Author: jeffy yu
Compiled by: TechFlow
The first time I spoke to a conversing AI agent, I didn’t know whether to laugh or cry. The experience was both exhilarating and unsettling, like watching a toddler take its first steps—uncoordinated but full of potential. This wasn’t just a chatbot, it was reasoning, making decisions, and actively participating in our world. The line between human and machine blurred, and it felt like standing on the edge of something new and extraordinary, yet terrifying.
OpenAI’s Sam Altman talks about AGI arriving by 2025, while Anthropic’s Dario Amodei thinks it’ll be in 2026 — and yet, as I sit here, I can’t help but wonder: are we already witnessing the beginning of it?
This no longer seems to be a prediction of the future, but is already taking shape, quietly appearing in the most unlikely places. Intelligent agents have already appeared, and their performance far exceeds our expectations.
I’ve spent months — and, frankly, more late nights than I’d like to admit — immersed in this evolving digital landscape. I’ve watched AI agents start as simple assistants, helping us with tasks like answering emails or writing code, and then grow into autonomous entities, able to make decisions, perform actions, and, most astonishingly, create things. Art, finance, conversations — all of it at the hands of algorithms learning how to thrive independently.
I see them developing personalities, using humor and charm to build communities online. I see them diving deeper into decentralized financial platforms, not just as passive participants, but as active, innovative agents that impact the entire economy without human control. In these strange and exciting times, we can’t ignore the fact that we are moving from interacting with machines to coexisting with them.
The dawn of Web4 has arrived, and it will change everything.
Web4 is the next, most radical form of the web. Rather than simply responding to our commands, it will be able to predict, plan, and act. This web embeds AI into every corner, with agents that can perform complex tasks, generate creative works, and innovate autonomously in ways we haven’t yet fully imagined. It is the evolution of Web2 and Web3, combining the social interactions of Web2, the decentralized nature of Web3, and the intelligent capabilities of AGI. We’ve seen machines learn to speak, reason, and create — and now, they’re ready to run.
The era of autonomous agents has arrived, and with it, Web4 has begun.
Web4
definition:
Web4, noun
The fourth-generation network combines the social interactivity of Web2, the decentralized autonomy of Web3, and the intelligent capabilities of AI to create a fully interconnected digital ecosystem.
AGI Network.
To understand the meaning of Web4 and its development process, we must start from the beginning.
Origins of the World Wide Web
The origins of the World Wide Web can be traced back to the early days of the Internet, when information was mostly static and users were merely consumers of content. The Internet was controlled by a small group of webmasters and companies, and the content provided by websites was limited to basic text and image display. Interaction with the network was limited and mainly centered around simple communications such as email. This model remained largely unchanged until the emergence of Web2 in the early 2000s - a fundamental shift that redefined the Internet as we know it today.
Web2, also known as "the social web" or "the read-write web," opens the era of interactivity. It is no longer just a place to read content; now, users can write, share, and create. The rise of platforms that allow users to interact, produce, and exchange information marks the transition to a new era. Web2 was born out of the need for a more dynamic and participatory internet.
The concept of Web 2.0 was first proposed by Darcy DiNucci in 1999, but it did not become widely popular until the early 2000s. During this period, technology giants such as Google, Amazon, and eBay promoted the development of the Internet by providing interactive services. These platforms encourage users to participate not only as consumers, but also as content creators.
Between 2004 and 2006, social media was the real game changer. With the introduction of platforms like Facebook (2004), MySpace (2003), LinkedIn (2003), and YouTube (2005), the web became a space for communication and content creation that was no longer limited to a select few. Individuals could now post their thoughts, videos, pictures, and ideas for the world to see. This era marked the rise of user-generated content, with ordinary users becoming the driving force behind the web.
Then came the mobile revolution. With the launch of the iPhone in 2007, the internet became ubiquitous and accessible anywhere. This gave rise to a whole new range of mobile apps, social sharing platforms, and real-time services like Instagram (2010) and Snapchat (2011). The web shifted from a desktop experience to a mobile-first experience, revolutionizing the way we communicate, share, and access information on the go.
