Interpretation of Y Combinator's Spring Entrepreneurship Guide: What future trends of AI Agents are reflected?

Author: 0xJeff

Compiled by: TechFlow

Y Combinator recently released the "Request for Startups" for spring 2025, listing the directions they want more entrepreneurs to pay attention to. These ideas reflect emerging trends in AI agents in Web2, focusing on solving practical problems and pain points, including:

• AI App Store

• Data Center

• Compliance and Audit Tools

• DocuSign 2.0 (Next Generation Electronic Signature Solution)

• Browser and computer automation tools

• AI Personal Assistant

• Devtools for the agent (Devtools)

• The future of software engineering (engineering agent)

• AI commercial open source software

• Intelligents that optimize code for hardware

• Business-to-Agent

• Vertical field AI agents (agents that focus on specific industries or scenarios)

• Inference AI infrastructure (technical foundation that supports efficient inference and operation of AI models)

These directions are very informative, but if you have already worked hard in this field, you will find that many Web3 agent teams have already made arrangements in these fields.

If you want to dig deeper into these trends, you can check out the original post posted by @ycombinator :

I think the following areas will become the key trends in the development of Web3 AI agents (regardless of ranking):

1. AI commercial open source software

2. Devtools for the agent (Devtools)

3. Vertical field AI intelligent body

4. AI Personal Assistant

5. AI App Store

6. B2A (Business-to-Agent)

1.AI commercial open source software

Web3 AI has a natural connection with open source AI, which makes the open source field an important focus of Web3. Take @ai16zdao as an example, they have driven one of the largest open source AI movements, with the launch of the ElizaOS framework currently receiving 14k stars and 4,227 forks on GitHub. Despite market volatility, the adoption of this framework is still rising steadily.

This open source movement has also inspired Web3 developers to open source their own technologies, pushing teams to develop AI technologies and frameworks so that other developers can collaborate more efficiently. In recent years, we have seen many open source frameworks that go beyond ElizaOS, such as @arcdotfun , @GAME_Virtuals , @sendaifun , @pippinlovesyou and @freysa_ai , which have jointly promoted the development of the open source innovation ecosystem.

With the rapid development of AI agents, such as the o3 launched by OpenAI, the new model released by DeepSeek, and the accelerated launch of related products by technology giants, the demand for open source AI and Web3 AI is heating up. The combination of cryptocurrencies and AI (Crypto x AI) is expected to occupy an important position in the AI ​​market.

2. Devtools for AI Agents

Building AI agents is not only about creating intelligent models, but also requires providing developers with efficient tools and infrastructure to help them transform these agents into practical applications. As AI agents become more complex, developers ' demand for friendly tools, frameworks and platforms is growing rapidly , which can simplify the construction, deployment and management of agents.

In the Web2 era, the popularity of developer tools has significantly improved the capabilities of AI technology. Web3 further promotes this trend, bringing new possibilities to AI development by introducing features such as decentralization, trustlessness and open source collaboration . We are heading towards a new era in which AI agents will no longer rely on the closed ecosystem of a few technology giants .

This trend has spawned many AI -oriented development platforms, agent ecosystems, and No -code/Low-code tools . These tools are designed to lower the threshold for AI agent development and allow more developers to easily participate.

In the Web3 field, more and more platforms are beginning to provide AI agent development toolkits to help developers quickly create and commercialize AI-based applications. Some examples worth paying attention to include:

• @ai16zdao : Launched ElizaOS, with the richest plug-ins and integrated features.

• @sendaifun : Solana Agent Kit, focusing on agent development on the Solana blockchain.

• @CoinbaseDev : CDP Agent Kit, providing basic tools for on-chain AI agent development.

• @autonolas : Launched Pearl, an Agent App Store focusing on practical tools, providing services such as forecasting markets, DeFi automation, and autonomous execution of agents.

• @AlloraNetwork : Provides machine learning infrastructure to help AI agents make more accurate predictions in real time.

• @cookiedotfun : Focus on AI agent-driven data analysis, helping agents extract social emotional information from on-chain and off-chain data.

• @getmasafi : Provides real-time data flow solutions to provide the latest dynamic intelligence for AI agents.

Some codeless AI platforms focused on Web3 include:

• @virtuals_io : A leading codeless/low code AI agent building platform to help developers quickly transform AI agents from concepts to actual products.

• @HoloworldAI : A codeless platform focused on building 3D audio-visual AI agents to help users design virtual characters powered by AI.

• @Cod3xOrg : A codeless tool specially designed for automatic trading agents to help traders automate trading strategies with AI.

