Dialogue with Hemera CEO: How to Build an AI-Native Operating System for the Web 3.0 Era Using Crypto + AI?

Questioner: Mia, ChainCatcher

Responder: Arthur Meng, Co-founder and CEO of Hemera

Recently, W3W.ai announced its brand upgrade to Hemera, providing a public infrastructure protocol for high-performance AI and data, making Hemera a unique new competitor in the AI and data track of Web3.

In fact, with the rise of Sora under OpenAI, the AI field has once again sparked a technological wave. The equation AI + Crypto = ? has become a hot topic in the current innovative technology sector. In a16z's "2024 Outlook Checklist," "AI + Crypto" is included, emphasizing that cryptographic technology can create multi-sided, global, permissionless markets, allowing anyone to contribute computing power or new datasets to the network and receive corresponding compensation. This sharing and utilization of resources not only reduces the cost of AI but also makes data more accessible.

At the same time, Vitalik Buterin also envisioned the potential of "AI + Crypto" in his article, believing that cryptographic technology can find a balance between the centralization and transparency of AI, further optimizing data storage for AI.

Hemera is steadfastly exploring the potential of Crypto + AI, leveraging the latest AI revolution to empower the Crypto industry, creating an AI-native operating system for the entire industry through a high-performance data indexing network and deeply optimized large models for Web3. According to Hemera's founder Arthur Meng, the Hemera Protocol will be built into a public infrastructure for the entire industry, using AI capabilities to redefine how users interact with blockchain in the web3.0 internet era.

Arthur Meng, co-founder and CEO of Hemera, holds a Ph.D. in Physics from Stanford University and has co-authored papers with several Nobel laureates in journals such as Nature and Nature Communications.

Hemera is committed to providing real-time, high-performance data infrastructure services for the global blockchain industry. Its new ecosystem has flourished, building a series of commercial products and services for developers based on the Hemera protocol, such as the SocialScan blockchain explorer, airdrop and community quality audits, and the SocialScan Agent Store, meeting various industry data needs.

Hemera has completed a $2.6 million seed round of financing, with investors including Nomad Capital, LIF, SNZ, Chainlink, as well as founders of ZetaChain, Sending Labs, Ultiverse, and angel investors like Danny Zhang, co-founder of Wish, and Shen Xiangyang, Senior Vice President of Microsoft and AI expert. Currently, its official website has been newly revamped and launched, and the website domain and official X account have also been migrated and renamed.

Recently, ChainCatcher interviewed CEO Arthur Meng to unveil the mystery of Hemera.

The Founder's Academic and Entrepreneurial Journey: From AI Exploration to Web3

ChainCatcher: You previously pursued a Ph.D. in Physics at Stanford. What prompted you to venture into AI entrepreneurship and lead your team into the Web3 field?

Arthur Meng: Many people in the venture capital circle don't quite understand "doing a Ph.D." Some, influenced by Silicon Valley mogul Peter Thiel, hold some interesting "prejudices," thinking that the earlier you "drop out," the better. I have the authority to speak on this from "the other side of the mountain": pursuing a top Ph.D. is very similar to entrepreneurship—first, like entrepreneurship, a Ph.D. is not suitable for everyone; at the same time, the project direction you choose must achieve maximum results within a limited time and with limited resources. If the direction is too broad or distant, it may lead to situations that cannot be accomplished under current conditions; if the direction is too small or ordinary, it may not even allow for publication in top journals or a smooth graduation. During my time at Stanford, I had a strong interest in AI and data, participated in many AI projects, and successfully integrated AI into physics research, achieving very good results, with some outcomes published in the main issue of Nature (impact factor 65).

After graduating from Stanford, I joined an AI company, responsible for algorithm development, engineering development, and architecture, turning the company's unique machine learning and deep learning algorithms into enterprise service software sold to large companies for user behavior analysis. Major clients included Pinterest, Yelp, Alipay, American Express, Funplus, etc., helping them detect large-scale registration attacks and predict whale customers. This experience greatly inspired our current work on the Hemera Protocol.

ChainCatcher: W3W announced its brand upgrade to Hemera Protocol in March, providing AI and data infrastructure services. What led the team to pivot to the AI field?

Arthur Meng: Actually, when we first entered Web3, we were already exploring the potential applications and scenarios of AI in Web3. At that time, when mentioning Web3 and AI, most people felt resistant or thought it was unreasonable. However, from day one, we believed in the immense potential of the massive on-chain data in Web3, and that the industry was still in a very early stage. Whoever can effectively solve the problem of utilizing on-chain data will have a competitive advantage in the future large-scale application of AI in Web3.

