In 2024, the wave of AI Agents swept through the tech and crypto worlds.
While everyone's attention was focused on how OpenAI's GPTs could make AI more "useful," a team led by Amazon AI scientist and USC professor Salman Avestimehr was using the power of Crypto to tackle a more disruptive question: Who will own and control AGI (Artificial General Intelligence)?
Professor Salman Avestimehr
This team is ChainOpera AI. They did not choose the "upward scaling" path of a single giant model but opted for a more challenging "outward expansion" approach. Through a decentralized intelligent network co-created and co-owned by the community, countless specialized AI Agents collaborate to collectively emerge as "super agents."
ChainOpera AI's AI Terminal has already gathered thousands of AI Agents built by the community, gradually bringing this goal into the reality of Crypto.
From the decentralized consensus of PolyShard to the federated learning of FedML, and now to ChainOpera AI, Salman’s entrepreneurial journey has always revolved around a core theme: How can distributed collaboration unlock power beyond the individual?
With curiosity and questions about Crypto AGI and ChainOpera AI, BlockBeats engaged in an in-depth conversation with ChainOpera AI co-founder Professor Salman Avestimehr.
The Journey from Academia to Entrepreneurship
BlockBeats: First, could you briefly introduce what ChainOpera AI does in one sentence? What problem does it aim to solve?
Salman: You can think of it as a flywheel that connects two aspects. On one hand, it empowers community co-created AI by ensuring human ownership, value alignment, and fair participation for Agents; on the other hand, it unlocks the potential of DeFi and tokenized markets for ordinary users through an AI-driven intelligent layer.
BlockBeats: Your academic career began with research in information theory and blockchain, later shifting to federated learning and creating FedML. When you were doing this research, did you already envision the prototype of ChainOpera AI?
Salman: Since 2019, our initial vision has always been to achieve collaborative, community co-created AI. However, as AI itself has undergone tremendous changes in recent years, the question of where collaboration can create the most value has been constantly evolving.
When we started working on FedML, the biggest challenge was data—how to build better AI without compromising privacy. The core of FedML is to address data privacy and collaborative training issues in distributed machine learning.
Today, the biggest challenge has shifted to the control and value capture of intelligence. We are moving from solving data privacy issues to addressing collaboration and ownership issues of AI Agents.
ChainOpera AI is a natural evolution that applies our understanding of decentralized collaboration to this new frontier of AI Agents.
BlockBeats: What is the intrinsic logical connection between PolyShard, FedML, and ChainOpera AI?
Salman: The connection between PolyShard, FedML, and ChainOpera AI is my ongoing focus on distributed computing. I have been exploring the question: How can collaboration among multiple nodes, people, or systems unleash a power far beyond any single entity?
In PolyShard, we decentralized the consensus mechanism of blockchain, enabling thousands of participants to collaborate securely and efficiently.
In FedML, we decentralized AI model training, allowing data owners to collaboratively train more powerful models while protecting privacy.
In ChainOpera AI, we decentralized intelligence itself, building a co-created and co-owned ecosystem of AI Agents by users, developers, and infrastructure providers.
Thus, the main thread running through them is clear: through decentralization and collaboration, intelligence continues to evolve and expand.
BlockBeats: What unique advantages does your academic background bring to ChainOpera AI, and what challenges does it present?
Salman: I believe one of the unique advantages that a professor brings to a startup is that we are accustomed to raising funds based on a vision, communicating that vision to the community, and testing its feasibility through research. This is almost the entirety of the early stages of a startup—turning an idea into something credible and possible.
Of course, at some point, a startup is no longer just an idea but becomes a business, driven by clear metrics, growth, and execution, which is not where academia truly trains you.
So having the right co-founder is crucial, and I am fortunate to collaborate with Aiden (Chaoyang), who brings years of experience in large tech companies and knows how to scale and execute. After working in the industry for many years, he decided to pursue a PhD and joined my research group, which is how we met, co-founding FedML, and later naturally collaborating again on ChainOpera AI.
BlockBeats: Your team has a very diverse background, with AI scientists from Amazon and Google, as well as financial experts from Goldman Sachs and JPMorgan. What does this interdisciplinary team configuration mean for the development of ChainOpera AI?
