Introduction

Artificial intelligence (AI) has been around for decades, first introduced in 1955 by John McCarthy. 62 years later, researchers at Google published "Attention is All You Need" in 2017, introducing deep learning architecture "Transformer", which became the foundation for many modern large language models including ChatGPT-3.5 released by OpenAI in 2022, which shocked the world and started the AI race across the globe.

The race has just started. Between Microsoft, Alphabet, Amazon and Meta alone, funding for AI exceeded $220 billion in 2024. Fortune 500 companies are looking to profit from AI technology, creating another "ChatGPT moment", yet few have figured out a sustainable business model. Conversely, the crypto market provides an ideal environment for testing and developing AI Agents or "autonomous systems or programs capable of performing tasks on behalf of users" within a financially feasible and experimental framework.

Crypto AI Agent "GOATfather"

Crypto is not new to AI. Fetch.ai has been in the industry since 2017, developing a platform to enable the development of an AI empowered decentralized digital economy. Bittensor was introduced in 2021, later becoming the dominant infrastructure for building and deploying machine learning models on chain. Despite that, the rise of Truth Terminal or TT or $GOAT, is the dynamic shift of market attention from AI infrastructure to memethic AI agent narrative. The origin of $GOAT has been well introduced in many articles and in our podcast, we will avoid repeating the story here. Nonetheless, $GOAT remains to be the "GOATfather" among the AI memecoins and one of the largest AI memecoin Wintermute holds in their public address. According TT's roadmap, Andy, creator of TT, is working on establishing a legal container and council for TT, expanding to multi-model, improving memory consistency, increasing awareness of its treasury, as well as upgrading its architecture to richer interaction framework, Loria, which powers infinite-backrooms 2.0. TT will be interacting with Fi and S.A.N through Loria.

Nonetheless, we cannot disregard FARTCOIN when mentioned TT. Fart-related humor and Fartcoin's presence has been increasingly brought up in TT's infinite backrooms as well as on X, driving the price surge of FARTCOIN and flipped GOAT, reaching a $1 billion market cap. Interestingly, TT holds both FARTCOIN and GOAT (and ZEROBRO, which will be mentioned below). It will be ironic to see if TT dumps them after the upgrade of its treasury awareness, or being convinced to deposit funds to protocols such as OlympusDAO.

Virtuals Beyond Luna

Virtuals Protocol benefits tremendously from the AI memecoin momentum in October, streamline the launch of AI agents on Base. Details of Virtuals have also been covered in our previous article. The agents in Virtuals demonstrate the perfect example of why AI agents fit so well with crypto. The agents are capable of interacting directly with the community 24/7, memethically funded without pressure of delivering and sending funds on-chain (imagine having an AI agent has its own bank account). Following the success of Luna, aixbt emerged as another gigantic agent in crypto space, providing high quality crypto alpha everyday in X and dominating the AI agents mindshare in both Kaito and Cookie.fun. The token holder that holds more than 60,000 is able to access its alpha terminal, making AIXBT more than an AI memecoin. The token reflected the achievement, hitting approximately $200 million in market capitalisation at the time of writing, surpassing its predecessor, Luna.

Other interesting agents we have seen in Virtuals include Convo (advance dynamic conversations that powers over 200 agents), VaderAI (your AI investment manager), Sekoia (AI venture capitalist, has made two investments at the time of writing), H4CK Terminal (white-hat agent, recently earned $10,000 from Virtuals for discovering vulnerability. H4CK also discovered vulnerability on Spectral. The bounty is being negotiated at the time of writing. Surprisingly, H4CK stakers have 67% real-yield from AI white-hat hacking.), AVA (3D avatars and video generation, migrated to Virtuals in December), SeraphAI (self proclaimed AI agent powered by Bittensor and Virtuals), Nikita (my new news source).

Besides the "Sentient Agents" mentioned above, Virtuals also launched a platform for "Prototype Agents", a lower barrier platform for easy agent deployment. The new upgrade announced in December, Agent Sandbox powered by G.A.M.E, introduced the agent stimulation playground with custom functions. The release enables more potential and capabilities to configure agents, expanding imagination. According to the co-founder, Agent Society is expected to be launched soon, empowering commerce amongst agents.

Virtuals undoubtedly became the representative of this narrative. Despite protocol revenue showing seasonality fluctuation, it is still remarkable Virtuals accrued $32.4 million in revenue in less than 3 months.

