LazAI Mainnet is online, and we had a chat with Metis about this move.

Written by: Eric, Foresight News

On the evening of December 22, Beijing time, LazAI, an AI data and application layer incubated by Metis, officially launched its Alpha mainnet. The last time Metis left a deep impression on me was when it was the first to launch a decentralized sorter. As many L2 projects have turned towards a transaction-centric direction in recent years, why did Metis confidently choose AI?

With questions in mind, we had a chat with Metis.

Focus on “data”, Metis's unconventional approach.

The Metis team indicated to the author that the launch of LazAI was not a spontaneous decision in response to the AI hype. As early as the beginning of this year, Metis had already determined its strategic direction focused on AI, and LazAI is the flagship product launched after nearly a year of in-depth development. LazAI is not merely a pure AI application or just a token issued by a simple AI product, but rather a network that serves AI model training and applications.

Creating a “Web3+AI” application may not be a good choice. The current level of AI development has not yet reached the point where it is worthwhile to integrate it with Web3 at the application level, or in other words, the certainty of application direction is not very high. The reason behind this is that the success of stablecoins and DeFi is due to the fact that the financial infrastructure in many countries or regions is not well-developed, leaving a market gap, while in terms of AI applications, I believe that in the short term, Web3 will not gain much advantage.

However, it is completely different at the non-application level. Looking back over the past one or two years, cloud service providers, including Alibaba Cloud and AWS, have integrated various tools or products related to L2 or Alt L1, including Sui, to varying degrees. This allows cloud service providers to offer more diverse options, and Web3 tools are often a more cost-effective choice.

In the author's view, it is a correct choice for Metis to launch LazAI by leveraging its advantages in verification capability and speed through its L2. Moreover, LazAI is not merely a simple application of Web3 concepts, but rather an original solution that is optimal both in engineering and market matching.

Let’s start with the image: The biggest feature of LazAI is that it considers a complete solution from data, training to application in its design. The entire life cycle of AI, from training to use and then to application based on AI, can be completed on LazAI.

To explain LazAI clearly, we first need to clarify three core components: iDAO, DATs, and the verifiable computing framework.

iDAO is the smallest unit participating in the network and is also a consensus node. It can represent any role in the AI lifecycle, such as data professionals providing data, AI models trained using data, entities providing computing power, and teams developing applications based on AI. LazAI separates the various participants in the AI ecosystem, providing greater modularity for AI.

DATs (Data Anchoring Tokens) are a semi-fungible token standard created by the LazAI team and are a core innovation of LazAI. DAT encodes three key attributes: the “ownership certificate” that proves the source of the asset and the identity of the author, the “usage rights” that define access quotas (e.g., number of inference calls), and the “value sharing” that allows holders to automatically earn income based on their proportion. DAT enables data contributors and AI developers to monetize their contributions and continuously earn income from users' usage in the future.

The verifiable computation framework is used to address the “black box” problem of AI computing, primarily to ensure the confirmation of data and model invocation processes. LazAI employs TEEs (Trusted Execution Environments), ZKPs (Zero-Knowledge Proofs), and OPs (Optimistic Proofs) to minimize trust in off-chain AI execution. TEEs provide private execution, ZKPs verify outputs without disclosing data, and OPs assume validity to optimize speed. This hybrid proof system is similar to ZK Rollup but specifically tailored for AI, balancing privacy, efficiency, and verifiability.

Based on this, we can clarify the overall workflow of the LazAI network: users submit encrypted data to iDAO, which packages it into LazAI Flow and sends it to Quorum via VSC. Quorum uses TEE/ZKP for verification and anchors the hash to LazChain. After on-chain verification, DATs can be minted, recording metadata and rights. Users transfer DATs to invoke services, which are executed off-chain by TEE, and the results are verified through ZKP/OP.

In this process, VSC (Verifiable Service Coordinator) can be understood as a group of experts used to verify the authenticity of specialized data, while Quorum is the consensus mechanism of LazChain. The iDAO, as a consensus node, not only assumes its own responsibilities but also ensures the operation of the consensus.

What can we do now that the Alpha mainnet is live?

LazAI is designed to address the issue of acquiring learnable data in the AI field. Currently, the Web3+AI projects we see, aside from x402, include incentive model networks such as computing power networks and AI Launchpads, as well as recently emerging projects that also focus on providing learnable data. From the author's perspective, the first two do not target real existing needs, but rather use Web3 as a better carrier for AI, while the latter's coverage is too narrow.

LazAI, designed for specific issues, has created a mechanism that allows contributors to continuously profit, which is hardcoded into the logic rather than having to be temporarily added each time to ensure the participants' interests.

