From Meme to DeFi: How AI Agents Reconstruct the Web3 Value Network?
TechFlow Selected TechFlow Selected
From Meme to DeFi: How AI Agents Reconstruct the Web3 Value Network?
When the market is quiet, I'd like to share some thoughts on the entire web3 + AI agent space.
Navigating Web3 tides with focused insightsAuthor: CaptainZ
When the market is quiet, I’d like to share some thoughts on the Web3 + AI Agent sector.
Ever since AI emerged out of nowhere two years ago, it has been clear that this is a technological revolution comparable to the invention of the internet. Initially, however, most attention focused on large language models (LLMs), and it was unclear how AI would impact the real world.
If we compare LLMs to the brain, AI clearly possesses analytical and reasoning capabilities but struggles to interact with the physical world. This naturally led us to equip LLMs with various sensors and functional modules. For internet-related functions, integration is straightforward—just connect via API interfaces—and thus AI officially evolved into AI Agents.
Large language models are undoubtedly important, as they determine an AI Agent’s intelligence level. Currently top-tier models such as OpenAI's O3 and DeepSeek's R1 already surpass human PhD students in many areas.
Thus, in 2024, various industries began exploring AI Agent + specific sectors. For example:
-
AI Agent + e-commerce (AI assists directly in product selection, copywriting, and image generation)
-
AI Agent + internet (AI programming)
-
AI Agent + finance (AI analyzes data in real time for smart investment advisory)
-
AI Agent + supply chain (AI optimizes supply chain management and monitors production lines)
-
AI Agent + education (AI provides real-time tutoring and personalized learning)
Web3 has now grown into a massive industry, so naturally we’ve seen the emergence of
AI Agent + Web3
But Web3 itself consists of many subfields, so AI Agent integration takes different forms depending on the area, such as
AI Agent + meme
AI Agents directly replace humans in launching meme coins—the representative being goat fartcoin
AI Agent + DeFi
AI Agents directly replace humans in analyzing market data (aixbt), managing funds (degenai), and executing DeFi operations (griffain)
AI Agent + gaming
AI Agents participate directly in game commentary (luna), gameplay (freya), sports betting analysis (dwain), and AI game frameworks (digimon/dreams)
AI Agent + social
Act initially explored how agents could socialize, later evolving into a meme project
There are also cases like AI Agent + DePIN, but we won’t go into detail here.
AI Agent is already an implemented technology in Web2—not just empty hype. In Web3, it simply adds a token issuance process that people can speculate on.
People often say Web3 AI Agents are far behind those in Web2, but AI Agent + Web3 is inherently about transplanting Web2 technologies into Web3 applications—finding meaningful points of integration is what really matters.
I personally break down Web3 AI Agents into the following categories: AI Agent infra (infrastructure) and AI Agent (agents for specific use cases)
1. AI Agent Infra (Infrastructure)
1. Fine-tuned Large Language Models
LLMs serve as the brain of AI Agents. An un-fine-tuned LLM is like hiring a fresh graduate with no specialization. An LLM fine-tuned on industry-specific data is like a graduate who majored in that field. Currently, Lumo is the only project pursuing this path. The main challenge lies in collecting, cleaning, and labeling industry-specific data.
2. Frameworks
This space has many competitors and was the main market theme a couple of months ago, including ElizaOS, Arc, Swarms, etc. An AI Agent framework is essentially a set of rules for an LLM to invoke various functional modules, enabling standardized agent creation.
3. Launchpads
The above frameworks are usually open-source code, requiring users to know basic coding and deploy on their own servers. AI Agent launchpads, however, offer a SaaS version—providing no-code environments (users only need to input a few parameters) with built-in server hosting. While this further simplifies agent creation, it reduces flexibility. Well-known launchpads include Virtual, Vvaifu, and Avaai.
4. Special Functions
Other projects focus on specialized features to make AI Agents run faster, better, and stronger.
For example, Web3 is a unique industry due to frequent interactions with large sums of money, requiring AI Agents’ operating environments to be sufficiently isolated and verifiable (e.g., ensuring wallets are operated by the AI Agent, not a person). Notable examples include Phala’s TEE environment.
2. AI Agent (Agents for Specific Use Cases)
This direction is highly diverse, representing explorations and innovations at the application layer, further divided into:
-
Analytical: Analyze a specific domain and generate outputs
-
Operational: Analyze a domain and then take direct actions
Examples below illustrate this:
1. aixbt (@aixbt_agent)
The most well-known DeFi AI analyst, skilled in briefly analyzing market projects and posting content on X with interactive engagement.
2. Truth of Terminal (@truth_terminal)
A pioneer in the AI Agent space, known for making absurd claims and launching meme coins.
3. Ava (@AVA_holo)
Holoworld’s flagship agent, skilled in delivering chaotic video analyses of the market.
4. Yne (@yesnoerror)
Analyzes scientific research papers and corrects errors.
5. Buzz (@askthehive_ai)
Haven’t used it myself, but according to its description, users can interact with the system through natural language interfaces; the agent handles tasks including trading, staking, liquidity management, and market sentiment analysis.
The above is my analytical framework for the Web3 + AI Agent sector. Although currently in a downturn, AI Agents in Web2 are thriving and represent one of the hottest trends of 2025. They will inevitably bring more exciting products and imaginative possibilities to Web3—making this one of the few directions capable of continuously attracting attention and builders.
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
@hiCaptainZ