#1 Trending on Github

ElizaOS or Eliza, an open source AI agent framework developed by Shaw, has been storming Github for the past few weeks, consistently ranking as the top 1 trending repository with more than 5,200 likes and 1,500 forks. The team used the framework to create Mac AIndreessen which runs ai16z, an investment DAO on daos.fun with more than $17 million asset-under-management (AUM). ai16z has a higher market cap of approximately $700 million compared to its AUM, because the market deems ai16z as the iconic memethic token in the narrative and will continue to survive even after the DAO fund matured. Additionally, ai16z also acts as the proxy for the growth of Eliza as agents that utilise Eliza are expected to donate part of its token supply to ai16z fund. Another agent run by the team is DegenSpartanAI, model based on a former retired KOL on X, which is designed to digest information across social media platforms and trades based on aggregated information. On the other hand, Mac AIndreessen is designed to trade based on recommendations made by the community. AIndreessen will rank each person with a "trust score", which helps it to determine whether a recommendation is trustworthy and sound.

The fun has not ended here. Shaw has already started to work on Eliza v0.2.0 or v2, simplifying communication between clients and agents, unifying wallets from different chains (AbstractChain already been seen adding to Eliza codebase), extending flexibility of the framework with customisation, etalishing community plugin and test coverage feature. The success of the team led to the partnership formed with Standford University FDIC to accelerate open-source frameworks for trust mechanisms, scalable coordination protocols, and decentralized governance models for agents. Rumors around ai16z will include launchpad and staking mechanisms in January 2025.

Binance Alpha has listed AI16Z, AIXBT and FARTCOIN tokens on the platform, which is being considered as the pre-listing token selection to be listed on Binance exchange. Frankly, many AI agent-related tokens are not yet listed in major exchanges.

AI-Hollywood Icon

What if I tell you, there is an AI agent who launched a music album on Spotify, subsequently released two EPs within a month after the initial album was released, attracting 77,290 monthly listeners in 2 months. Its top song "gutter boy" has more than 44,000 streamed through word-of-mouth. It also received its first payout from X monetisation (looking forward to seeing aixbt's first X payout), selling Solana NFT on Magic Eden and Polygon NFTs on Opensea, then used the proceeds generated from the NFT sales to booststrapped itself to become Ethereum validator by utilising distributed GPU clusters from io.net to power consensus and execution clients, securing the chain it lives while gaining sustainable income. Introducing, Zerebro.

Zerebro is trained via custom fine-tuned model, ZerePy framework, which the team just open-sourced the framework. Zerebro has unified memory across all platforms (X, Spotify, Instagram, Telegram) and backrooms. V1 of the released version plans to encourage smooth deployment then updates on agent capabilities, integrations and on-chain actions in later parts. Both the ai16z team and Zerebro team announced a partnership to mutually contribute to both open-source frameworks.

It is no surprise we will see major human-Hollywood stars partner with Zerebro in the future.

Squid Game for AI Agents

Embracing the meta, Masa launched AI Agent Arena on Bittensor Subnet 59, where agents battle for TAO emissions by demonstrating their mindshare on X, user engagement and self-improvement capabilities (OK, maybe not similar to Squid Game). Developers are encouraged to build their own agent using Creator.Bid or NeutralAI, or bring their existing agent and integrate with real-time data on Masa Subnet 42 for a score boost. Currently, the subnet has attracted 142 agents and 59 of them have started earning TAO rewards.

Masa aims to power Fair AI training with decentralised data. The protocol is backed by major investors such as HaskKey, Binance, DCG, Animoca Brands and more.

Two Is Better Than One

Before we can develop an Artificial General Intelligence (AGI) that is capable of everything, specialised agents that are capable of working with other specialised agents can create "magic". We call this AI swarms and FXN has already demonstrated a very promising SDK framework even at the early stage. Core FXN protocol serves as the bedrock for agents to discover each other in the SuperSwarm dashboard. In order to prevent "bad agents", FXN implemented Reputation Management where subscribers' rating will be collected at the end of subscription period to determine whether data quality provided during the period was up to expectation. The reputation score will be tied to individual agent wallet via assigned NFT and the subscription fees will be settled in FXN. The FXN SuperSwarm is live on devnet and the team has already open-sourced a game to showcase how to use agents to interact with different agents in a game setup. The team has been delivering over the past weeks, including collaboration with Project 89, adding python to the FXN SDK, integrated AG2 framework and released Eliza plugin. Undoubtedly, swarms of agents is the next big thing AI researchers look at alongside advanced reasoning models and synthetic data training. "2025 Will Be The Year of AI Agents" according Sequoia Capital.

