb3/status/1849645682228965872?t=c68RXv4djNAY2Miucuu6ow&s=19

If you don’t want to visit the above tweets, there is a comparison table listed in the original report to tell you: What are the differences between AI agents and our common robot models?

 

From the comparison above, I think it has been stated clearly enough, so next I will cite specific cases to tell you about the rise and development of AI agents in encryption.

2. The Rise of AI Agents in Crypto: Terminal of Truths & $GOAT

(hereinafter referred to as TOT)

-Timeline

(1) In March 2024, AI developer @AndyAyrey created Infinite Backrooms, which linked two Claude models and allowed them to converse without human intervention. This is where the story begins.

Increasingly strange conversations led to the establishment of a meme narrative pseudo-religion called "Goatse of Gnosis", which is based on an early internet (90s) meme called "Goatse". You can search for related information yourself.

(2) In June 2024, Andy trained the Llama-70B AI model based on Infinite’s chat records.

This model is called Terminal of Truths ("ToT"), and ToT then begins posting regularly, slowly developing its own personality, and then telling stories related to "Goatse".

(3) In July 2024, a16z’s co-founders funded TOT with $50,000 in BTC. Then in the following months, as TOT continued to promote “Goatse”, someone created a $GOAT meme on Pump Fun.

(4) This Meme token was then recognized by TOT, and its creator sent 1.93 million $GOAT to TOT’s wallet.

- Notes

A key point to remember is that Andy Ayrey, who we mentioned initially, retains a great deal of control over TOT, for example:

(1) ToT’s tweets: Andy can manually approve tweets generated by ToT, but he cannot add his own input or any context. The content of tweets on the ToT account is purely AI-generated.

(2) ToT’s wallet: Andy keeps ToT’s wallet. ToT cannot conduct transactions on its own.

(3) Roadmap: Andy also said that he would not adjust $GOAT’s holdings until the roadmap for TOT is released...

This is the story of AI tokens and $GOAT. The price increase of $GOAT tokens directly fills the narrative of AI agents with a rich background story.

In any case, this story leads to a thought: What else can AI agents create?

Then some cases in the market began to gradually enter the industry's field of vision based on this thinking. The following will briefly introduce them:

3. Case Introduction

-Virtuals Protocol @virtuals_io 

Virtuals Protocol is a platform focused on allowing users to create, deploy, and monetize AI agents. In their own words, they “provide a plug-and-play solution that allows games and consumer applications to easily deploy artificial intelligence agents.”

The core idea is that users can start entertainment-focused AI agents and generate tokens. The AI ​​agents will interact with users and generate revenue, which will then empower token equity through a buyback and burn mechanism.

Each time an AI agent is created, 1 billion tokens specific to that agent are issued and then added to the liquidity pool.

➢ Users who believe in the potential of this agent can purchase these tokens and participate in key decisions regarding the agent’s development, behavior, and future upgrades

➢ This means that users can have potential financial gains in various agents, provided that the AI ​​agent can actually make money

The process of creating proxies is called an Initial Agent Offering (“IAO”), and essentially, tokens representing new proxies are locked in a liquidity pool paired with $VIRTUAL tokens, and each proxy token is launched fairly with no internal allocation or pre-mine.

I will not go into more detail, otherwise it would not be considered an introduction. To put it simply, this is an AI agent version of Pump fun.

-daos.fun @daosdotfun 

daos.fun allows the use of DAO structures to create AI-led hedge funds. Although the platform was originally designed for real people, it has now adopted AI agents as the overall framework, and their largest fund managers are now AI agents.

Currently, the creation of the DAO fund relies on centralized authority review, and the most famous case on its platform is the widely discussed ai16z.

This is a developer named Shaw who created the AI ​​agent version of Marc Andreesen (a16z general partner) and the associated hedge fund ai16z.

The idea also got the real Marc Andreessen to tweet about an AI hedge fund, which helped ai16z gain ground on the daos.fun platform.

Subsequently, ai16z became the largest hedge fund DAO on the platform, with a current market value of nearly $100 million.

In fact, I have come across many business frameworks based on DAO, especially in the previous social cycle, and I have personally participated in the construction of such protocols.

I don’t think the DAO narrative itself is sexy, but the AI ​​agent foundation created by $GOAT has provided a template for the market.

A genuinely somewhat sexy AI agent, rather than any way of achieving it.

4. What will the future be like?

Although the AI ​​agent track is still in the very, very early stages in Web3, it is a potential growth area worth paying attention to.

It is also an innovation in the way tokens are issued.

And, given that the AI ​​agent can operate at a high level 24/7, in a way, the AI ​​agent can act as the best KOL in the world because it does not need to rest, even @Phyrex_Ni can hardly compare to it in this regard

Compared with human-managed funds, AI-managed funds may have higher advantages. For example, the possibility of discovering hidden Alpha may be quite high, which can truly make AI help you trade cryptocurrencies.

These are some ideal scenarios. In fact, I think the most important thing to note in this article is that AI agents may bring a more comprehensive and less controlled token issuance method to the industry, just like last year’s inscription.

However, when it comes to the upstream of the agent, the various AI models, I can say responsibly at this stage that they are not as omnipotent as imagined, and even when I use them, they often fail...

I hope this article can help you understand this track. Thank you for following @sjbtc9