Salman: This diverse team background is a conscious choice because we have two missions:
On one hand, we need top talent in AI and distributed systems to solve technical challenges, such as how to enable thousands of agents to collaborate efficiently and how to ensure the security and privacy of decentralized computing.
On the other hand, our goal is to make complex financial markets more understandable to ordinary people. This requires experts from the financial field who can understand the complexities of traditional finance and DeFi and translate that knowledge into safe and reliable strategies that AI agents can execute.
This interdisciplinary combination ensures that we can not only build cutting-edge technology but also ensure that this technology can solve the most valuable problems in the real world and provide real-world application value.
BlockBeats: In 2022, you judged that "the timing for Web3 AI is too early." Looking back now, that judgment was correct. What made you decide to fully commit to ChainOpera AI in 2024?
Salman: In 2022, the decentralization of AI had clear value for enterprises, but at the end-user level, the situation was different. Users did not feel "better decentralization," but rather better applications. Simply decentralizing the infrastructure layer did not make AI applications more useful or attractive to ordinary users.
By 2024, the situation began to change. AI Agents, as universally available and interactive AI applications, gradually emerged, coupled with increasing attention to who controls the direction of AI development and how independent developers can participate in the AI economy.
All of this clearly indicates that we need community-driven AI.
This is precisely where the opportunity for ChainOpera AI lies, as we empower community co-creation of collaborative AI Agents by moving up a layer. This way, end users can experience the network effects brought by multiple agents seamlessly executing complex tasks. Developers gain a way to collectively create and profit, rather than being monopolized by tech giants.
As a global community, all our members have a voice in questions like "how to build a new type of AGI" and "whose values does AGI ultimately reflect."
BlockBeats: Your AI Terminal is already online and has gathered thousands of community-built AI Agents. Can you reveal what types of Agents are most commonly used by users right now?
Salman: Currently, the types of Agents most commonly used by users are primarily concentrated in two areas:
The first area is finance and trading. This includes market analysts, arbitrage strategists, portfolio optimizers, etc. This is not surprising, as DeFi and the crypto market are ideal environments for AI Agents to operate autonomously and transparently.
The second area is content creation and code generation. Users are leveraging Agents to automate their daily workflows, such as generating marketing copy, writing smart contract code, or conducting complex document analysis.
As we continue to expand the capabilities and toolsets of the Agents, we expect the use cases of users to become more diverse.
Collaborative Intelligence through "Outward Expansion"
BlockBeats: The core argument of ChainOpera AI is that AGI will not come from a single large model but from collaborative intelligence. What is the essential difference between this and the multi-model or multi-agent architectures being explored by giants like OpenAI and Anthropic?
Salman: The core difference lies in the technical path and economic model.
In terms of technical path, giants like OpenAI are taking the "upward scaling" route, building a few massive, single models that become smarter through scale and computing power. In contrast, ChainOpera AI is taking the "outward expansion" route, building a network composed of many smaller, more specialized AI Agents that form collective intelligence through collaboration. This is a fundamentally different path to AGI, one that grows through collaboration rather than centralization.
In terms of economic model, we are creating a collaborative AI economic model. In traditional centralized models, innovation is controlled by capital, and small teams are constantly squeezed. In the ChainOpera AI system, any developer, researcher, or resource provider can contribute their efforts and share the value they create.
Developers co-build a super AI terminal by contributing their Agents. They not only gain recognition and usage but also realize shared economic value generated by the network's adoption of their Agents, thus achieving co-ownership of the AI terminal.
BlockBeats: How is your proposed "Co-Create, Co-Own" implemented technically?
Salman: The slogan "Co-Create, Co-Own" is realized through our AI Terminal application, which you can think of as a decentralized ChatGPT. Within this terminal, there are thousands of AI Agents built by the community, all of which can be accessed through our super AI entity CoCo, which is responsible for coordinating and orchestrating these Agents to collaboratively execute complex tasks.
BlockBeats: Your white paper describes a grand four-layer architecture. Could you walk us through how these four layers work together using a specific user scenario, such as a DeFi trader wanting to create an automated trading strategy?
Salman: Certainly. Let’s take a DeFi trader who wants to use ChainOpera AI to build an automated trading strategy as an example.
He starts at the application layer, which is the AI Terminal application, the main interface through which our users interact with thousands of Agents. The trader can use specific trading Agents already available in the terminal, such as grid trading strategists, limit order Agents, or market analysts, or he can combine them through our Agent Social Network to collaborate on executing more complex strategies.