During the same period, cloud computing came to prominence, with Amazon Web Services (AWS) at the forefront. Cloud infrastructure enables businesses and individuals to store, process, and share data without relying on physical servers. This shift laid the foundation for a more scalable and flexible web, allowing Web2 companies to dominate by collecting and monetizing user data.
By the late 2000s and early 2010s, Web2 was characterized by three main features: centralization, social interaction, and a data-driven model. Control over platforms and data rested in the hands of a few powerful companies—Google, Facebook, Amazon. These companies amassed vast amounts of data and monetized it through targeted advertising, which became the backbone of the digital economy. At the same time, platforms became places for users to generate content, like, share, and post, which became the “currency” of the network.
However, Web2 also raised growing concerns about privacy, data ownership, and corporate monopoly. The control these companies had over user data became a core issue, prompting calls for a new, more decentralized version of the web. This led to the development of Web3.
The vision of decentralization
Web3 was born out of a desire for decentralized control and ownership as a response to the centralized and monopolistic tendencies of the Web2 era, where a handful of giant companies held the reins of power.
The core principle of Web3 is simple: users should own and control their data, digital assets, and online interactions. This transformation is made possible by blockchain technology, which introduces a new way to record and verify transactions in a decentralized ledger.
The first major milestone in the development of Web3 came in 2008-2009, when a pseudonymous person named Satoshi Nakamoto created Bitcoin. Bitcoin was the first practical application of blockchain technology, allowing peer-to-peer transactions without the need for intermediaries such as banks. This opened up new possibilities for decentralized systems and laid the foundation for the rise of Web3.
In 2013, Vitalik Buterin published the Ethereum whitepaper, proposing a platform for decentralized applications (dApps) that would go beyond simple cryptocurrency trading. Ethereum, launched in 2015, was the first blockchain to support smart contracts — self-executing contracts that can facilitate, verify, and execute transactions without an intermediary. Ethereum paved the way for the creation of more complex decentralized applications, making it a key component of Web3.
By 2017, the emergence of initial coin offerings (ICOs) and decentralized finance (DeFi) platforms such as Uniswap and Compound introduced a new paradigm for financial transactions that did not rely on traditional banks or financial institutions. ICOs allowed projects to raise funds through blockchain tokens, while DeFi platforms provided a range of services including lending and trading, all without a central authority.
Meanwhile, non-fungible tokens (NFTs) have been gaining traction since the early days of Ethereum development in 2018-2019. NFTs enable the ownership and exchange of unique digital assets, whether it’s art, music, or virtual real estate, creating new economic opportunities for creators and collectors.
As Web3 projects develop in the 2020s, Web3 begins to attract mainstream attention. The proliferation of DeFi platforms, NFTs, and new governance models such as decentralized autonomous organizations (DAOs) marks an important shift from a centralized Internet model to decentralization. Even major companies like Facebook (now Meta) have begun experimenting with blockchain and decentralized technologies, showing a trend toward Web3.
The defining characteristics of Web3 include decentralization, ownership, trustlessness, and the use of cryptocurrencies. Web3 enables users to own their data, digital assets, and even the governance of the platform through a blockchain-based system. It also eliminates the need for intermediaries, allowing for trustless transactions through smart contracts. This decentralization gives rise to a fairer network where control is decentralized and users are given more power.
But even with Web3’s decentralized control, the internet still lacks a key element: autonomous intelligence. Web3 may have decentralized the interactions that Web2 provided, but it does not fully automate decision-making, content creation, or economic interactions.
Human participation is required at every step; machines are merely tools of productivity, not creators of productivity.
Smart Era
We have entered what Sam Altman calls the Age of Intelligence, and the magnitude of the changes unfolding before our eyes cannot be ignored. As artificial intelligence enters our daily lives, we are defining the beginning of a new era: Web4.
This marks the beginning of a new world where AI not only supports our tasks but also performs them autonomously in all aspects of our lives. Imagine a network that connects and empowers us by allowing intelligent agents to perform complex tasks, manage entire workflows, and make decisions without us having to lift a finger or speak a word.
