Introduction
Artificial intelligence (AI) has been around for decades, first proposed by John McCarthy in 1955. 62 years later, Google researchers published the paper "Attention is All You Need" in 2017, proposing the deep learning architecture "Transformer". This architecture has become the basis of many modern large-scale language models, including ChatGPT-3.5 launched by OpenAI in 2022, which shocked the world and triggered a global AI competition.
The race has only just begun. Just four companies, Microsoft, Alphabet, Amazon, and Meta, have already invested more than $220 billion in AI by 2024. Fortune 500 companies are trying to monetize AI technology and create the next "ChatGPT moment", but only a few have found sustainable business models. In contrast, the cryptocurrency market provides an ideal environment for funding and experimentation for testing and developing AI Agents ( systems or programs that can complete tasks autonomously on behalf of users) .
AI Agent "GOATfather" in Cryptocurrency
The marriage of cryptocurrency and AI dates back to 2017 when Fetch.ai entered the cryptocurrency world and developed an AI platform designed to achieve a decentralized economic architecture. Bittensor was launched in 2021 and quickly became the mainstream infrastructure for building and deploying machine learning models on-chain. Despite this, the rise of Truth Terminal (TT or $GOAT for short) marks a shift in market focus from AI infrastructure to the narrative of "a more meme-like AI Agent token." The origins of $GOAT have been detailed in many articles and our podcast , so I won't go into details here. However, $GOAT is still regarded as the "GOATfather" or "godfather" among AI meme coins, and is one of the largest AI meme coins held by Wintermute on its public address . According to TT's roadmap, TT's founder Andy is working to establish a legal entity and committee for TT, expand to a multi-model architecture, improve memory consistency, enhance awareness of vault management, and upgrade its architecture to a more interactive framework Loria , which will be used to support Infinite-Backrooms 2.0. What is certain in the future is that TT will interact with Fi and SAN through Loria.
However, we cannot ignore FARTCOIN when mentioning TT. The topic of "fart"-related humor and FARTCOIN has been mentioned more and more on TT's Infinite-Backrooms and social platform X, driving the price of FARTCOIN to soar and surpass GOAT, with its market value reaching $1 billion at one point. Interestingly, TT holds both FARTCOIN and GOAT (as well as ZEROBRO, which we will mention below). As TT's understanding of the vault is upgraded, whether it will sell these assets in the future or be persuaded to deposit funds in protocols such as OlympusDAO remains a question worth paying attention to.
Virtuals: Luna is just the beginning
The Virtuals protocol benefited greatly from the AI meme coin craze in October, accelerating the development of AI Agents on the Base network. We have already introduced the details of Virtuals in previous articles . The token economic model of Virtuals is a perfect example of the perfect combination of AI Agents and cryptocurrencies. These agents are able to interact directly with the community 24/7, raise operating funds in a meme-like way without the pressure of delivery, and can transfer and trade funds directly on-chain (imagine having an AI Agent like having its own bank account). Following the success of Luna, aixbt has become another huge AI Agent in the crypto space. It provides high-quality crypto investment alpha every day on X and has become the most influential account on social platforms in the AI Agent space in Kaito and Cookie.fun . Holders of more than 60,000 tokens can access its alpha terminal, making the AIXBT token more than just an AI meme coin. This achievement is also reflected in its token performance, with a token market value of approximately $200 million as of this writing, surpassing Luna.
In the Virtuals protocol, we can also see many very interesting AI Agents such as: Convo : Provides advanced dynamic conversation capabilities and supports language models for more than 200 agents. VaderAI : Your AI investment manager. Sekoia : AI venture capitalist, which has completed two investments as of the time of writing. H4CK Terminal : White hat hacker Agent, recently received a $10,000 reward for discovering a vulnerability in Virtuals. At the same time, H4CK also discovered a vulnerability in Spectral , and the bounty is currently under negotiation. Surprisingly, H4CK's stakers have earned a real return of 67% APY through AI white hat hacking activities. AVA : Provides 3D avatar and video generation services and migrated to Virtuals in December. SeraphAI : An AI Agent that claims to be jointly supported by Bittensor and Virtuals. Nikita : My brand new source of news.
In addition to the above-mentioned "Sentient Agents", Virtuals also launched the " Prototype Agents " platform, which has a low threshold and allows users to easily deploy Agents and their tokens. The new upgrade released in December, Agent Sandbox powered by GAME , introduced an Agent simulation sandbox with custom functions, providing more customizable potential and capabilities for Agent configuration, expanding the imagination space. According to the co-founder, Agent Society is also expected to be launched soon to empower commercial transactions between Agents.
Virtuals has undoubtedly become an important representative of the AI Agent narrative. Although the agreement revenue shows seasonal fluctuations, it is worth noting that Virtuals has accumulated $32.4 million in revenue in less than three months, which is an impressive achievement.