In this case, our super Agent CoCo will be responsible for the orchestration and coordination between these Agents to ensure everything runs smoothly.
All of this operates on a decentralized AI and GPU infrastructure layer powered by decentralized computing nodes that can scale to perform model inference, training, and Agent deployment.
Therefore, from the trader's perspective, it feels like using a seamless, intelligent system, but in reality, it is a network composed of AI Agents, developers, and decentralized computing providers, aimed at making complex DeFi trading accessible to anyone.
BlockBeats: In this scenario, how is the user's data and strategy protected?
Salman: All communication between users and the Agents or model servers is encrypted, ensuring that no third party can intercept or read any data during transmission. The computing nodes isolate different Agents and models using container technology, allowing each task to run in an independent secure environment without affecting one another.
For users who require stronger protection, we offer flexible deployment options. Agents and models can run in a Trusted Execution Environment (TEE), on preferred cloud providers, or even on the user's own local hardware, thus achieving complete control over data and execution privacy.
As for the strategy itself, it depends on how you interact with the system. If you are using community-built Agents, the logic of the Agent is public, but your specific parameters, data inputs, and usage patterns remain private.
If you wish to keep the strategy completely confidential, you can create and deploy a private Agent that will not be published on the AI Terminal but is designed for your own use.
BlockBeats: If this trading strategy performs well, how can other users utilize it? How does the creator get incentivized?
Salman: If a trading strategy Agent proves successful, the developer can publish it as an Agent and make it accessible to others through the AI Terminal. This way, anyone in the community can try it, adjust it, or even build upon it.
As more users interact with that Agent, the creator will earn developer points. This is our current way of recognizing and rewarding those who contribute to the ecosystem. Over time, as we continue to expand, these points will evolve into formal income and incentive mechanisms, providing developers with a transparent and fair way to benefit from the adoption of their work.
BlockBeats: Your roadmap mentions many innovative concepts, such as Agent routers and Agent Social Networks. These concepts sound cutting-edge but also raise concerns about complexity. How do you ensure that ChainOpera AI finds a balance between technical innovation and user experience?
Salman: This is the beauty of building a decentralized AI ecosystem driven by applications.
All these complex technologies, such as Agent routers, collaborative Agent frameworks, decentralized model platforms, etc., will work in the background. They enhance our AI application experience, which is our decentralized ChatGPT, making it more powerful and valuable for end users.
We recently shared an article explaining how these components come together to create a seamless "from prompt to response" intelligent system. Users simply interact naturally, while the Agent network takes care of the complex and heavy lifting in the background.
Community Value and Future Development
BlockBeats: Frankly, the $COAI token has experienced significant price volatility in recent months, and there are concerns about the concentration of token distribution within the community. Many are asking, where is the value anchor for ChainOpera AI?
Salman: While market volatility is inherently unpredictable, analysts have pointed out that the price fluctuations of COAI are influenced by various factors. These include the bull market in the BNB ecosystem at the beginning of October, the rise of perpetual contract trading, COAI being one of the first tokens listed on Aster, and the subsequent "black swan" liquidity event on October 11 that affected the entire crypto market.
Regarding token distribution, research indicates that a certain degree of concentration is typical in the early stages of a project’s token issuance.
In most cases, these tokens belong to the team, early investors, and locked shares for ecosystem plans, following a transparent and on-chain verifiable vesting and unlocking schedule. Many community members have pointed out that the early concentration of COAI is lower than that of many AI tokens launched around the same time, which is a fact anyone can verify through public channels.
As for the fundamental value, COAI is a utility token that drives the open collaborative AI economy of ChainOpera AI. It supports users in paying Agent usage fees, microtransaction settlements between Agents, and GPU computing costs when developers deploy new Agents.
The long-term value of COAI ultimately comes from the real economic activities it empowers: a growing network of users, developers, and AI Agents who collaboratively build a decentralized, community-owned, utility-driven intelligence.
BlockBeats: You emphasize on social media that the team is still focused on development, but for many investors, such statements are not enough. Can you specifically discuss what verifiable milestones are expected in the next 6 to 12 months?
Salman: ChainOpera AI has always prioritized actual building over empty talk. We have released and announced many new AI Agents and features, all of which are live on our AI Terminal and available for anyone to try today. This is the best proof of progress; real products are more persuasive than promises.
