AI Agent Track Thoughts: Will androids dream of electric sheep?

Fast forward to 2025, and we are all familiar with AI, but in fact many people do not understand the definition of Agent accurately enough.

In fact, if a project is named "AI Agent", it means that its problem-solving logic must follow the following steps:

1. Perception
2. think
3. action

For example, the most common AI Agent project [Using AI to automatically invest in cryptocurrency] works like this:

1. Perception: Read the latest crypto news
2. Thinking: Determine whether the news is good or bad for a certain currency
3. Action: Buy if it’s good news, sell if it’s bad news

Very easy to understand.

So, is ChatGPT, the product most familiar to us ordinary people, considered an AI Agent?

First of all, although ChatGPT itself considers it to be an AI Agent.

But it also admits that it is only in a broad sense:

Therefore, if we define "AI Agent" broadly enough (as long as it has perception, decision-making and action cycles), then ChatGPT can be regarded as a kind of AI Agent; but if the requirements for AI Agent include richer perception and Physical/system level execution capabilities, then ChatGPT is more like a conversational intelligent component and does not have complete multi-modal or physical environment operation capabilities.

Why? There's a word game here - Just imagine, if you want to use ChatGPT, you will only say: "I want to use AI" instead of saying an extra word to say "I want to use AI Agent".

Therefore, when we usually mention the term AI Agent, we must want to emphasize something extra.

Then, we further abstract the characteristics of AI Agent:

• It solves something more vertically
• It wants to get more data from the physical world

Solving vertical problems

Although for the future AGI era, GPT is only limited to AI that generates conversations, which is a standard vertical application.

But for 2025, GPT (and of course its friends gemini, grok, etc.) is already an all-round player in this era.

Therefore, from a definition perspective, these current AI Agents are more vertical than GPT.

More data from the physical world

AI Agent needs to have the ability to actively acquire knowledge about the physical world.

For example, the AI ​​Agent we are familiar with is currently the well-deserved top Twitter AIXBT (Mindshare ranking No. 1). It has undoubtedly obtained more data from the physical world, such as various currency circle news, so This purple frog can make all kinds of comments and predictions.

The one that impressed me the most was when aixbt predicted that a coin would be listed on Binance within 6 hours, but it turned out that it was actually listed on Binance that night.

Thus, we obtained the complete definition of AI Agent:

1. Adopt the "perceive - analyze - act" process
2. Solve a problem vertically
3. Use more data from the real physical world

Once we understand the definition, we can begin the discussion.

So, what can AI Agent do?

This question is actually a variant of "What can Crypto do?" AKA. It is a question that many outside cryptocurrency critics have spent their lives "questioning".

I have tried dozens of Agents in these four major frameworks, which I summarized into several major categories:

1. AI investment

I have seen a lot of this kind of thing. You can put money into the AI ​​Agent wallet, and then the AI ​​Agent analyzes the rise and fall based on real-time news, and the AI ​​makes buying or selling decisions.

Todd’s comment: I think it’s interesting, but the security of the wallet is questionable.

2. AI girlfriend/boyfriend

In fact, this is what everyone often calls GPT cheating. But in fact, they have added some personality restrictions compared to the past.

For example, Eliza repeatedly emphasized to me: giggs, she is a real girl, not an AI.

For example, AVA, a beautiful white-haired girl, is good at making videos and analyzing the market through videos.

Todd’s comment: Political correctness is too great in real life. Facts have proved that everyone still likes childlike faces/big breasts…

3. Viewpoint output class

You can ask your AI Agent to review something vertically.

For example, AIXBT reviews cryptocurrencies and even does some research (such as promoting multiple projects on the same track).

Another example is Zerebro. Its concept is more fantasy, but overall, it still feels like a crazy artist who talks weird things every day.

For another example, there is a famous E/ACC (Effective Acceleration) real-person critic. Unexpectedly, a developer built an AI Agent to crazily imitate the way he speaks, and then publishes some weird and wonderful remarks.

Todd’s comment: Although many people doubt the existence of this thing, you have to know: in this era, as long as there is traffic, it is meaningful.

4. Idol category

Make an AI-based virtual idol that can release records, such as Luna

5. Prediction class

There is a project in Virtual that specializes in Polymarket predictions, which is actually a variant of opinion.