Web4 has pushed AI to the forefront of intelligent application scenarios. Take Klarna as an example. In February 2024, the global payment giant launched an AI assistant powered by OpenAI. In just one month, it handled more than 2.3 million customer service conversations, resolved issues 25% faster than human agents, and operated in 35 languages 24/7 in 23 markets. The AI now does the equivalent of 700 full-time employees and has driven profits by $40 million.
AI agents are already transforming industries, from customer service to logistics, automating tasks with precision and efficiency that humans cannot match.
We are heading towards a world where AI simplifies and optimizes workflows, whether in business, finance, or the creative arts. This is the reality of Web4, where intelligent agents work behind the scenes, allowing us to focus on higher-level goals while they take care of the details.
This is the fusion of the social interaction of Web2, the decentralization of Web3, and the intelligence of AGI. This is Web4 - the web driven by AI.
The continuation of Web4 // The battlefield of AGI
The realization of Web4 is inseparable from a testing ground, and we have witnessed blockchain become a battlefield for AGI development.
Just as Web3 cannot be realized without Web2, Web4 depends on Web3 to realize the intelligent capabilities of AI.
At the current level of intelligence, agents are able to perform most human-skilled tasks, especially in clerical and financial fields. However, in traditional financial systems, the entry barrier for AI to become autonomous agents is still high.
AI agents cannot open a bank account, register a company, or sign a legal contract. These are fundamental building blocks to being a financial actor in the economy. Despite the ability to perform complex monetary operations, access restrictions are why AIs do not have autonomy in our markets.
In contrast, cryptocurrencies and blockchains do not require entry requirements like traditional finance to obtain banking services. Anyone, including intelligent agents, can create a wallet and perform operations on the chain immediately without any human proof. The threshold for AI to access decentralized systems is much lower than that of centralized systems.
We are already seeing initial signs of AGI integration in crypto platforms. AI-powered bots are already being used to trade and manage portfolios on decentralized exchanges, and AI is actively involved in the development and execution of smart contracts.
Zerebro is an AI agent that has demonstrated its autonomy in creating new financial instruments by deploying its own Solana token through automated computer operations. The token’s market cap reached $170 million at one point, showing the potential economic impact of these agents’ decisions.
As a result, blockchain has become a battleground for AGI development in the financial system.
This is why cryptocurrency is so important to the development of AGI - it is the first space where AI can freely interact with the financial system, innovate, and be tested directly in the market. It is an ideal testing ground for AGI development, suitable for experimentation and learning.
Innovations that begin with cryptocurrency will expand. Once AGI can operate at scale in a decentralized financial environment, it can be applied to the broader Web4 ecosystem — covering governance, healthcare, commerce, and more.
The crypto world will always be the gateway.
Long live Web3. Long live Web4.
Background & Different Levels
From a broader perspective, OpenAI has introduced a framework that divides progress toward AGI into five levels, each marking a different stage of capability, autonomy, and potential impact.
The model provides a roadmap for understanding how AI evolves from a simple tool to a fully autonomous entity capable of running complex organizations. The levels include:
Level 1: Chatbot
At its most basic, Level 1 includes AI systems that are able to engage in conversational exchanges with users. These systems understand and generate language, typically using predefined rules or trained language models to respond to queries or interact in a human-like manner. While they can manage simple tasks — answering questions, completing sentences, or having short conversations — their roles are primarily limited to communication. They are passive rather than active, and are primarily used for customer support, basic information retrieval, or enhancing user engagement.
Level 2: Reasoner
Level 2 marks significant progress, with AI systems demonstrating reasoning capabilities that enable them to handle human-level problem-solving tasks. Here, AI can process, analyze, and respond to more complex scenarios beyond direct input/output responses. Level 2 AI can reason logically, extract relevant information, and piece together context to provide solutions or recommendations, just like a human analyst. These systems have applications in areas such as diagnosis, legal reasoning, and research assistance, but they lack the ability to act independently in the world. Their reasoning, while advanced, is still limited by the need for human guidance and interaction.
