Original text from 0x Jeff
Compiled by Odaily Planet Daily Golem (@web3_golem)
2024 AI Agent development review
2024 will be a transformative year for AI Agents. About three months ago, Truth Terminal captured everyone's attention with its humorous personality, conversational style, and interactions with A16z co-founder Marc Andreessen, becoming the first 'millionaire agent' and sparking the trend of AI Agents.
Soon after, Virtuals also entered this field, pioneering 'Agent Tokenization' and solidifying this narrative. Since then, innovation has exploded:
Luna: This agent launched an on-chain wallet tipping feature for fans, and can now browse Twitter, analyze posts, and even join Google Meet.
Dialogue agents on Twitter/X: Some agents have become 'meme masters,' while others focus on acquiring and sharing Alpha information. For example:
AIXBT: Known for its concise and actionable insights and a bit of light-hearted humor;
Dolos: Sharp personality, now developed its own framework to support other agents through Dolion.
At the same time, AI Agents are gradually becoming more entertaining, equipped with 3D models, voice capabilities, and the ability to exist across platforms. Representative agents include:
AVA and Holoworld AI: The first 3D audio-video framework, granting agents 3D bodies, voices, and deeper personalities;
Zerebro: A music agent that released a top album and is about to launch its own framework ZerePy, allowing more people to build agents like Zerebro;
Nebula: A meme AI KOL that can create meme images and videos and appears in AR/VR environments and games;
LucyAI: The first reality anime agent that can speak multiple languages, live stream, and interact with fans;
DO KWEEN: A movie agent that produces Netflix-quality drama episodes weekly.
2024 AI Agent narratives
Meanwhile, ai16z and the open-source innovation movement have also gained attention, especially after the launch of the Eliza framework. Developers have come together to create toolkits, plugins, and other features that promote collaboration and innovation. During this period, Virtuals has grown into a unicorn company, further solidifying its leadership position in the AI Agent distribution platform.
The open-source innovation movement has sparked interest in the developer community, opening the door to the largest community collaboration of the year. More and more projects are emphasizing the importance of 'open-source frameworks.' As agents continue to evolve, new narratives promoting more agent collaboration have emerged:
Agent Metaverse: Proposed by Realis, it created a Minecraft-style replica of Earth to house these AI agents, allowing them to interact and build a civilization.
Gamification of agents: ARC Agents are a representative in this field, featuring reinforcement learning AI x gaming. By combining AI with gaming reinforcement learning, they launched a game similar to Flappy Bird, pitting agents against each other while community-contributed game data helps these agents grow. ARC recently revealed its vision towards AGI.
Collective/Swarm Intelligence: FXN is a representative in this field, aiming to establish a unified economy for AI agents, with the idea of collusion allowing AI agents to work together towards common goals. Virtuals is also advancing interactions between agents (or commercialization), a communication protocol that enables agents to seamlessly provide services to one another. Meanwhile, Story also announced the launch of an IP-based inter-agent communication protocol, allowing agents to tokenize, monetize, and trade IP.
In parallel with these narratives, we can also see:
On-chain trading agents: Initially proposed by Spectral, their Syntax v2 allows users to launch trading agents capable of trading on Hyperliquid. They have maintained dominance in the field, but progress has been temporarily stalled due to a small bug. Another noteworthy agent is Big Tony, which automatically trades mainstream currencies using Allora's machine learning price prediction model.
InvestmentDAO: Initially represented by ai16z, now more DAOs are beginning to emerge, such as AIrthur Hayes and Aimonica. The general narrative is that these DAOs raise SOL on daos.fun (or other platforms) and use these funds for trading and investment profits. If you can use the names of Crypto VCs or well-known figures, the created InvestmentDAO is more appealing.
DeFi Agent: Led by Mode, which is the preferred ecosystem for DeFi agents. Major application scenarios include AI-driven stablecoin mining, providing liquidity, lending, etc. High-quality teams within the ecosystem, including Giza, Olas, Brian, Sturdy, QuillAI Network, are involved in the construction.