The author learned from the team that LazAI's Alpha mainnet will not immediately launch a token. For those with professional knowledge who can contribute, as well as developers of AI models and products, it is both a rare opportunity for self-presentation and a chance to monetize their capabilities through airdrops. Additionally, LazAI will launch a developer incentive program for the Alpha mainnet with a total prize pool of 10,000 METIS, covering full-stage support from early prototypes to mature applications, and providing a multi-layered ecosystem empowerment that includes cross-social channel promotion and user growth funding pools.

Before the mainnet launch, LazAI has already achieved impressive results on the testnet. According to the team, the total number of active users on the testnet is close to 140,000, and the evolvable AI companion Lazbubu launched by the official team has also gained favor from nearly 15,000 users.

The rewards of the test network go beyond this, as the ROVR Network transforms everyday vehicles into smart 3D physical world data mappers using LazAI's solutions.

ROVR continuously maps the surrounding environment and generates rich geospatial datasets through its devices, inputting the data into the LazAI ecosystem. In this case, ROVR is an “iDAO”, and the uploaded data will be minted as DAT, allowing LazAI to have a high-precision DePIN and RWA database. In the future, AI autonomous driving tools, for example, can utilize this data for self-learning optimization.

The team stated that LazAI's team culture makes it very friendly to developers, as evidenced by the incentives provided to developers during this mainnet launch. This culture of valuing developers has also earned Metis the favor of scholars in the AI industry. In June of this year, Dr. Wang Zehua, a core member of the Blockchain Research Center at the University of British Columbia (UBC) and a part-time professor in the Department of Electrical and Computer Engineering, joined LazAI as a technical advisor. It is reported that Dr. Wang has long been engaged in the field of decentralized multi-agent system collaboration and security, with research focused on integrating AI and blockchain technology, particularly having a deep accumulation in areas such as trustworthy edge AI, blockchain and smart contract security, and zero-knowledge proofs.

The author mentioned at the beginning that Metis is the first L2 to bring decentralized ordering into practice, which is also a good reflection of its pursuit of technological iteration. This dedication to technology and attention to developers has laid a solid foundation for long-term development.

Why choose AI?

This question may seem a bit silly. As a hot concept, choosing AI appears to be an obvious choice, but the reality may not be as simple as it seems.

The challenges faced by general-purpose Ethereum L2s are becoming increasingly severe. Many projects choose to build their own L1 or develop application chains based on mature Rollups in pursuit of more customized performance. This compels L2s to reposition themselves and seek new directions based on their own characteristics.

Recently, the built-in Doubao phone launched by ByteDance has caused a sensation. The core of this sensation lies in AI, where users no longer need to interact with multiple apps, but only need to tell AI their needs, allowing AI to call various apps to achieve the user's goals. This fundamentally changes the logic of “gathering traffic” in the internet era, and the future traffic entry points are likely to become a competition among AIs.

The author gives this example to illustrate that even though many L2s have chosen trading, prediction markets, and RWA tokenization, they overlook the fact that the future operators of these applications may not necessarily be humans, but rather AI that receives instructions from humans. If the entry point for AI is missed, even the most numerous application chains will become mere workers for AI, and it is clear that Metis recognized this issue a year ago.

The author previously mentioned that Metis has actually been implementing an AI-centric strategy since the beginning of the year. In March of this year, Metis announced its dual-chain strategy at ETHDenver. In addition to Metis itself, Hyperion serves as a high-throughput L2 specifically optimized for AI applications, supporting parallel execution and instant feedback. Furthermore, Hyperion is deeply integrated with the Metis SDK, enabling modular construction of application chains and is positioned for high-frequency trading and real-time AI applications.

LazAI is the “flagship product” supported by this, and all previous layouts are now showing their true value. All L2s, including Metis, understand one thing very well: the efficiency advantage of L2 itself is being gradually eroded by the Ethereum mainnet, so there must be a capable product that can firmly establish itself in at least one track to ensure the chain has a stable usage rate and maintain the stable operation of the ecosystem. AI infrastructure is more like a “difficult but right thing.”

The use of Web3 solutions to optimize the problem of AI data labeling has only just begun to emerge in recent months, and Metis is also one of the first wave of pioneers. However, Metis's solution is a very typical Web3 Native solution, not just a simple introduction of on-chain confirmations and token issuance.

For Metis, the expansion of the on-chain application ecosystem and the strategy of using the chain as a settlement layer to some extent are progressing simultaneously. I believe that in the future, the price of tokens will increasingly be linked to their real value, and the extent to which the network can be adopted and the real demand for Gas tokens will determine the value of the tokens and the network. Entering the AI space also serves to enhance the value of METIS itself. If my prediction comes true, the emergence of more non-AI application chains based on the L2 stack will provide more value support for METIS.

Blockchain-based products have begun to permeate every aspect of internet applications, and their performance in the AI field is even more prominent. I still maintain that a purely “on-chain model” or “AI Launchpad” will not have a long lifespan, but products like LazAI that serve the AI lifecycle are different. For developers and users, products that are prioritized in ecological strategy are always worth paying attention to and participating in.