Source: Chi_Labs
Host Blake: Today we have invited Michael, the co-founder and CEO of 0G Labs. I am Blake, the host from Chi Labs.
Michael: At 0G Labs, our mission is to make AI a public good, which means providing infrastructure that everyone can contribute to, especially with transparency, security, and verifiability. This requires a completely different system from black-box AI. We have built a modular multi-layer system: an infinitely scalable Layer1 (execution layer), a storage network designed for AI workloads, a computation network for inference and fine-tuning (TEE-verifiable), a service marketplace, and AI alignment nodes that monitor drift and malicious behavior. Together, this forms a decentralized AI operating system.
Why choose to create 0G in 2023?
Host Blake: Why did you decide to launch 0G in 2023? What was the spark at that time?
Michael: At that time, we saw ChatGPT take off from OpenAI, and we believed it was a breakthrough moment for AI. We could finally interact with machines as if we were talking to people, receiving natural language responses, and the machine's answers were almost as effective as humans. This was truly a landmark breakthrough.
But then we started to think: what will happen five to ten years from now? What if some larger-scale use cases, especially at the societal level, are really driven by AI? For example, an airport run by AI. This kind of future made us very uneasy, because in black-box AI systems, it's hard to know where the data comes from, who labeled the data, what exactly happens inside the model, which version of the model you are getting, and how review decisions are made.
What pain points does 0G want to solve for the AI and Web3 industries?
Host Blake: What pain points does 0G want to solve for the AI and Web3 industries?
Michael: To truly own AI and be able to build "infinitely scalable" AI on-chain, we need to achieve several key infrastructure breakthroughs.
First, on the performance level: how to create a powerful enough Layer1 that can support the heaviest data throughput use cases. In many cases, AI data centers can reach throughput of hundreds of GB or even TB per second. So we must design a high-performance system architecture.
Second, on the research level: we need to figure out how to properly achieve AI alignment, and how to understand and build AGI in a decentralized context. Will AGI come from a single giant model, or from the combination of many small models? We lean towards the latter, but this requires a lot of research investment.
How do you hope 0G will be defined in 5 years?
Host Blake: How do you hope 0G will be defined in 5 years?
Michael: Our ultimate goal is to attract Web2 AI Builders into Web3.
When I envision 0G five years from now, I hope it will be the core hub for all mission-critical AI applications. For example, in manufacturing facilities, multi-robot systems, airports, logistics systems, and other societal-level scenarios, 0G can provide blockchain-based security and alignment mechanisms to ensure the operation of these systems. On this level, I hope 0G will always be at the forefront, at the core.
Of course, for some individual use cases, people can still use centralized AI or edge device AI to meet their needs. But when it comes to societal-level, mission-critical applications, 0G must be at the forefront, and all of this should be promoted in a community-centric way, ultimately making AI a public good.
What is the core architecture of a decentralized AI operating system?
Michael: I already mentioned some of this in the introduction. The core of 0G is a layered structure, with each layer performing different tasks needed to build large-scale AI applications.
It usually includes the following parts:
●Computing Power / DePIN Layer: This part is not owned by us, but is connected to computing resources provided by partners (such as Aethir, Akash).
● Software Layer:
○ Layer1 (execution layer): infinitely scalable.
○ Storage layer.
○ Data availability layer.
○ Computation layer: supports inference and fine-tuning, using TEE (Trusted Execution Environment) to ensure verifiability.
○ Service marketplace layer: similar to an App Store, where various services and public datasets can be plugged in, fully open.
○ AI alignment nodes: similar to a "police force," responsible for scanning model drift and negative behavior, and maintaining system health through penalizing staking.
Each of these components has a deep design space. For example, Layer1 itself is a modular architecture: modular execution layer, modular consensus layer, modular DA layer. Each layer is highly optimized and can scale infinitely; we are not limited by traditional blockchain systems. For example, some L2s may only achieve 500 TPS, but we use sharding, allowing any number of shards to be deployed as needed by applications, achieving any scale of TPS. This design philosophy runs through every part of the system.