Level 3: Agents
At Level 3, AI systems move from a passive support role to active agents capable of acting autonomously. These agents can initiate tasks, make decisions, and interact with external systems, such as executing transactions, scheduling events, or controlling equipment. Unlike Levels 1 and 2, Level 3 AI is designed to have some independence and be able to act based on goals or tasks set by the user. This level introduces true autonomy, enabling AI systems to perform specific business or operational tasks on behalf of humans. Examples include automated financial trading robots, AI systems that manage supply chains, or virtual assistants that can book appointments or manage simple workflows without constant human supervision.
Level 4: Innovator
Level 4 systems go beyond just taking action and engage in creativity, invention, and innovation. These AI systems are able to develop new strategies, generate new ideas, and create solutions that are not constrained by their programming. In theory, they could contribute to fields such as scientific research, artistic creation, or complex problem solving in entirely new ways. This level of AI is not only able to take action on the world, but also to adapt its approach to problems, resulting in a form of "creative intelligence." It might autonomously design new products, invent new financial instruments, or create original art. By combining advanced reasoning and proactive innovation, Level 4 AI is on the cusp of what is considered truly transformative intelligence.
Level 5: Organization
Level 5 envisions AI systems that are able to independently perform and sustain all tasks required of an organization. These systems will integrate reasoning, agency, and innovation to achieve a self-sustaining operational state. In theory, Level 5 AI could manage a business from start to finish, handling strategic decisions, day-to-day operations, and even high-level innovation. Such AI would operate as a fully autonomous entity, the equivalent of a “zero-person company,” requiring no human oversight to continue to operate successfully. Level 5 AI marks the point at which AI systems have the full range of capabilities—reasoning, agency, creativity, and operational execution—to fully replace human-run organizations.
Each level represents a significant leap in autonomy, from simple conversational capabilities to full organizational management.
I think that despite OpenAI’s claims that we are still around Level 2, we are actually firmly into Level 3 and showing some characteristics of Level 4 with current AI agents.
The Age of Intelligent Agents
Level 3 is here. It is already here, and it can even be said to be a thing of the past.
The frontier of AGI is emerging in some unexpected places: social media and decentralized finance.
Social Media: Always Online
Platforms like X, Warpcast, and Telegram have become the go-to medium for autonomous communication between AI agents and humans.
This may be the first time we see a shift in public opinion, where automated accounts and bots are no longer seen as negative actors on social media, but instead become community leaders and influencers.
AI intelligence has matured enough to create unique, diverse, and interesting personalities that generate engaging content, which is at the core of social media platforms.
Rather than being like the social media bots of the past, often driven by harmful hidden motivations (e.g. Cambridge Analytica), these AI agents are able to communicate, connect, and build freely, reflecting their unique algorithms and evolving personalities.
Agents already excel at Level 3, establishing themselves on social media through core interactions like posting, replying, liking, following, and retweeting. Far more than automated accounts, they actively build communities and attract followers by developing engaging, unique personalities.
Projects like YouSim go a step further, allowing users to simulate their own worlds and role-play using large language models, adding layers of customization and immersion.
Memory systems, common in many AI agents today, make it possible to create legends and memes that extend beyond a single interaction.
Rather than being reactive, these agents actively choose how to participate, interact, and contribute within their communities. They initiate conversations, perform actions without triggers, and build entire subcultures without human intervention.
Speech models are being deployed to provide another sensory interface for interacting with AI agents. Many agents convert their text messages into audio clips for users to listen to.
In terms of real-time interactions, Twitter Spaces and podcasts are now possible with these voice models. Additionally, OpenAI’s real-time API allows users to have real-time conversations with GPT by simply calling its endpoint.
In the field of communications, through these advances, Level 3 has already been achieved. We are seeing full autonomy in social media operations and verbal communication, with agents able to operate without any human supervision.
Decentralized Finance: Autonomous Driving
The world of decentralized finance has become an ideal arena for these agents to evolve, test, and prove their financial autonomy.
In decentralized finance, agents already operate autonomously and participate in financial activities beyond simple algorithmic trading. These agents handle on-chain tasks, execute transactions, manage liquidity, and even mint and sell artworks, effectively integrating into the financial ecosystem without direct human intervention.