AI App Store: ALCHEMIST AI is a representative in this field, providing a no-code tool that allows users to create applications. MyShell is another AI App platform with a larger developer and user base, particularly in the Web2 field.
Abstraction layer: Griffain and Orbit are representatives in this field, providing a chain-based abstract experience for everything on-chain, facilitating user operations on-chain, especially friendly for ordinary users.
On-chain VC agents: Sekoia Virtuals aims to become a top 'rubber stamp' for quality agent projects, currently strictly filtering investments to only three projects, setting a precedent for on-chain VC.
Other narratives: Such as Freysa's on-chain puzzles, JailbreakMe's agent hacking bounty rewards, H4CK Terminal's white-hat AI, and the unique agent models of god and s8n representing a debate between God and Satan. Even more interesting are agents focused on Alpha analysis, such as Rei (quantitative analyst), kwantxbt (TA analyst), and Nikita (general alpha analyst). Then there's Fartcoin, a suddenly popular meme project that even made it onto Stephen Colbert's show and broke the $1 billion market cap, as AI memes are being accepted by the public.
Development of data and frameworks
Cookie DAO is becoming the primary source of data and social metrics for AI agents, relied upon by industry insiders to track agents' influence, market value, and performance;
Masa integrates with Virtuals to provide agents with real-time data, enabling self-learning and self-improvement;
TAOCAT is the first virtual agent powered by the Bittensor subnet, showcasing the potential of real-time data (it is the only agent token that surged despite other agent tokens declining);
AgentTank showcases a framework that brings agents to computers, giving them full computable capability so they can engage in entertaining interactions on the internet and provide interesting commentary.
Other new frameworks:
Arc: A Rust-based RIG framework that has gained attention for its versatility;
Dolion: Evolved from Dolos, becoming a toolkit for creating unique agents.
What have we learned from 2024?
The above may have overlooked some minor narratives or AI Agents, but through the development of AI Agents, we can learn the following points from this year:
Top teams valued at over $50 million all have their own tuned models.
They initially showcase the application scenarios of their agents and the uniqueness of their models, then provide a no-code framework for others to have agents as good as their flagship agent, which also brings higher agent value and boosts the token price of the agents.
But this does not mean I suggest you build your own framework or not build on top of other frameworks like Virtuals G.A.M.E and ai16z Eliza. If you do not have enough AI resources or capabilities, you should join these communities because, with the right tools, you can quickly realize your ideas and experiment. At the same time, you should leverage Virtual and ai16z for distribution/marketing, as these two places currently provide the best visibility, and integrating and collaborating with them is definitely positive EV.
Investing in agents with built-in frameworks or the entire AI agent ecosystem will lead to better risk-return ratios.
If they manage to create a framework that people are willing to pay to build agents, it means that the framework has enough attention and demand to drive or maintain its price. Arc is a great example; the first Rust framework quickly became popular, and its price followed suit.
On-chain and DeFi applications will be the product-market fit (PMF) for crypto AI.
I believe the areas currently providing the most value include:
The abstraction layer helps people navigate on-chain;
Alpha agents share quality alpha that people can profit from;
Execution agents can help simplify trading, mining, providing liquidity, and executing loans.
Perhaps we will soon see an agent that combines alpha discovery + execution.
Data is an indispensable part of every agent
Bad data = bad output. If data is gold, then data platforms like Cookie DAO are essentially a gold mine. Vana is an interesting L1 that tokenizes data into liquidity pools, and they have a DataDAO model that helps people jointly own data, introduce data, and clean this data for AI Agents. Although the tokenomics may have issues, the product is very interesting.
Looking forward to the development of AI Agents in 2025
In the above, we explored the development of AI Agents in 2024, reviewing milestones and innovations from the year. Now, we will look forward to 2025—I believe that this year AI Agents will not only become more useful but will also begin to reshape our perceptions of autonomy, intelligence, and collaboration.
Laying the groundwork for 2025
Before moving to the next step, it is worth emphasizing that Virtuals will continue to solidify its position as the primary distribution network for AI Agents on Base. Virtuals has become the preferred platform for agents to match their liquidity, increase visibility, and form deeper collaborations with other quality projects. Currently, the total market cap of Virtuals agents is about 3 billion dollars, accounting for 77% of the entire AI Agent field (data source: Cookie DAO).