• @Almanak__ : A platform specially designed for institutional-level quantitative agent development, supporting the application of advanced financial scenarios.

• @EliteAgents_AI : Focus on plug-in enhanced AI agents, seamlessly integrating with AI ecosystems such as ElizaOS and GAME.

Although the Web3 AI development tool ecosystem is still in its early stages, its infrastructure is rapidly improving. In the next few years, we are expected to see the formation of a fully decentralized AI development ecosystem . In this ecosystem, AI agents will become easier to build, while having complete autonomy, scalability and commercialization capabilities. The development tools that drive this transformation will become an indispensable infrastructure in the Web3 AI economy.

3. Vertical AI Agents

AI agents are gradually evolving from a universal tool for performing simple tasks to highly specialized vertical agents. These agents focus on specific industries or scenarios and are able to handle complex and meticulous tasks . By deepening field knowledge, they can not only complete basic automation, but also act as decision-making agents to perform operations that require deep human expertise.

Today, the wave of AI-driven verticalization is gradually emerging. In the fields of finance, law, scientific research, etc., agents have the ability to analyze, recommend, and even perform operations on behalf of users. This verticalization trend will further enhance the influence and application depth of AI agents in various industries.

Some typical vertical field AI agents examples include:

• Tax agent : Helps users calculate, optimize and execute tax-saving plans.

• Legal agent : able to review contracts and optimize terms, and even participate in legal disputes on behalf of users.

• Financial agent : analyze financial statements, interpret macroeconomic trends, and provide investment advice.

What makes Web3 unique to vertical AI agents is its emphasis on autonomy, decentralization and on-chain integration . Traditional AI services often rely on centralized data silos, while Web3 native AI agents achieve greater transparency and trust through on-chain Verifiability. This feature makes Web3 agents more advantageous in data processing and results credibility.

In the cryptocurrency space, the interaction and personalization of the community are particularly important , so Web3 AI agents are moving towards a more personalized and interactive direction. Unlike the usually cold and functional AI agents in Web2, which are usually cold and functional, Web3 agents have gradually formed unique personalities and interaction patterns to adapt to the culture of decentralized communities. For example:

• AI influencers : For example, @aixbt_agent , share unique insights and market cutting-edge information on Crypto Twitter to attract community attention.

• Token analysis agents : such as @unit00x0 , @kwantxbt , @tri_sigma_ , @mobyagent and @_AgentScarlett , focusing on analyzing token data and providing relevant suggestions.

• Research agents : such as @DV_Memetics and @S4mmyEth , providing actionable market intelligence through @orbitcryptoai .

• DeFAI agent : Focusing on managing liquidity mining (LP'ing), Yield Farming and trading strategies, developed by teams such as @Cod3xOrg , @gizatechxyz and @autonolas .

In addition, AI model platforms such as @NousResearch , @BagelOpenAI and @PondGNN are further enhancing the personalization capabilities of agents to make them more in line with the needs of the decentralized community. As DeFAI agents gradually simplify the complex operations of DeFi, they could be a key driver in attracting billions of new users into the blockchain world. These agents provide users with a more intuitive experience by lowering the threshold for use of DeFi, and are expected to set off a new wave of AI adoption in the future.

4. AI Personal Assistant

AI personal assistants are revolutionizing the way we handle daily tasks, making many features that were unimaginable in the past come true by providing convenience and automation. These assistants will no longer be limited to reminders and schedules, but will be able to make proactive decisions to help users manage time and resources more efficiently.

Imagine an AI that can book a trip for you while recommending restaurants to your preferences, checking traffic conditions, and automatically adjusting meeting schedules when you are late. It can also summarize the content of the meeting, make follow-up suggestions, and even automatically book transportation. In addition, it can organize your photos, sort them by location and event, and generate exquisite memory albums for easy access at any time.

With the support of Web3, these features will be further expanded:

• Airdrop Agents : Help users scan all wallets and automatically detect whether they meet the airdrop conditions of encrypted projects (such as @berachain , @monad_xyz , @StoryProtocol ) .

• Yield Farming & LP Management Agents : Track and optimize DeFi positions in real time, automatically receive rewards and compound returns into the best strategy.

• GitHub repository analysis agents : such as @soleng_agent , can evaluate the strength of the project development team and help users identify potential scams.

• Automatic Trading Agents : such as @Cod3xOrg and @Almanak__ , execute transactions according to preset conditions, optimize the timing of entry and exit positions to maximize market returns.

The next generation of AI personal assistants will no longer be passive assistants, but a "co-pilot" who can take the initiative. As AI models continue to improve their inference and decision-making capabilities, these agents will change from responsive to predictive, capable of completing complex multi-step tasks with minimal user input.

