AI Agents Are Here to Stay: Adapt or Be Left Behind

ear, we have witnessed the growth of Crypto for AI, taking root in the Crypto field and giving rise to several new tracks, such as "computing power assetization," "model assetization," and "data assetization," with representative projects being ionet, Olm, and Sahara.

Over the past year, we've seen a surge in interest in Crypto for AI, which has taken root within the broader Crypto domain and spawned several new tracks, including "computing power assetization," "model assetization," and "data assetization," with representative projects being ionet, Olm, and Sahara.

With Trump's victory, the crypto craze has once again picked up steam globally, while AI for Crypto is also on the rise, rapidly generating many new things that emerge from the collision of AI, AI Agents, and Crypto.

AI Agents issuing tokens

AI Agents posting and calling trades

AI Agents managing DAOs and fund trades

AI Agents trading autonomously

To provide a rough overview, the development of AI Agents in the Crypto field currently has two paths:

1. Top-down, developed by AI concept projects, which are more infrastructure-oriented. We will analyze this in detail in the next article describing the panorama of Web3 AI infrastructure.

2. Bottom-up, led by the AI Meme craze, driven by independent developers.

This article mainly discusses the second path.

Table of Contents:

1. OpenAI's official AI Agent is about to be released, which may trigger an AI Agent frenzy

As a leader in AI, OpenAI has long divided the path to the ultimate form of AI, AGI (Artificial General Intelligence), into five stages:

Basic AI (Emerging AGI): The initial stage of AI development, referring to AI capable of basic conversation and information processing, such as ChatGPT. It relies heavily on pre-trained datasets, and the AI's "IQ" (understanding and reasoning ability) is very limited;

Reasoners: An advanced version of basic AI, capable of advanced logical reasoning and solving complex problems;

Agents: AI begins to have the ability to create content or perform actions without human input, or at least to execute tasks under human guidance. Most current AI Agents are still at a relatively early stage, mainly completing complex tasks that basic AI cannot achieve through planning, reasoning, and tool invocation;

Innovators: AI at this stage can not only solve existing problems but also conduct independent research and development, innovating and evolving through learning, forming a virtuous cycle, and getting closer to humans;

Organizations: The final stage of AGI, where such AI systems can intelligently allocate tasks, work collaboratively, and complete complex tasks, similar to a team or organization, achieving a 1+1 greater than 2 effect.

If ChatGPT is at the first stage mentioned above, then the recently launched reasoning model o1 can certainly be classified into the second stage. The o1 model performs almost at the level of a PhD student in handling complex subjects like materialized biology. In the field of mathematics, it achieved an astonishing 83% accuracy rate in the International Mathematical Olympiad (IMO) exam, while GPT-4o could only correctly solve 13% of the problems.

However, while users' attention is still focused on second-stage models like o1, OpenAI has quietly extended its "tentacles" into the third stage: Agents. OpenAI is positioning AI Agents as the next "trump card," as the Next Big Thing, steadily moving forward with a clear release date set for 2025.

"Agents will be the next major breakthrough" ------ OpenAI CEO Sam Altman

"Letting GPT autonomously execute tasks will be a major focus next year" ------ OpenAI CPO Kevin Weil

Next year, AI Agents will become the focal point of competition among various AI giants, with OpenAI facing strong competitors, including Anthropic's Computer Use and Google's upcoming AI Agent named Jarvis, set to launch in December.

Recently, Bloomberg reported that OpenAI plans to release an AI Agent tool named Operator in January, capable of autonomously completing tasks on a computer under user instructions, such as writing code and booking itineraries.

It seems that the AI Agent framework Swarm, launched by OpenAI a month ago, is just an appetizer, with numerous real-world work and life use cases waiting for OpenAI and developers to implement. For example,

2. AI Agent + Crypto, where to start?

The productivity revolution brought by AI Agents will gradually permeate various industries, leading to higher efficiency, better services, and personalized experiences.

In the crypto world, it seems that the current Crypto is inseparable from finance, so the most natural starting point for on-chain AI Agents is DeFi, but first, AI must gain financial autonomy. Traditionally, AI faces many restrictions in financial activities, such as being unable to open bank accounts and lacking legal identity. In the Crypto world, registering an on-chain identity and wallet for AI Agents is a very natural thing. We can pre-install crypto wallets and smart contracts, set spending limits, execute transactions, and manage funds, granting AI Agents financial autonomy.

When AI Agents possess both autonomous decision-making capabilities and financial autonomy, the question of "what can they do" suddenly opens