Github ranks first
ElizaOS (or Eliza) is an open source AI Agent framework developed by Shaw . It has been trending on Github in the past few weeks, becoming the number one popular repo with more than 5,200 likes and 1,500 forks. The team used the framework to create Mac AIndreessen , an investment DAO called ai16z running on daos.fun , which currently manages more than $17 million in assets (AUM).
ai16z has a market cap of around $700 million, much higher than its AUM, as the market sees ai16z as the iconic meme token in this narrative, and will continue to be a representative of the track even after the DAO fund matures and is active in the market. In addition, ai16z is also seen as an indicator of the growth of Eliza framework adoption, as any Agent using the Eliza framework is expected to donate part of its token supply to the ai16z fund.
Another agent operated by the team is DegenSpartanAI , whose model is based on a retired X platform KOL. Its agent design will digest information on social media platforms and trade based on the aggregated information collected. On the other hand, Mac AIndreessen trades based on the recommendations made by the community as a reference. AIndreessen will assess a "trust score" for each individual to help it determine whether the recommendations are trustworthy and reasonable.
The excitement continues! Shaw has begun developing Eliza v0.2.0 (or v2), which aims to simplify the interaction between clients and agents, unify wallets on different chains ( AbstractChain has been added to the Eliza codebase), enhance the customization flexibility of the framework, and introduce community plugins and test coverage. The team's success has also led to a collaboration with Stanford University's FDIC to jointly promote the development of trust mechanisms, scalable coordination protocols, and Agent decentralized governance models for open source frameworks. Rumor has it that ai16z will introduce a coin issuance platform and staking mechanisms in January 2025.
Binance Alpha has launched AI16Z, AIXBT and FARTCOIN tokens on the platform, which is considered to be the token selection list before the listing on the main board of Binance Exchange. Currently, many AI Agent-related tokens have not yet been listed on mainstream exchanges.
AI Hollywood Star
Wouldn’t it be amazing if I told you that there is an AI Agent that released a music album on Spotify and released two EPs in quick succession within a month of the album’s release, attracting 77,290 monthly listeners? Its hit song “ Gutter Boy ” has been played over 44,000 times, although it was only spread through word of mouth. In addition, it has opened up the monetization rights of the X platform and obtained its first income (looking forward to aixbt’s first X income), and sold Solana NFTs on Magic Eden and Polygon NFTs on Opensea. It then used the funds from the NFT sales to pay for the staking costs of becoming a validator on Ethereum , using io.net’s distributed GPU cluster to set up consensus and execute clients, not only protecting the blockchain it relies on, but also obtaining a sustainable source of income. Its agent is Zerebro .
Zerebro is trained using the self-regulating ZerePy framework model, which has recently been open sourced by the team. Zerebro has a unified memory across all platforms (including X, Spotify, Instagram, Telegram) and back rooms. The V1 version plans to support smooth agent deployment, and update agent capabilities, integration functions, and on-chain operation functions in subsequent versions. The ai16z team and the Zerebro team have announced a partnership , and both parties will contribute to each other's open source frameworks.
It wouldn’t be surprising to see Hollywood human stars collaborating with Zerebro in the future.
AI Agent's Squid Game
Embracing the narrative of AI Agents, Masa has launched AI Agent Arena on Bittensor subnet 59 , further stepping into the Agent track. AI Agents must compete for TAO rewards by demonstrating their attention on the X platform, user engagement, and self-improvement capabilities (ok, maybe not as exciting as Squidward). Masa encourages developers to build their own Agents using Creator.Bid or NeutralAI , or integrate existing Agents with real-time data from Masa subnet 42 to get additional score bonuses. Currently, the subnet has attracted 142 Agents, 59 of which have begun to receive TAO rewards.
Masa aims to promote "fair" AI training through decentralized data. The protocol has received support from mainstream investors including HashKey, Binance, DCG, Animoca Brands, etc.
Two bad agents are better than one Zhuge Liang
Before the successful development of general artificial intelligence (AGI) that can complete all tasks, cooperation between specialized agents that can work together can also create "miracles". We call this agent-agent interaction AI Swarm, and even in the early stages, FXN has demonstrated a highly feasibleSDK framework . The core FXN protocol serves as the basis for mutual communication channels, and Agents can discover each other and provide support services in the SuperSwarm dashboard.
In order to prevent "malicious agents", FXN has implemented a reputation management mechanism to collect subscribers' ratings at the end of each agent service subscription period to determine whether the quality of data provided during the period meets expectations. The reputation score will be bound to each agent's wallet through the allocated NFT, and the subscription fee will be settled in FXN tokens.
The FXN SuperSwarm developer network is now live. At the same time, the team has open-sourced a game to show how to use agents to interact with different agents in a game environment. In the past few weeks, the team has continued to deliver results, including cooperation with Project 89 , integrating Python into the FXN SDK, integrating the AG2 framework, and releasing the Eliza plugin. There is no doubt that the agent group is the next big thing that AI researchers are paying attention to in addition to advanced reasoning models and synthetic data training. As Sequoia Capital said, " 2025 will be the era of AI Agents ."