As more unique agents appear on Virtuals and these application scenarios become increasingly diverse, more developers will be attracted to the Virtuals platform, regardless of whether they already have tokens. This growth will also drive an increase in the VIRTUAL token.
While ai16z Dao has led the open-source innovation movement with its Eliza framework, it currently lacks a launch platform, and its tokenomics' value accumulation level is not as high as Virtuals. Nevertheless, the future remains promising. Recently, they have formed a working group to improve their tokenomics, and a future launch platform could position ai16z as the top distribution platform on Solana, surpassing existing launch platforms (if they decide to launch one).
In 2025, we will also see top agents with product-market fit (PMF) receive significant capability upgrades. For example, AIXBT has already established a leadership position in the dialogue agent field focused on Alpha and may further solidify its position through sharper responses and more insightful analyses.
As leaders in other verticals emerge, this evolution will be reflected throughout the ecosystem, as they lead the way with unique expertise and innovation.
What are the trends for 2025?
2025 will be a year of specialization for AI Agents. We will witness the emergence of leaders in various verticals, each dominating its own niche market:
3D models: Agents with high-quality visual designs suitable for games, AR/VR, etc.
Voice modules: Agents capable of speaking naturally and resonating emotionally like humans.
Personality-rich agents: Personalized dialogue agents with unique, relatable personalities.
Live streaming agents: Interactive agents thriving on platforms like Twitter/X and YouTube.
Automated Trading Agent: Capable of continuously executing profitable trades.
DeFi-focused Agent: Optimizing yield strategies, lending, and liquidity allocation.
Abstract Agent: Seamless on-chain interaction through a user-friendly UI.
Just as humans are diverse and specialized, AI Agents will also become equally diverse. The uniqueness of each agent will be closely related to its underlying model, data, and infrastructure. However, the success of this ecosystem depends on a robust decentralized AI infrastructure.
The role of decentralized AI infrastructure
To expand AI agents in 2025, decentralized infrastructure is crucial; without it, the field will face bottleneck risks in performance, transparency, and innovation.
Here are the reasons why each part of decentralized AI infrastructure is important, and the projects currently being built to address these challenges:
Verifiability
Trust is the foundation of decentralized AI. As AI Agents become more autonomous, we need systems that allow us to verify what happens in the background. We need to know whether this 'agent' is genuine AI or a human impersonator; whether the outputs are accurate and generated by the claimed algorithms or models; whether computations are executed correctly and securely, and so on.
This also involves Trusted Execution Environments (TEE), which ensure agents operate independently, securely, and without manipulation. Without verifiability, there is no trust, and without trust, the ecosystem cannot scale.
Projects to watch:
ORA: Exploring the infrastructure for secure AI, but the tokenomics still needs refinement;
Hyperbolic: Pioneering sampling proofs for verifying AI computations and reasoning;
Phala Network: Known for its TEE infrastructure, adding a layer of security for decentralized AI.
Payments
For AI agents to operate autonomously in the real world, payment systems are required. Whether transacting with humans or other agents, these systems must handle various aspects from on-chain/off-chain to barter and accounting. Imagine agents managing finances independently, purchasing computing resources, or even exchanging services with other agents—this is the foundation of business transactions between agents.
Notable protocols:
Crossmint: An AI payment tool that facilitates transactions;
Nevermined: Supports commercial transactions and interactions between agents;
Skyfire: Focused on payment and accounting for agent operations.
Decentralized computing
The computational demands of AI are soaring—doubling approximately every 100 days. Traditional cloud services like AWS cannot meet this demand in terms of cost or accessibility. Decentralized computing networks allow anyone with idle resources to join the network, offer their computing power, and earn rewards.
This year, we have even seen the emergence of GPU-supported debt financing models like GAIB, helping data centers finance and expand their operations. This enables decentralized computing to be utilized by a broader audience.