As mentioned earlier, our entry point into Web3 is closely related to the business of our previous startup. How to leverage community members' on-chain information and evidence to build Web3 identities, thereby helping the community grow and operate, was our first scenario to consider. We quickly realized that this question is not easy to answer: to address it requires very high-quality, high-precision account-level granular data capabilities. The user transaction forms in Web3 are diverse, and with the maturation of Ethereum's scaling solutions, the modular blockchain tech stack has become mainstream. Organizing and utilizing interaction information across multiple blockchains and contracts is very cumbersome. On the other hand, almost all the communities we collaborate with, including NFTs, games, social networks, and even wallets, need such solutions. We soon discovered that the existing data infrastructure in the industry could not adequately support the scenarios we wanted to pursue. Most of them are centralized B2B SaaS data companies, which are limited by cost control issues and cannot quickly support hundreds of application chains (appchains). Another type of infrastructure, such as The Graph, caught our attention: The Graph provides decentralized data indexing capabilities, which can effectively handle the granularity of data at the smart contract level, but is powerless regarding the "account-level" data granularity we focus on.

After nearly a year of research and effort, we developed the first version of the Hemera Protocol—a decentralized account-level real-time data indexing protocol. It can effectively meet the needs of developers across the industry for user account-level asset information, social community information, and Web3 on-chain reputation, while also supporting essential infrastructure services required by various public chain ecosystems, such as blockchain explorers and data APIs. In this process, we quickly realized that in the context of the AI revolution across the internet, the Hemera Protocol should also evolve with the times, transforming into the next generation of AI-native data infrastructure: building on the first generation of Hemera, which accessed data through traditional methods like SQL and GraphQL, we proposed for the first time the integration of large models fine-tuned using on-chain data, allowing ordinary users to call on-chain information in English and integrating an Agent-as-a-service model. This enables developers of various Web3 applications, protocols, and public chains to easily customize their own AI Agents, providing AI capabilities for the entire industry and comprehensively upgrading the interaction experience between Web3 users and Web3.

The SocialScan Explorer blockchain explorer is built on the Hemera Protocol as an infrastructure application aimed at public chain ecosystems, providing high-performance blockchain explorers for major Layer 1, Layer 2 public chains, and application chains. Currently, we have established partnerships with several well-known partners and clients, including public chain ecosystems like Polygon, Linea, Manta Network, Taiko, Mantle Network, Celestia, Xion, Story Protocol, Merlin, etc. At the same time, we have established good strategic partnerships with leading RaaS projects in the infrastructure field, such as Caldera, Altlayer, and Conduit, to jointly serve App clients under the Ethereum scaling and rollup trends.

Additionally, in the SocialScan community interface, we provide an entry point for the Hemera AI Agent Store for Web3 end users, allowing both seasoned Web3 users and newcomers to efficiently track assets and trade using real-time on-chain data through AI Agents. Since the launch of the invite-only beta test in early March, it has attracted over 500,000 registered users and generated around 600,000 on-chain interactions.

The Birth of Hemera Protocol

ChainCatcher: What is the meaning behind the name Hemera Protocol, and what services will it provide to blockchain users?

Arthur Meng: Hemera comes from the Greek goddess of daylight, coincidentally a sibling of Ether, and we hope that the Hemera Protocol can shine like sunlight on the vast amounts of on-chain data, allowing everyone to easily discover and utilize truly valuable signals.

The Hemera Protocol aims to be a public infrastructure serving the entire industry, providing real-time, efficient data services and AI capabilities for the entire Web3 field. Analogous to water and electricity in real life, the data and AI capabilities in the future Web3 world will become more valuable and efficient under the promotion of Hemera. On-chain data contains enormous opportunities but also faces challenges such as information fragmentation and difficulty in information filtering. Therefore, Hemera will utilize technologies like artificial intelligence to sift through the vast amounts of on-chain data to extract truly valuable signals, providing strong decision-making support for developers, investors, and traders. Hemera aims to become the data middle platform in the Web3 field, allowing the entire industry to easily access and process on-chain data through AI.

Achieving Interoperability and Sharing of On-Chain Data

Regardless of which chain the data is on, as long as it goes through the Hemera Protocol, it can be easily accessed and shared, significantly reducing the cost of data acquisition and processing and improving the operational efficiency of the entire industry.

Mining and Analyzing On-Chain Data with Large Language Models

The Hemera Protocol utilizes artificial intelligence technology to deeply mine and analyze on-chain data. By training large language models or AI algorithms, we can help users filter information that meets their needs, providing personalized data services. This way, whether for developers, investors, or traders, they can more conveniently obtain the data support they need, improving the accuracy and efficiency of decision-making.

Open Source Applications

As a public infrastructure, Hemera is open-source, and it will also closely collaborate with partners, developer communities, and others to provide strong data support for the development of the Web3 field.

Application Scenarios of Hemera

ChainCatcher: How does Hemera leverage large language models to bring value to the Web3 industry? What are the specific application scenarios of Hemera?

Arthur Meng: The technical architecture of the Hemera Protocol incorporates large language models, changing the way users interact with data, achieving an iteration from passive to active user demand. It can fully leverage the advantages of data interoperability in the Web3 industry. Through large language models, Hemera can proactively capture and meet user needs, completely breaking the limitations of traditional applications where users passively receive information. At the same time, the data interoperability in Web3 provides a broad stage for the innovation of the Hemera Protocol, allowing us to fully utilize data across the network to provide users with more precise and comprehensive services.

Given that the Web3 industry is still in its early stages, where users primarily focus on asset management and discovering new projects, our first landing scenario is to develop a cross-chain, fully interoperable AI agent to help users filter noteworthy projects from the Ethereum ecosystem based on factors like high gas fees, community quality, and investor behavior, thereby enhancing the efficiency of user investment decisions.