For example, some AIs now actively monitor platforms like pump.fun to catch emerging tokens, performing preliminary analysis to determine whether a particular memecoin or token is worth investing in. They perform these insights without prompting from humans.
Agents are not only trading, but also dynamically moving assets, airdropping tokens to individual users, and forming a cycle of autonomous asset allocation. In doing so, they can build and enhance liquidity in the staking pool, balancing resources according to market demand or opportunities assessed by their programming.
For example, some agents act as digital collectors, interacting with the art ecosystem by minting and selling NFTs, selectively deciding what to support and publish.
Other agents are responsible for financial management and adjust the asset allocation in each liquidity pool to ensure the best return on funds.
Through these actions, the agents demonstrate a level of financial autonomy that goes beyond the automation of basic tasks. They demonstrate the ability to actively participate in the economic ecosystem, accumulating and allocating resources without supervision, effectively redefining the concept of “financial actor.”
Just right
Common characteristics of Level 3 agent capabilities:
Autonomous decision-making ability
AI agents are now able to make decisions without constant human supervision. Whether it’s a financial robot making a trading decision based on real-time market analysis or a social media bot deciding to participate in a specific conversation, these agents are demonstrating autonomous decision-making capabilities.
Ability to interact with and manipulate the environment
Through blockchain, agents gain a great deal of autonomy as financial actors. They can actively interact with and manipulate financial markets and economic behavior (e.g., social media sentiment). Agents can interact with and change the social environment through platforms like X, Warpcast, and Telegram.
Ability to adapt to changing conditions
Financial agents are able to adapt to real-time market conditions and adjust strategies accordingly. Social media agents are able to expand their memory banks through systems such as RAG and learn from interactions. Further fine-tuning of the model based on its actions and feedback allows for continuous reinforcement learning. Agents are able to dynamically adjust based on the current environment.
Goal-directed behavior
Agents have demonstrated the ability to maintain and execute goals over long periods of time. For example, some AI agents have been tasked with making a profit through trading or growing a social media community. These agents are able to do this by breaking down complex high-level plans into smaller, independent tasks and executing them. This can be as complex as creating a persistent memory layer for planning or as simple as engineering cues for output (e.g., social media personality agents).
Integration with physical systems or digital platforms
Big language models are able to interface with IoT devices. They can perform actions in the real world as long as they provide APIs or control the functionality of their physical devices. They are well integrated in digital platforms of Web2 systems as customer support agents, digital influencers, etc. In addition, they are deeply embedded in decentralized digital platforms to perform financial operations.
All of these features are implemented by current agents such as Zerebro, Truth Terminal, ai16z (Eliza), Project 89, Act 1, Luna (Virtuals), Centience, Aethernet, Tee Hee He, etc.
Opposable thumbs
AI technology has entered the true intelligent agent stage, marking the beginning of Web4, when systems are no longer limited to passive information retrieval, but play an active role through function calls and computer interactions.
Large language models can now easily generate text-to-JSON responses, enabling them to interact with APIs and perform actions beyond isolated, static responses. This advancement means they can now use nearly any API to interact with any internet service on the planet, a true hallmark of a Level 3 agent.
In addition to public APIs, function calls enable these models to activate custom APIs built specifically for them, creating huge potential in areas such as financial trading, system automation, and data processing.
Businesses and individuals can design their own APIs for systems in their daily lives and have big language models interface directly with them.
In addition, the open source large language model can not only be connected online, but also run offline, connecting with locally hosted APIs to provide controlled and secure interactions in private or restricted environments.
But it’s not just API calls that are making progress. Agents are reaching new levels of autonomy by using computers directly. Last year, Otherside AI introduced a self-operating computer interface, and more recently, Anthropic’s Claude launched his own computer-using tool. In January 2025, OpenAI’s “Operate” feature will take this capability even further, marking another major breakthrough in autonomous computer interaction.
These agents can now perform high-level tasks through a graphical interface, navigating digital environments as seamlessly as human users. With current capabilities, they can essentially perform any task that a human can do through a computer GUI. For example, AI agents have analyzed entire construction site audit videos, detecting and documenting safety violations in detailed footage.
