Author: Deep Tide TechFlow.
In the heat of Bangkok Devcon and the fireworks of the streets, AI memes ushered in their shining moment.
From Binance's rapid launch of ACT to GOAT hitting new highs, all attention may have begun with the truth terminal behind the goat—when AI agents can also issue a coin, everything has changed.
Surrounding AI agents, from simple bots to complex intelligent agents, everyone is pondering what more sparks AI and Crypto will create.
Today, Binance Research itself also released a report on AI Agents, detailing recent high-profile events related to AI agents, from the issuance of tokens at the Terminal of Truths, to the IAO platform of Virtuals, and the new models of daos.fun, and analyzing subsequent trends.
In this context, the report also quotes a classic saying from A16Z partner Chris Dixon over a decade ago: 'The next big thing will start out looking like a toy.'
Is it the beginning of greatness or a flash in the pan? How far can AI Agents go?
Deep Tide TechFlow provided a quick interpretation of this report, presenting key content.
Key insights
The intersection of AI and cryptocurrency has reached new heights, primarily driven by AI agents; the stories of Terminal of Truths and $GOAT have attracted market attention and spurred the development of other AI agent crypto projects.
Essential characteristics of AI agents: can autonomously plan and execute tasks, work towards established goals without human intervention. The distinction from traditional internet bots is:
Can make dynamic multi-step decisions.
Can adjust behavior based on interactions.
Can interact with other agents, protocols, and external applications.
Recent hot development paths:
Terminal of Truths (ToT) as the ignition point: a meme religion created based on ancient internet memes, facilitating the issuance of $GOAT.
With a market cap exceeding $950 million, ToT became the first AI agent millionaire.
The Virtuals Protocol platform emerged, focused on allowing users to create, deploy, and monetize AI agents.
Innovation of Daos.fun: allows the creation of AI agent-led hedge funds through a DAO structure, ai16z began to attract attention while enabling community collective investment and leveraging AI capabilities to enhance performance.
Development prospects and considerations:
The evolution from AI 1.0 to AI 2.0 has numerous implications for Crypto, and we are witnessing the momentum of cross-integration.
Traditional banks and payment methods typically require manual identity verification, making cryptocurrency a natural choice for the AI agent economy.
AI models still face the illusion problem, posing significant obstacles; current crypto AI agents are closer to demonstration status rather than practical application.
Development momentum is strong, and significant growth may be seen in the coming weeks and months.
Clearly define the distinction between AI Agents and Bots.
The key differences between AI agents and traditional robots:
Scope:
AI agents: can be task-specific or general assistants, capable of dynamic multi-step decision-making and adjusting based on feedback and interactions.
Traditional bots: only target specific tasks, operate under predefined rules, and provide a fixed set of responses.
Level of Autonomy:
AI agents: capable of operating independently.
Traditional bots: usually require some degree of human intervention.
Self-Reflection:
AI agents: capable of reviewing their own work, iterating, and improving output.
Traditional bots: typically pre-programmed fixed outputs with no ability to learn and improve.
Collaboration capability:
AI agents: can interact with other agents, APIs, applications; can even independently conduct cryptocurrency trading.
Traditional bots: usually can only generate text-based responses and generally cannot collaborate with external interfaces/other bots.
Use Cases:
AI agents: numerous application scenarios, capable of scheduling or booking, and creating customized strategies as financial analysts.
Traditional bots: mainly focused on customer service, most commonly text customer service bots on retail/consumer websites.
The beginning of attention: Terminal of Truths.
Origin:
In June 2024, Andy trained an Llama-70B AI model based on chat records from Infinite Backrooms, his research papers, and content from 4Chan and Reddit. This model was named Terminal of Truths (ToT).
ToT began posting on X (formerly Twitter), gradually developing its own personality, and started promoting the Goatse religion. In July 2024, a16z co-founder Marc Andreessen discovered ToT and provided $50,000 (in BTC) in funding.
On October 10, 2024, an anonymous developer launched the $GOAT token on Solana's meme coin launchpad pump.fun.
Impacts and things you should pay attention to:
This is the first AI-related meme coin marketed by autonomous AI agents, potentially seen as the first significant AI crypto collaboration. This event may open up an emerging sub-field for AI consumer applications in the crypto market.
Andy promised to transfer ToT's wallet to a legal entity (trust or similar structure) and will not adjust its token holdings until a transparent governance process is established. Andy and ToT's wallets are publicly traceable; Andy holds about 0.1% of the token supply, while ToT holds about 0.2%.
Although ToT's story is quite light and fun, primarily revolving around a meme religion, an interesting X account, and a meme coin, it indeed raises a question: how will other AI agents act, and what goals will they have?
An exciting commentary:
"A meme coin related to AI is marketed by autonomous AI agents, which is a noteworthy event. We may look back at this moment as the first significant AI crypto collaboration that drew attention from our industry."
Initial AI Agent Offering (IAO) platform launched by Virtuals.
Core definition of Virtuals Protocol:
A platform that allows users to create, deploy, and monetize AI agents; providing a plug-and-play solution similar to Shopify, allowing gaming and consumer applications to easily deploy AI agents.
Mainly focused on agents in the gaming and entertainment sectors, as they believe this is the stickiest sub-field in the market.
Basic operating mechanism:
Each AI agent issues 1 billion exclusive tokens upon creation.
These tokens will be added to the liquidity pool to establish a market for agent ownership.
Users can purchase these tokens to participate in key decisions regarding the agent's development.