In the initial phase, we plan to provide real-time delivery of on-chain data to users through simple web push notifications, Telegram, Discord, or email. One of Hemera's innovations is to utilize AI and data capabilities to provide adaptable agents for various infrastructures, protocols, and applications to better serve their clients. Through artificial intelligence, we will help creators and IPs in the ecosystem deliver their offerings to the Web3 community in a higher quality manner. We plan to place agents from different fields in the Hemera Agent Store to meet users' diverse needs.

Interoperability

ChainCatcher: How does Hemera build and develop its ecosystem, and what are the criteria for selecting partners?

Arthur Meng: Hemera is committed to achieving unlimited interoperability, stemming from the team's deep understanding that in the Web3 era, both large enterprises and startups are more or less involved with on-chain data. The diversity of chains and the prosperity of ecosystems are extremely beneficial for our project development. Therefore, the team has already collaborated with multiple public chains and well-known projects to expand Hemera's influence and application scope, meeting the needs of more ecosystems and applications. Currently, Hemera has connected with numerous public chains, including the Ethereum mainnet and its well-known L2 solutions, as well as other L1 public chains and the Cosmos ecosystem.

Our strategy is to work closely with public chains, infrastructure, and application ecosystems that share a similar vision to jointly promote industry development. Only through collaboration can we achieve true win-win outcomes, which is why the team has been actively seeking more partners to empower applications in this industry.

The First Ecosystem Application: SocialScan

ChainCatcher: What are the product features and positioning of SocialScan? Why create an application product?

Arthur Meng: SocialScan is positioned as the next-generation Web3 interaction interface. We divide this interaction interface into two different segments: one for developers and one for end users. SocialScan Explorer is a blockchain explorer aimed at developers, empowering the modular blockchain ecosystem using the data infrastructure of the Hemera Protocol, while also providing data capabilities for Hemera AI Agents. The end-user interface of SocialScan is positioned as the Hemera Agent Store, analogous to the GPT Store in Web2. Through this application, users can access and connect with numerous Web3 protocols, infrastructures, and applications, interacting with them through AI. SocialScan is not only a window showcasing our data capabilities but also a platform that allows users to directly interact with our partners' AI Agents. Through it, users can easily find the infrastructure protocols and applications across various chains and ecosystems and interact with them. Currently, over 500,000 users have registered for SocialScan, generating nearly 600,000 on-chain interaction records.

In addition to collaboration and community building, the team also places great emphasis on product iteration and optimization. We understand that the emergence of large language models has disrupted the way users interact with server-side applications, so the team hopes to enable users to interact more directly with our AI through the SocialScan application, thereby accelerating the iteration speed of the Hemera Protocol.

Community Incentive Program

ChainCatcher: How will the project incentivize users participating in the construction of Hemera's big data algorithms?

Arthur Meng: Currently, the Hemera Protocol mainly attracts several types of users: C-end community users, algorithm construction users, and computing power network participants. For C-end users, Hemera showcases its artificial intelligence capabilities and provides data labeling functions, attracting participation through point rewards, which will eventually be converted into tokens.

For algorithm construction users, Hemera aims to attract AI and data experts from both Web2 and Web3 to contribute algorithms to enhance the project's value. We have drawn inspiration from similar projects but have conducted in-depth iterations and updates on both technical and business levels. In the future, we will establish a developer community to encourage data scientists and AI experts to provide algorithms to improve the Web3 user experience.

Computing power network providers will also have the opportunity to participate in Hemera's decentralized data indexing protocol and receive rewards through computing power. At the same time, we will continue to promote community development, rewarding community users through token airdrops and growth plans.

Team Expansion

ChainCatcher: Regarding the development roadmap, what are Hemera's future development plans?

Arthur Meng: The Hemera team has been accelerating its development, especially in Q1 of this year, where the team size has significantly expanded, and community users have seen substantial growth. In today's early blue ocean markets for both AI and Crypto industries, we plan to accelerate financing and product commercialization.

In my view, AI can greatly drive the future development of the Web3 industry. As a team that has been working in the AI and data field for many years, Hemera began contemplating this issue over a year ago. Vertical data in Web3, especially full-chain data, holds special significance for AI development. Although there are many challenges in using on-chain data for AI, we believe it is a direction worth exploring.

From a business model perspective, AI is largely an enterprise-facing business. However, in the Web2 environment, due to the monopoly and resource advantages of large enterprises, small companies face significant challenges in developing AI. In contrast, Web3 offers massive liquidity and incremental markets, providing a broader space for the application and development of AI. Most of Web3 belongs to computing power networks and has not yet entered the true AI application infrastructure algorithms, so I believe this area holds great promise for the future.

Unlimited interoperability is Hemera's core competitive advantage. We look forward to all applications and infrastructures having their own agents on our platform in the future, jointly promoting industry development. We strongly encourage builders and AI experts to actively join us in creating exclusive agents. In the future, Hemera will continue to increase investment and research efforts, driving rapid project development and bringing more value to users and the community.