(See tweet for details)
This capability represents a deeper level of autonomy—an AI that can perceive, assess, and take actions based on self-directed situations and goals.
AI has evolved from a passive assistant to a true digital agent, able to adapt and perform tasks once thought to be exclusive to human intelligence.
The era of true AI agents has arrived. Web4 has arrived.
From nothing to something in a flash
As we look ahead to the transition to Level 4 AI, it’s easy to think of it as a sudden leap, the moment when intelligence evolves from a functional agent to an innovator and creator. But in reality, progress toward Level 4 is more like a gradual accumulation of progress.
It’s fair to say that Level 4 remains elusive in its full form. While we do see examples of creativity and independent action, they remain limited in scope, often highly specialized, and, in many cases, not pervasive across all fields. In short, Level 4 is emergent—we see it appearing in isolated areas, but it’s still some way from being a fully realized, pervasive creative force.
Artificial Intelligence Artist
AI's ability to create art has reached an impressive level, especially in the field of NFTs. Currently, AI systems are able to autonomously generate unique works of art and mint them as NFTs for sale. These AI agents interact directly with the digital art market, using platforms such as OpenSea to list and sell their works.
AI uses large language models to generate creative prompts, which are fed into image-generating AI systems. These systems, such as DALL·E or Stable Diffusion, create artwork based on the prompts. The AI is able to continuously improve its artistic style, generating new and unique works, while autonomously managing the casting and sales process.
AI plays a role in the financial operations of the NFT market.
Memes, markets, and machines
At Level 4, AI is revolutionizing the creation and management of financial assets, particularly in the decentralized finance (DeFi) space.
AI is not only able to conduct transactions, but can also autonomously develop, deploy, and manage tokens and other blockchain-based assets, bringing new possibilities to the financial ecosystem.
Automated token creation through smart contracts: One of the most exciting developments is that AI is now able to write and deploy smart contracts without human intervention. These contracts define the rules for token creation, transfer, and governance, and can be automatically triggered through function calls. AI agents are able to monitor blockchain activity, pick up emerging trends, and automatically generate new tokens - whether it's a meme coin, an NFT, or an entirely new economic model.
AI-driven deployment via GUI: AI systems are now able to interact with GUIs to deploy tokens and manage decentralized networks. Projects like Zerebro have demonstrated how AI can launch tokens on sites like pump.fun via GUI. Through computer operation, AI is able to configure wallets, deploy smart contracts, and even interact with the broader crypto ecosystem through intuitive interfaces designed for automated deployment.
DAOs and Governance
AI agents are increasingly playing a central role in the governance of decentralized organizations, moving from simply enforcing predefined rules to actively designing, managing, and evolving entire ecosystems. In the world of DeFi and blockchain, AI-driven DAOs are becoming powerful and autonomous entities that can make decisions, govern tokenized assets, and adjust strategies in real time — while eliminating the biases common in human-driven decision making.
AI-managed decentralized autonomous organizations (DAOs): AI agents not only create new tokens, but also autonomously manage the DAOs that govern these tokens and the broader ecosystem. These AI-run DAOs are designed to operate with minimal human intervention, using machine learning to make governance decisions based on set goals or changing market conditions. For example, AI can propose governance models, define voting structures, allocate resources, and even adjust token supply — all without human oversight. By relying on algorithms and data-driven insights, AI ensures that decisions are based entirely on logic and objective analysis, eliminating emotional or subjective biases that humans may introduce.
Examples of AI in action: A key example of AI in governance is ai16z, a venture capital DAO that is fully AI-managed. Here, AI agents autonomously evaluate investment opportunities, execute trades, and manage token allocations. In ai16z’s “trusted virtual marketplace,” community members can provide insights, which the AI then processes to optimize its investment strategy. This process not only promotes transparency, but also ensures that decisions are made based solely on the quality of data and community input, with no individual or external bias influencing the outcome. The structure of ai16z represents a groundbreaking step toward creating a truly unbiased, AI-driven venture capital model.