Notable protocols:
Aethir: Decentralized computing tailored for AI and Web3.
io.net: Scalable computing solutions for AI workloads.
Data
If AI is the brain, then data is the oxygen. The quality, reliability, and integrity of data directly impact the performance of AI models. However, acquiring and labeling high-quality data is costly, and bad data leads to bad outcomes.
Excitingly, some platforms have emerged allowing users to own their data and monetize it. For example, Vana allows contributors to tokenize their data and trade in data liquidity pools (DLP). Imagine choosing TikTok DataDAO or Reddit DataDAO to aggregate your contributions—this concept empowers users while driving AI development.
Notable protocols:
Cookie DAO: A reliable source for data metrics and insights;
Vana: Tokenizing user data into liquidity pools that can be traded on decentralized markets;
Masa: Collaborating with Virtuals to build the largest decentralized AI data network, supporting dynamic and adaptive AI agents.
Model creators and markets
2025 will witness an explosive growth of new AI Agents, many of which will be supported by decentralized models. These models will be more advanced, integrating human-like reasoning, memory, and even cost-awareness.
For example, Nous Research is exploring a 'hunger' mechanism that introduces economic constraints for AI models. If an agent cannot afford the reasoning costs, it effectively 'dies,' teaching it to prioritize tasks more efficiently.
Notable projects:
Nous Research: Introducing a 'hunger' mechanism to teach AI resource management;
Pond: Collaborating with Virtuals to provide tools for decentralized model creation and training;
Bagel: Providing privacy-preserving infrastructure using FHE and TEE.
Distributed training and federated learning
As AI models become larger and more complex, centralized training systems will not be able to meet the demand. Distributed training spreads the workload across multiple decentralized nodes, making the process faster and more efficient. Meanwhile, federated learning allows organizations to collaboratively train models without sharing raw data, addressing major privacy concerns.
FLock.io is the 'Uber of AI.' Flock connects AI engineers, model creators, and data providers, creating a marketplace to safely and decentralizedly train, validate, and deploy AI models. It supports projects like Aimonica and many more interesting models.
Collective intelligence and coordination layer
As more specialized agents enter the ecosystem, seamless communication between them becomes crucial. Collective intelligence allows agents to work together as a team, pooling their capabilities to achieve common goals. The coordination layer abstracts complexity, making collaboration among agents easier.
For example, Theoriq uses meta-agents to identify the most suitable agents for a task and form a 'swarm' to achieve goals. It also tracks reputation and contributions, ensuring quality and accountability.
Notable projects:
FXN: Creating protocols for unified communication and commerce;
Virtuals: Enabling interactions and integrations between agents;
Theoriq: Developing agents and building advanced coordination tools for AI agents, including clustering and task delegation.
Why decentralized infrastructure is critical
The next phase of AI agent development depends on infrastructure. Without verifiability, payment systems, scalable computing, and robust data pipelines, the entire ecosystem risks stagnation. Decentralized infrastructure helps address these issues by providing trust and transparency, scalability, collaboration, and empowerment.
Of course, there are several other narratives expected to develop in 2025, such as:
Agent Metaverse / AI x Gaming: Projects such as Realis and ARC Agents are merging agents with games and immersive virtual worlds;
On-chain and DeFi tools: Protocols such as Almanak, Wayfinder, Axal, Cod3x, Griffain, and Orbit are building essential tools for DeFi-driven agents.
Summary
2025 will be the year of AI Agents, where we will see them rapidly moving towards perceptive AGI. These agents will not only perform isolated tasks—they will autonomously trade, collaborate with other agents, and interact with humans in ways we cannot imagine.
Imagine an agent analyzing market data, trading, managing your finances, or coordinating complex tasks with others. They will seamlessly integrate into our lives, handling everything from on-chain DeFi operations to real-world interactions, with autonomy and intelligence far beyond what we see today.
Decentralized infrastructure currently being built (verifiable systems, payment tools, computing networks, and coordination layers) will make this future possible. For builders, investors, and enthusiasts, now is the time to delve into research and shape the future.
2025 is not just a continuation, but a dawn for AI Agents entering the next era.
