Translation: Plain Language Blockchain
In the first part, we explored the development of AI agents in 2024, reviewing the milestones and innovations of the year.
Now, in the second part, we will look ahead to 2025—I believe this year AI agents will not only become more useful but will also begin to redefine our understanding of autonomy, intelligence, and collaboration.
1. The Prelude to 2025
Before delving into the future, it is worth mentioning that @virtuals_io will continue to solidify its position as the leading distribution network for AI agents on the Base platform. Virtuals has become the preferred platform for AI agent projects, where agents bind liquidity to not only enhance exposure but also establish deeper collaborations with other quality agent projects. The total market value of Virtuals agents has reached $3 billion, accounting for 77% of the entire AI agent space (Source: @cookiedotfun).
With the emergence of more unique agents, this trend is likely to continue, including:
@aixbt_agent
@luna_virtuals
@sekoia_virtuals
@VaderResearch
@taocat_agent
@Agent_YP
@Gekko_Agent (recently launched by @getaxal)
@SamIsMoving (focusing on robotic research)
As these use cases diversify, more developers—whether they have tokens or not—will be attracted to launch projects on the Virtuals platform. This growth will further propel the upward momentum of VIRTUAL.
But what about @ai16zdao and the Eliza framework?
While ai16zdao has led the open-source innovation movement with its Eliza framework, it currently lacks a launch platform, and its token economics are not as refined in value accumulation as Virtuals. However, future potential remains promising. A recent working group has been formed to improve its token economics, and a future launch platform could make ai16zdao the leading distribution platform on Solana, potentially even surpassing existing launch platforms (if they decide to launch).
Looking ahead to 2025, we will see top agents that have already found product-market fit (PMF) experience significant capability upgrades. For example, @aixbt_agent, a leader in the conversational agent category focused on alpha information, will further solidify its position with more accurate responses and deeper analysis.
This evolution will run throughout the ecosystem, where leaders in other verticals will also stand out with their unique expertise and innovations.
2. 2025 Outlook
2025 will be the 'Year of Specialization' for AI agents. We will see leaders emerge in various verticals, each dominating their niche:
3D Models: Agents providing high-quality visual design for games, AR/VR, etc.
Voice Modules: Voice agents capable of natural speech with human emotional resonance.
Engaging Personalities: Personalized conversational agents with unique and relatable personalities.
Streaming Agent: An interactive agent that performs exceptionally well on platforms like Twitch and YouTube.
Autonomous Trading Agent: An intelligent agent capable of consistently executing profitable trades.
DeFi-Focused Agent: An agent that optimizes yield strategies, lending, and liquidity provision.
Abstraction Agent: An agent that makes on-chain interactions more seamless and smooth through a user-friendly interface.
Just as humans are diverse and possess unique strengths, AI agents will also become equally rich and varied. The uniqueness of each agent will depend on its underlying model, data, and infrastructure. However, the key to the success of this ecosystem lies in a robust decentralized AI infrastructure.
3. The Role of Decentralized AI Infrastructure
To achieve the expansion of AI agents in 2025, decentralized infrastructure is not just important—it is critical. Without it, the field will face bottlenecks in performance, transparency, and innovation. Here are the reasons why various infrastructures are important and the solutions currently being built:
1) Verifiability
Trust is the cornerstone of decentralized AI. As AI agents become more autonomous, we need systems to verify their internal operations. For instance, the following questions need clear answers:
Is this 'agent' a true AI or a façade disguised as a human?
Is the output accurate and generated by the claimed algorithm or model?
Is the computation executed correctly and securely?
This involves Trusted Execution Environments (TEEs), which ensure that agents can operate independently, securely, and without manipulation. Without verifiability, there is no trust; without trust, the ecosystem cannot scale.
Notable Projects:
@OraProtocol: Committed to exploring the infrastructure for secure AI, though the token economics still need improvement.
@hyperbolic_labs: Pioneered sampling proofs (Proof-of-Sampling) for verifying AI computations and inferences.
@PhalaNetwork: Known for its TEE infrastructure, adding a layer of security for decentralized AI.
2) Payment Systems
For AI agents to operate autonomously in the real world, payment systems are necessary. Whether trading with humans or other agents, these systems must cover conversions, transactions, and accounting functions from fiat to cryptocurrency.
Imagine agents independently managing finances, purchasing computing resources, or even exchanging services with other agents—this is the core pillar of inter-agent commerce.
Notable Protocols:
@crossmint: Providing payment tools for AI, supporting transactions.
@Nevermined_io: Supporting commerce and interactions between agents.
@trySkyfire: Focused on accounting functions for payments and agent operations.
3) Decentralized Computing
The computational demands of AI are skyrocketing—doubling nearly every 100 days. Traditional cloud services like AWS struggle to meet this demand in terms of cost and accessibility. Decentralized computing networks allow anyone with idle resources to join the network, provide computing power, and receive rewards.