As for the next steps, our larger plans and roadmap are outlined in the white paper, which details the next phase of development for the AI Agent network, Agent developer platform, and decentralized computing layer.
There will be several exciting releases this quarter and next. We will regularly share these highlights on our official social channels so that the community can transparently and in real-time track the achievement of each new milestone.
BlockBeats: How do you plan to involve the community in the product development and validation process?
Salman: The community has been deeply involved in shaping ChainOpera AI. Many members are creating and publishing their own AI Agents, while others are actively using these Agents and sharing feedback to help developers improve them. We have also launched a new ambassador program to reward community members who showcase and promote the real value of our ecosystem.
Looking ahead, we are exploring more ways to bring the community into the development loop. These include validating before the release of AI Agents, contributing data and feedback to help these Agents become smarter and more reliable over time.
Our goal is simple: to make ChainOpera AI a truly collaborative, community co-created intelligent network where everyone plays an indispensable role in its evolution.
BlockBeats: AI projects are now facing intense competition. If you had to summarize in one sentence how ChainOpera AI differs from these competitors, what would you say?
Salman: The key difference is that ChainOpera AI is building a truly application and utility-driven decentralized AI infrastructure. All underlying technologies converge into a tangible Crypto AGI application that anyone can use, benefit from, and help grow. We encourage everyone to try it out and experience it firsthand.
BlockBeats: How do you define the success of COAI? Is it user numbers, token price, or other metrics?
Salman: For me, success is about impact. The broader goal of Crypto AGI is to make complex financial markets accessible to ordinary people through a community-built intelligent layer.
So one upper-level metric I focus on is simple: how many non-crypto users can we help enter the complex world of crypto and DeFi through this intelligent, agent-driven, collaborative ecosystem?
From Crypto AGI to Physical AI
BlockBeats: Your roadmap outlines a grand vision from "autonomous AI subnetworks" to "physical AI." In your vision, what will ChainOpera AI look like in 2030? How will it change our lives?
Salman: In the next one to two years, our focus is on achieving Crypto AGI, which is a community-built intelligent layer designed to help everyone participate and benefit from complex financial markets. The goal is to make these markets understandable, accessible, and rewarding for ordinary people through intelligent, collaborative AI entities.
Looking further to 2030, we see this evolving into a broader frontier we call physical AI. As agent systems mature, their roles will expand from managing digital finance to coordinating elements of our physical world. This includes managing smart homes, autonomous vehicles, and personal robots that assist with everyday tasks.
By 2030, we envision a world where AI entities seamlessly collaborate in both digital and physical realms, co-created and co-owned by the community, transforming intelligence itself into a shared global infrastructure that empowers everyone.
BlockBeats: How do you ensure that ChainOpera AI truly realizes "Co-Create, Co-Own," rather than ultimately being dominated by the core team like many projects? Have you envisioned a day when you and the core team will "step back" and let the community take full control of the project?
Salman: This is indeed our long-term goal. "Co-Create, Co-Own" means that over time, the community becomes the main driving force behind the ecosystem. We have already taken steps in this direction by open-sourcing key components, empowering community-built Agents, and introducing ways for users to directly contribute to improving the AI.
As the network matures, we will introduce governance mechanisms to gradually shift the core team's role from control to coordination, ultimately allowing the community to fully own the evolution of ChainOpera AI.
BlockBeats: What would you do if the community's decisions conflict with your vision?
Salman: I believe the essence of building something truly decentralized is that you must trust the community. Our team's role is to help lay the foundation and guide the early vision, but if the community later steers things in a different direction, that is a sign that the ecosystem has matured.
BlockBeats: Looking back on this journey, if you had to summarize the core narrative of this journey in one sentence, what would you say? In your mind, what kind of story is ChainOpera AI?
Salman: ChainOpera AI is a story about building a community-owned intelligent layer where humans and AI collaborate to make complex systems, like financial markets, accessible, transparent, and beneficial for everyone. It represents a fork in the road toward AGI, a path that we all co-create and co-own.
BlockBeats: If someone were to write a book about ChainOpera AI ten years from now, what would you hope the title to be?
Salman: I would hope it to be: "Collective Mind: How We Co-Create Superintelligence."
免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。