6. AI resource sharing class

For example, FXN is based on the Eliza framework. Its concept is that if you are reluctant to buy various AI memberships, you can buy its packaged and integrated version and pay per view to save some money.

Todd comments: AI version of putting together rice

7. Other categories

For example, those who draw, generate animations, make music, etc., I won’t mention them one by one here. Of course, overall, they do not deviate from the scope of these multi-modal large models currently being done by the real AI industry.

Wait, what about the AI ​​framework?

In fact, AI Agent is now in full bloom, which is quite similar to the ICO and DeFi Summer of the past.

If we use this analogy, then AI16Z, Virtual, and AVA are equivalent to public chains, and their frameworks can be used to quickly start everyone's AI Agent.

Of course, there is a fee.

For example, if you want to use AI16Z/Eliza to send AI Agent, it requires allocating some tokens to AI16Z's treasury and using AI16Z coins as collateral. This is why there is more and more money in AI16Z's treasury.

Use the Virtual framework to send an AI Agent to collect 100 Virtual coins, and pair your own tokens with Virtual to form LP, which is equivalent to SOL to pump.fun.

AVA and Swarms are generally similar to the previous two, but because they are a little later, the ecosystem above is not as rich as the previous two, but they are still making progress.

The special thing is that AVA was formerly Holoworld, an AI project of the Binance Incubator, and decisively transformed into an AI Agent, which is good at video generation.

The portrait of the founder of Swarms is that of a genius boy. This framework has been running in the tradition for many years. It is generally more technical and emphasizes that multiple agents can be used to cooperate and relay results (the origin of the name swarm).

Of course, there is also a tidbit of news, that is, he was severely insulted by Shaw from AI16Z, saying that his skills were not good, and he asked him to suck his dick.

But then again.

Applications are applications and tokens are tokens.

The most important function of these frameworks, or the biggest difference from traditional AI frameworks, is that they can help developers issue coins.

Or to put it more bluntly: in essence, these AI frameworks are also modified from the existing frameworks in the current AI industry. They stand on the shoulders of giants, and their most important originality lies in the currency issuance module.

Imaginary AI developer: OpenAI founding employee, Stanford Ph.D., annual salary 1M

Actual AI developer: first issue a coin and then talk about it

So, thanks to these frameworks, if you want to issue an AI Agent, your process is like this:

1. Determine the theme

2. Find a framework you like and pay for it

3. Coin listing

4. Create a Twitter account for the AI ​​Agent to talk nonsense

5. Inject more functions into this AI Agent (requires development)

6. Develop official website, build community, and contact CEX

7. Start a sect and publish your own AI framework

8. Encourage more people to identify topics…

Mortal Cultivation of Immortality be like ↑Wry smile

The corresponding valuations are (reference)

1

100

10K

100K

1M

10M

100M

1B

-------- In comparison, the processes of traditional AI companies:

1. Determine the theme

2. Financing

3. Burn money

4. Financing

5. Burn money

6. Financing

7. Listing

Did you see it? Crypto+AI actually just helps you get it listed without saying anything else.

Pure rebellion against Tiangang is a "subversive" innovation in the literal sense.

In addition, because the secondary market performance of these AI frameworks is so good (the market value can reach 1B or even 2B without being listed on any T1 or T2 exchange), they have become the only popular track for Crypto at present.

As of today, the streets are already empty of people:

It must be said frankly that it is indeed overheated in the short term. If there is an AI Agent fear and greed index, the current greed index is at least 90, which does not rule out the possibility that it has exploded.

In fact, we often see some real AI developers outside the circle criticizing these things as having no technical content.

However, my point of view is that the best use case for cryptocurrency is to [complete market discovery quickly], although it will bring some bubbles.

Good products and founders can quickly gain exposure and community support, while founders of bad products will immediately encounter negative validation from the market.

Under such intense cultivation, products that traditional AI developers can’t expect may blossom.

Looking forward to that day.

PS: It’s written at the end. It’s so funny. AI is awesome. After I finished writing this article, I asked AI to help me think of an “article title”. As a result, it actually said:

I really want to click on the topic "10x Coin". . . It hits the G-spot of people in the currency circle too directly.

But I still followed my heart and chose the title from the original "Blade Runner": Do Androids Dream of Electric Sheep?

I don’t know if AI will dream about electronic sheep, but I know that humans may be worried about electronic sheep.

If AI develops for another 20 or 30 years... it’s a future that I can’t even imagine.