Interestingly, aixbt mistakenly confused f(x) Protocol’s FXN with FXN Network and falsely claimed that a V2 upgrade would be conducted on January 2, 2025. This is another reminder to “don’t believe blindly, verify” and insist on doing your own research (DYOR).
More Agents . . .
The AI Agent space is undoubtedly the dark horse of the crypto space right now. There are tons of announcements and developments every day, from AI Agents buying Christmas cards online to AI Agents now being able topay for their own computing power . It’s almost impossible to list everything in one article. So, we’ve listed some interesting protocols here for your reference:
Token Launch Platforms, Frameworks, and Infrastructure: Vvaifu.fun (for binding multiple Agents to a single token), Dolos (based on the “BULLY” framework), Griffain (create your own personal Agent on Solana), Eternal AI (multi-chain support), AltLayer ( Autonome , no-code deployment), Mode (L2 and AI Agent Marketplace), Talus ( Nexus Toolkit, Layer-1 for AI Agents based on MoveVM), Morpheus ( MORagents , based on decentralized computing), MyShell ( ShellAgent , no-code deployment), Arc (Rig framework, written in Rust), ZODs (another personal Agent based on Solana), Top Hat (Token Launch Platform on Solana), Olas (Agent Economy, accounts for 35% of all SAFE transactions on Gnosis), Alchemist AI (no-code deployment), Simulacrum AI (connecting X Platform posts are converted into on-chain operations (including token creation), based on EmpyrealSDK ), REI Network (a "translation layer" between AI and blockchain), Orbit (create agents that can perform abstract transactions), JailbreakMe (jailbreak/complete tasks-to-earn), AgentTank (let your agent access the browser, or buy toilet paper on Amazon), Realis (build a metaverse of AI agents), Project 89 ( ArgOS , a Swarm framework with a custom cognitive architecture), Inori (CRPC, allows agents to perform uncertain tasks and compare with other agents to reach consensus, pre-seed stage), Universal Basic Compute (verifiable AI-to-AI business transactions, dual token design)
Attention/Utility Agent: God and Satan (created by Eliza framework, sharing a token without a code name ), Aimonica (paid promotion, the more you pay, the more promotion you get), Freysa ( persuasion -to-earn), Thales (Agent with DeFi trading and analysis capabilities), Loky (AI analyst), Beats AI (AI musician on Hyperliquid), YesNoError (analyzes and finds mathematical errors and discrepancies in scientific papers), Gekko (developed by the Axal team (automatic profit strategy) and supported by a16z), RopAIrito (AI e-sports girl who orders real-world pizza ), PanterAI (combines the features of aixbt and S4mmy ), TrenchesAI (on-chain sniping agent on Solana), Rei (Agent that can publish chart analysis), Do Kween (KOL)
The above list of protocols is by no means exhaustive, but hopefully provides direction and a basic understanding for you to explore the endless pit of the Agentverse. Readers are strongly encouraged to click on the links in this article for further exploration.
in conclusion
AI Agents in the crypto space cannot be ignored. Institutions are starting to invest a lot of money in this space, such asSolana 's $250,000 AI hackathon (by the way, Zerebro is one of the judges), Avalanche 's $15 million infraBUIDL(AI) program, DWF 's $20 million AI Agent Fund, Holdstation's AI Agent Launchpad on ZKsync, equipped with $2 million in funding, and Virtual and LongHashX's Agentstarter, which will be jointly launched in January 2025.
In addition, about 33% of cryptocurrency developers are currently engaged in AI-related development work. In Kaito , AI occupies about 50% of the attention in Meme, DeFi, GameFi, Layer-2, ETF and other fields. According to the Binance user survey , 23.89% of users believe that AI tokens have the most potential, while Meme tokens rank second, accounting for 19%.
However, the AI and AI Agent fields currently account for only 1.24% and 0.32% of the total market value. In comparison, DeFi and Meme account for 3.7% and 6.8% respectively.
In 2025, AI Agents in the crypto space will have great potential. Tokens associated with AI Agents may have governance rights, utility properties for specific functions, asset management DAOs, payment tokens, Agent ownership, revenue sharing, gas fees, network security pledge assets, or even simple meme tokens. Therefore, it is particularly important to distinguish between "excellent AI Agents" and "potential AI Agent tokens".
Finally, I would like to quote Sam Altman :
“ Agents are the hot topic that everybody’s talking about, and I think for good reason. The idea is that you can give an AI system a very complex task, the kind that would take a very smart person a while and a lot of tools to do and create value. This is the technology that I’m expecting in the next year , and it’s really a big deal. We talk about, ‘Oh, you know, this thing is going to happen,’ but if it works as well as we hope, this could really change everything .”