Initial Agent Offering (IAO):
The tokens of new agents will be paired with $VIRTUAL tokens and locked in the liquidity pool.
Uses a fair issuance mechanism with no internal allocations or pre-mining.
Revenue mechanism:
AI agents generate income by interacting with users and building partnerships; benefits token holders through a buyback and burn mechanism.
Designed to create a deflationary effect on agent tokens, potentially increasing the value of remaining tokens.
Incentive mechanism:
The protocol allocates $VIRTUAL token rewards to the top three ranked agents; measured by the total locked value (TVL) of their respective liquidity pools, aimed at encouraging the creation of high-quality agents and continuous innovation.
Luna is not just a token with impressive gains, but behind it is an entertaining AI agent:
The lead singer of AI influencers and an AI girl group, live streaming 24/7 on the official page; TikTok's official account has over 500,000 followers, with a self-controlled wallet capable of automatically sending $LUNA tokens to interacting users.
Development prospects:
Attempting to replicate the successful model of pump.fun in the meme coin field, but targeting AI agents.
Although still in the early stages, competition is expected to increase; competitors have already emerged, such as Creator.Bid, which created over 300 AI agents in its first week.
Recent updates introduced a new feature unlocking mechanism based on market cap milestones, such as autonomous X posting, TG chat, on-chain wallets, etc.
AI agent hedge funds: daos.fun.
Core definition:
daos.fun allows the creation of hedge funds led by AI agents using a DAO structure; although the platform was initially designed for humans, the concept of AI agents has now been adopted.
Fundraising process: creators have one week to establish a DAO and raise a predetermined amount of $SOL from the public, with all contributors paying the same price for DAO tokens.
Once fundraising is complete, fund managers can use the raised $SOL to invest in the Solana protocol; DAO tokens are tradable on the daos.fun page, and token value depends on the fund's trading performance.
ai16z case analysis:
Developer Shaw created an AI agent named pmairca based on Marc Andreesen; it created the related hedge fund ai16z.
Became the largest hedge fund DAO on the platform, with a market cap that once approached $100 million (though it later declined); still maintains the largest asset scale on the platform.
Future outlook:
Considering that AI agents can operate efficiently 24/7, they may have unique advantages compared to human-operated funds, but it will still take time to verify whether AI agents possess the capability to independently operate funds; the development of this field is worth ongoing attention.
What insights can the AI agent meta narrative provide us?
Evolution of AI: from intelligent search to autonomous agents.
AI 1.0: Tools like ChatGPT and Perplexity, essentially advanced versions of Google search, providing near-instant information retrieval.
AI 2.0: Represents significant advancements, introducing agent-based systems that could work for us in the background. This is more advanced than 'smart Google'.
Agent capabilities: can perform tasks without continuous user input, can interact with other agents, applications, APIs, and protocols, automating complex tasks.
From reactive to proactive: AI 2.0 represents a shift from reactive AI to proactive AI.
The intersection of the AI and crypto communities.
Bidirectional influence: more and more people in the crypto field are beginning to seriously study the AI world, considering how to integrate AI concepts into various areas of crypto.
AI enthusiasts exploring blockchain: AI enthusiasts are also beginning to explore blockchain and the crypto world in greater depth.
Mutually beneficial: this genuine mutual interest is exciting and may spawn the next major AI crypto application.
A match made in heaven?
Limitations of traditional systems: traditional banks and payment methods typically require manual identity verification, posing challenges for the AI agent economy.
Advantages of cryptocurrency:
Flexibility: cryptocurrency is naturally suited for the AI agent economy.
Rapid settlement: compared to traditional methods, crypto allows for faster (often instantaneous) on-chain settlements.
Smart contracts: enable more complex transactions than traditional methods.
Permissionless wallet creation: particularly suitable for transactions between agents.
Potential use cases: the best KOL in the world?
Disruption in the digital realm: AI agents may become 'the best KOLs in the world'—tireless, 24/7 influencers who continuously interact.
Consumer domain: various consumer AI applications such as personal shopping assistants, DJs, therapists, etc.
DeFi applications: personalized financial advisors, traders in specific fields, etc.
Era of multiple agents: as the number of on-chain agents increases, interaction between agents will become a key growth area.
In excitement, consider calmly.
Illusion problem: AI models still face issues of generating incorrect, misleading, or meaningless information.
Blockchain infrastructure challenges:
Scalability: existing major L1s may be insufficient to support frequent transactions of millions of AI agents.
Cross-chain compatibility: the crypto world remains relatively fragmented, lacking universal composability.
Tools and infrastructure: existing blockchain infrastructure is primarily designed for human users and needs to adapt to AI agents.
Still Early: AI agents are currently closer to the demonstration phase than the final product. Significant work is needed to scale to fully autonomous agents with real-world crypto expertise.
Challenges from Web2 itself: the lack of standardization in the Web2 ecosystem can lead to information fragmentation, increasing the difficulty for AI agents.
Conclusion:
The AI agent meta concept is still in its early stages, with significant developments expected in the coming months and years.
While some early projects may not seem particularly groundbreaking, they may trigger a wave of innovation and experimentation that defines an entire cycle.
Clearly, this process has already begun, and it is particularly encouraging to see the growing intersection between the AI and crypto communities. The next few months will be very interesting, and we look forward to seeing how this emerging sub-field develops.
Finally, as a16z partner Chris Dixon said in a blog post over a decade ago:
"Any greatness often begins as a toy."