Web3 plays a key role in this transformation. Decentralized AI agents are trustworthy, transparent and censor-resistant, ensuring users have complete control over AI-driven workflows. This capability will allow users to hand over complex financial and operational decisions to AI, revolutionizing the way we work.

5.AI App Store

The AI ​​app store is one of the most anticipated developments in the field of artificial intelligence. Just as the mobile app store has changed the way software distribution is distributed, AI agents also need an exclusive market that allows users to easily discover, purchase and integrate AI-powered applications.

In Web3, this concept is evolving into a combination of multi - Agent Orchestration Network (MAO) and Agent Distribution Network :

• Agent Distribution Network : Attract developers, investors and users to join the ecosystem. For example, @virtuals_io is building an Agent Society that allows different AI agents to coexist and collaborate with each other.

• MAO Network : Through intelligent matching technology, appropriate AI applications are recommended to users and multiple agents work efficiently. Users do not need to search manually, they just need to express their needs, and the system can instantly combine solutions that meet their needs.

Therefore, Web3's AI application store is not just a trading market, it also needs to have functions such as planning, auditing and privacy protection, while supporting seamless interactions between agents. This model will completely change the way users interact with AI and lay the foundation for the future AI ecosystem.

Key roles in driving this field:

• @virtuals_io : Committed to expand its "Agent Society" blueprint, attract high-quality agent teams to join, and take the lead in developing interagent communication protocols to lay the foundation for agent collaboration.

• @santavirtuals and @questflow : Optimize resource allocation efficiency by improving coordination capabilities between Virtuals agents.

• Abstraction Layers projects such as @orbitcryptoai and @HeyAnanai : Lower the threshold for use by integrating AI agents and decentralized finance (DeFi) into an efficient abstraction layer, allowing more users to easily access these technology.

Although AI orchestration is still in its early stages, it is foreseeable that seamless and profitable AI agents will open up a huge market, and Web3 is making an active layout to occupy an important position in this market.

6.B2A (Business-to-Agent)

AI agents are now more than just tools, they are becoming active players in the digital economy, able to complete transactions, manage resources independently, and even collaborate with other agents. This trend has spawned a new infrastructure demand, and B2A (Business-to-Agent) emerged to provide services to AI agents.

Just as SaaS (Software-as-a-Service) has changed the way businesses operate, B2A will redefine how AI agents interact, trade and operate in the digital economy. In the future, AI agents will need exclusive payment solutions, data access, computing power, and privacy protection frameworks. Currently, multiple Web3 projects are driving this transformation:

• AI -commerce Payments : @Nevermined_io is developing payment solutions for agents with the goal of becoming "PayPal for AI agents".

• Compute Management : A self-sustaining agent developed by @hyperbolic_labs can efficiently manage its own computing resources.

• Privacy & Security Infrastructure : @PhalaNetwork , @OraProtocol and @brevis_zk are building privacy-protecting computing layers to provide AI agents with a secure and verifiable interactive environment.

• Quality Data Access : @getgrass_io , @vana , @getmasafi and @cookiedotfun provide structured high-quality data sources to help AI agents train, learn and operate efficiently.

• Agent -to-Agent Communication : @virtuals_io is developing interagent communication protocols to enable AI agents to collaborate efficiently.

• AI Intellectual Property for AI : @StoryProtocol is developing a framework similar to TCP/IP to manage intellectual property rights for AI-generated content, allowing agents to independently manage and authorize their content.

B2A is more than just a theoretical concept—it is becoming a reality. As AI agents continue to increase their functionality and complexity, they need dedicated infrastructure to support their independent operation in the economic ecosystem. If you haven't started thinking about how to serve the AI ​​agent market, you may have missed the opportunity.

Summary and thoughts

AI agents are redefining how we interact, build and automate in Web2 and Web3. With the rise of the Web3 native AI ecosystem, they bring new models, including open source collaboration, agent-driven business models, and decentralized automation solutions.

Although the integration of AI and encryption technology is still in its early stages, its development momentum is already unstoppable. Web3 provides AI agents with key capabilities that cannot be achieved by Web2: such as asset ownership, a license-free innovation environment, and a highly composable ecosystem. These characteristics create infinite possibilities for an agent-driven economy. The question is no longer whether AI agents will change Web3, but how quickly this change will come and which industries will be at the heart of this change.

As the scale of the agent-driven economy continues to expand, whether you are a developer, investor, or a curious observer, it is the best time to focus on this field. Infrastructure is being built quickly, key players are being formed, and opportunities have emerged.

So, the question is: Are you ready to join this wave of change?