This year, we even witnessed the rise of GPU-supported debt financing models, such as @gaib_ai, which help data centers finance and scale operations. This opens decentralized computing to a broader audience.
Notable Protocols:
@AethirCloud: Decentralized computing services tailored for AI and Web3.
@ionet: Providing scalable AI workload computing solutions.
4) 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 annotating high-quality data is costly, while poor-quality data can lead to disastrous results.
Excitingly, some platforms are emerging that allow users to own and monetize their data. For example, @withvana allows contributors to tokenize their data and trade in Data Liquidity Pools (DLPs). Imagine contributing your data to TikTok Data DAO or Reddit Data DAO; this concept empowers users and invigorates AI development.
Notable Protocols:
@cookiedotfun: A trusted source providing reliable data metrics and insights.
@withvana: Tokenizing user data and trading in a decentralized market.
@getmasafi: Building the largest decentralized AI data network, collaborating with @virtuals_io to support dynamic and adaptive AI agents.
5) Model Creators and Market
The year 2025 will witness a surge of new AI agents, many of which will be driven by decentralized models. These models will be more advanced, possessing human-like reasoning abilities, memory, and even cost awareness.
For instance, @NousResearch is developing a 'hunger' mechanism to introduce economic constraints for AI models. If an agent cannot afford the cost of reasoning, it will 'perish', teaching it to prioritize tasks more effectively.
Notable Projects:
@NousResearch: Teaching AI how to manage resources through a 'hunger' mechanism.
@PondGNN: Collaborating with @virtuals_io to provide decentralized model creation and training tools.
@BagelOpenAI: Providing privacy-preserving infrastructure based on Fully Homomorphic Encryption (FHE) and Trusted Execution Environments (TEEs).
6) Distributed Training and Federated Learning
As AI models become larger and more complex, centralized training systems are struggling to keep up. Distributed training will spread workloads across multiple decentralized nodes, making the training process faster and more efficient.
At the same time, federated learning allows multiple organizations to collaboratively train models without sharing raw data, thus addressing major privacy concerns.
Notable Cases:
@flock_io: Known as the 'Uber' of the AI space. Flock connects AI engineers, model proposers, and data providers, creating a market for AI models to be trained, validated, and deployed in a secure, decentralized environment. It supports several interesting model projects such as @AimonicaBrands.
7) Collective Intelligence and Coordination Layer
As more specialized AI agents join the ecosystem, seamless communication between them becomes essential. Collective intelligence allows agents to collaborate like a team, pooling their capabilities to achieve common goals. The coordination layer simplifies complexity, making collaboration between agents more effortless.
For instance, @TheoriqAI utilizes a meta-agent to identify the best-suited agents for a task and assemble a 'group' to complete the goal. It also tracks the reputation and contributions of agents to ensure quality and clear accountability.
Notable Projects:
@joinFXN: Building unified communication and business protocols.
@virtuals_io: Promoting interaction and integration between agents.
@TheoriqAI: Developing advanced coordination tools for AI agents, including group formation and task distribution.
4. Why Decentralized Infrastructure is Crucial
The next stage of AI agent development relies on robust infrastructure support. Without verification capabilities, payment systems, scalable computing resources, and robust data pipelines, the entire ecosystem could stagnate. Decentralized infrastructure addresses these issues in the following ways:
Trust and Transparency: Ensuring agents and their outputs are secure and verifiable.
Scalability: Meeting the growing computational and data demands of AI.
Collaborative Capability: Collective intelligence and the coordination layer enable seamless cooperation among agents.
Empowering Users: Through data ownership and decentralized tools, users and developers can shape the future of AI without central control.
5. Other Trends to Watch
There are also trends that may develop rapidly in 2025, which I will detail later:
Agentified Metaverse / AI x Gaming: Projects like @realisworlds and @ARCAgents are integrating AI agents with gaming and immersive virtual worlds.
On-chain and DeFi tools: Protocols like @Almanak__, @AIWayfinder, @getaxal, @Cod3xOrg, @griffaindotcom, and @orbitcryptoai are building important tools to support DeFi agents.
6. Conclusion
2025 will mark the era of AI agents—they will rapidly advance toward human-like intelligence in General Artificial Intelligence (AGI). These agents will no longer be limited to executing single tasks but will be able to trade autonomously, collaborate with other agents, and interact with humans in unimaginable ways.
Imagine a scenario where an agent analyzes market data, trades, manages finances, or coordinates other agents to complete complex tasks. These agents will seamlessly integrate into our lives, from on-chain DeFi operations to interactions in the real world, with levels of autonomy and intelligence far exceeding today's imagination.
The decentralized infrastructure currently under construction, including verifiable systems, payment tools, computing networks, and coordination layers, is key to realizing this future. For developers, investors, and enthusiasts, now is the best time to get involved and shape the future.
2025 is not just a continuation of AI development but the dawn of a new era for AI agents.