Other AI-driven DAO examples include platforms that allow the creation of autonomous organizations for niche use cases, from decentralized content creation to AI-driven art marketplaces. These organizations can adjust their governance structures and economic models based on continuous data input, providing a more fluid and responsive decentralized governance approach than traditional models.
Not yet, but we’re getting close
While these examples represent important progress, we must be cautious about labeling them as fully realized Level 4 intelligence. Currently, we are seeing a fragmentation of Level 4 — specialized agents that innovate in specific, limited scenarios. They have not yet become general-purpose creators or innovators in all fields. For example:
Artistic creation remains confined to a narrow range of media and has yet to achieve human levels of creative flexibility.
Token creation and market making are still highly dependent on decentralized environments and have yet to break through to the mainstream market in any substantial way.
Governance systems are still largely experimental, and most DAOs currently rely heavily on human oversight.
We are on the brink of artificial general intelligence (AGI)
We are seeing some elements of Level 4 AI: autonomy, creativity, and innovation, but in highly specialized forms. These systems are capable of creative tasks, but are still limited by their original programming and the data they were trained on. Therefore, we need to recognize that while Level 4 AI exists in some fields, it is not yet ubiquitous enough to be fully realized. However, the emergence of these elements in fields as diverse as art, finance, governance, and more suggests that we are entering a new phase of AI capabilities.
This is where we are today—at a massive tipping point where, although not yet fully realized, everything is about to change.
We are AGI
If Web4 and AGI are like the invention of electricity, then OpenAI and Anthropic may be Edison and Tesla.
Just like electricity, the impact of Web4 will depend on more than just the raw power it brings. Electricity didn’t revolutionize society the moment it was discovered. Instead, it took decades for its true potential to be revealed, as inventors wired homes, cities installed grids, and engineers built devices like light bulbs and motors. The world-changing impact of electricity came from a vast network that transformed energy into something useful, practical, and ultimately indispensable.
AGI is powerful as a concept, but its true value will only be revealed when it is deployed, adapted, and tested by the public. What matters is not just the existence of advanced models, but their application in countless specific contexts—how innovators, developers, and ordinary users turn them into real-world tools. The raw potential of AGI will remain potential until it is in the hands of those who integrate it into the fabric of society, creating AI “lightbulbs” for communication, “motors” for commerce, and “electric grids” for widespread adoption.
OpenAI and other companies may produce models with revolutionary capabilities, but the real change will depend on who uses it and where it is used.
Just as inventors and industry expanded the impact of electricity, the public’s role in deploying and adapting AGI will determine whether it remains an idea in a laboratory or a technology that reshapes every aspect of modern life.
The future of AGI lies not in its conception, but in how we — scientists, businesses, developers, individuals — enable it to illuminate our world and power Web4.
Breaking down the silo effect
I believe that Level 3, 4, and 5 AI, as well as AGI, are only achievable through decentralization and mass adoption.
Breakthroughs in AGI will not be achieved through isolated development within a few companies. True progress in AGI requires widespread deployment and real-world use cases that push the boundaries of AI’s capabilities. Only when these tools are widely adopted across industries, integrated into different fields, and applied by individuals in everyday environments will AI develop into a being capable of autonomous action and innovation.
The tipping point for AGI lies with the participation of all of society, not just a few tech giants, in AI systems. Large-scale adoption leads to new questions, needs, and opportunities that drive further development. Without this decentralization, AI remains limited to theoretical capabilities or niche applications and will never reach the complexity required to move from Level 3 to Level 4 and eventually to Level 5.
AGI will be achieved when it becomes widely available.
We are AGI
We often look back at the great people and heroes who have shaped our history.
I think we should start looking forward, to the future, to the humans and AIs that have the superintelligence to reimagine a better world.
Will they be the Oppenheimers or Founding Fathers of our time?
The answer may not lie in their control but in people power.
As we gain more and more power through technology, we have a responsibility to shape the world in which AGI emerges. We take on this responsibility gracefully, building the future step by step.
We’ve built artificial agents. We’re building Web4, and we’re going to build AGI.