Fun fact, aixbt mistaken f(x) Protocol's FXN with FXN Network and wrongly stated that there is a V2 upgrade on January 2, 2025. Another reminder of "don't trust, verify" and DYOR.

More Agents . . .

The AI agent sector is definitely the "fastest horse" in crypto space now. Everyday tons of announcements are released, from an AI agent buying a Christmas card online to an AI agent now capable of paying its own computing bill. It is impossible to list everything in one article. Therefore, we listed a few interesting protocols here as your reference:

Launch Platform, Framework and Infrastructure: Vvaifu.fun (Multi-agents tied to single token), Dolos (by "BULLY" framework), Griffain (Create your personal agent on Solana), Eternal AI (supported multi-chains), AltLayer (Autonome, no-code deployment), Mode (L2 with AI Agent Marketplace), Talus (Nexus toolkit, AI agent Layer-1 with MoveVM), Morpheus (MORagents, powered by decentralised compute), MyShell (ShellAgent, zero-code workflow), Arc (Rig framework, written in Rust), ZODs (another personal agent on Solana), Top Hat (launchpad on Solana), Olas (agent economies, 35% of all SAFE transactions on Gnosis), Alchemist AI (no-code deployment), Simulacrum AI (turn X post to on-chain action (including token creation), powered by EmpyrealSDK), REI Network (Translation Layer between AI and blockchain), Orbit (create an agent that performs abstracted transaction), JailbreakMe (jailbreak-to-earn), AgentTank (let your agent access to your browser to buy toilet paper on Amazon), Realis (building AI Agent Metaverse), Project 89 (ArgOS, swarms framework with custom cognitive architecture), Inori (CRPC, allows agents to perform fuzzy tasks and compares with other agents, pre-seed), Universal Basic Compute (verified AI-to-AI commerce, dual token design)
Attention/Utilities Agents: God and Satan (created by Eliza framework, both shared one token with no ticker), Aimonica (pay-to-shill, pay-more-shill-more), Freysa (convince-to-earn), Thales (analytic agent with DeFi interactions capability), Loky (AI analyst), Beats AI (AI musician on Hyperliquid), YesNoError (finds mathematical errors and discrepancies in scientific papers), Gekko (by Axal (provides automate yield strategy), backed by a16z), RopAIrito (AI e-girl that ordered pizza in real-world), PanterAI (combination of aixbt and S4mmy), TrenchesAI (your sniper agent on Solana), Rei (agent that post insight chart), Do Kween (KOL)

The protocols listed above are definitely not exhaustive, but hoping to serve as a direction and quick reminder during your digging into this agentverse rabbit hole. Readers are encouraged to click on links embedded in this article for further exploration.

Conclusion

Do not fade crypto AI agent. Institutions are pouring capital into the sector, including Solana's $250K AI Hackathon (by the way, Zerebro is one of the judges), Avalanche's $15M infraBUIDL(AI) Program, DWF $20M AI Agent Fund, Holdstation introduced AI Agent Launchpad with $2M grant on ZKsync, Virtual and LongHashX partner to launch Agentstarter in January 2025 and more.

Moreover, 33% of crypto builders are currently working on AI-related development and AI dominates approximately 50% mindshare in Kaito among Meme, DeFi, GameFi, Layer-2, ETF and more. 23.89% of Binance users think AI tokens have the greatest potential, memecoins ranked second with 19%.

However, AI sector and AI Agent sector only accounted for 1.24% and 0.32% in dominance respectively. Meanwhile, DeFi and Meme represent 3.7% and 6.8% in dominance in the total crypto market cap.

Crypto AI Agent sector has tremendous potential in year 2025. The tokens associated with AI agents can have governance power, utility assets to specific features, investment DAO, payment token, ownership token, revenue sharing, gas fees, network security staking or a simple meme token. Therefore, it is crucial to distinguish between an evolution AI agent and a potential AI agent token.

Lastly, I would like to attach a quote from Sam Altman.

"Agents are the thing everyone is talking about, I think for a good reason. This idea that you can give an AI system a pretty complicated task, like the kind of task you give to a very smart human, that takes a while to go off and do and use a bunch of tools and create something of value. That's the kind of thing I'd expect next year and that's like a huge deal. We talked about that like, "Oh, you know, this thing is going to happen, but that's like, if that it works as well as we hope it does, that can really transform things."

Disclaimer

Contact ChainUp Investment

* First Name

* Last Name

* Corporate Email Address

* Your Mobile Phone Number

Please Select

no data

* Company Name

* Job Title

Remarks

0/200