Compared to traditional L1, what are 0G's performance advantages?
Michael: Our core philosophy is parallelization. Whether it's storage, data throughput, or TPS, all can be infinitely scaled through parallelization, achieving DA layer throughput of tens to hundreds of GB/s, and hundreds of thousands of TPS.
How does 0G attract a developer ecosystem?
Michael: One of our design philosophies is to minimize the barrier from Web2 to on-chain deployment.
Through modular design, all infrastructure components can be tightly integrated. If you use each part of 0G, they can work together seamlessly. This provides a one-stop experience: developers don't need to go to 50 different places to assemble solutions. In addition, the ecosystem itself is part of the attraction. By joining the largest Web3 × AI ecosystem, you gain a competitive advantage and can interact with others within the ecosystem. At the same time, we provide a lot of support, including incentives, mentors, accelerators, investment, etc.
Host Blake: So developers will have mentor support?
Michael: Yes, we provide not only mentor support, but also investment and help from all aspects of the ecosystem.
Are there early signs of a Killer App?
Michael: At first, because Web3 is still very finance-oriented, early killer apps are likely to be tied to finance, such as:
● AI yield search agents
● AI loop leverage agents
● Other agents supporting financial operations
Over time, as the community participates in training and using models, users provide computing power/data and are compensated with tokens. Once the infrastructure is ready, this will also be a killer app direction.
What experiences can ordinary users gain in the 0G ecosystem?
Michael: As I mentioned earlier, such as mentor support, marketing, and investment. Our goal is to quickly build a large-scale ecosystem, attracting the best builders to deploy their models, data, and computing power. For this, every part of the team needs to invest a lot of resources. We always see ourselves as providers of competitive advantage, so that anyone who joins 0g can benefit from it.
JT: In addition to what other public chains offer, as Michael just mentioned, such as investment, we actually have many extended functions. The overall advantage of 0G is that in addition to on-chain functions, we also provide computing power, storage, and support for all-chain AI. As a user or project party, you can get different usage scenarios on the 0G platform, or you can deploy fully decentralized AI applications yourself, and ordinary users can do this too, which is something many other chains may not be able to achieve.
How to balance airdrop scale with long-term user retention?
Michael: This is a very tricky issue. On the one hand, we need to recognize the contributions of community members, such as their participation in the testnet or support in promotion, but on the other hand, we don't want people to just complete tasks for short-term incentives and then leave when the mainnet goes live.
Our solution is: set part of the airdrop as vested, such as Yapper's allocation or some NFT rewards. This mechanism allows everyone to stay in the ecosystem for a longer period of time.
We always adhere to long-term thinking—because our mission is to make AI a public good. This is not a one- or two-year thing, but a long-term endeavor that requires continuous effort and deep participation from the community.
JT: We have launched an airdrop checking and registration tool to give content creators like Yapper an incentive, and at the same time provide continuous rewards through the Infofy platform, but continuous contribution is required to unlock them.
How will 0G avoid FUD similar to Movement and RedStone?
Michael: To some extent, it is difficult to completely avoid FUD, because everyone wants to get the maximum airdrop. No matter what you give, there will always be people who think it's not enough. This is the difference between perception and reality.
We have also learned many key lessons: separating mainnet launch from airdrop/listing is not good for the community and easily creates unnecessary FUD. It is necessary to distinguish between real/long-term users and Sybil attacks, which is difficult, but we continue to invest in identification and filtering. We have implemented these lessons in current decisions, splitting the airdrop into different parts: social tasks/community rewards, testnet participation. The participation mindset of these two groups is different, and different reward dimensions have been designed over a long period to attract more long-term participants.
For testnet participants: we hope they will also deploy on the mainnet, because mainnet metrics are real. Participating in the mainnet qualifies for community rewards. In short, we design with a long-term mindset. However, screening real users and Sybils is difficult, but we continue to invest in improvement. The airdrop will be divided into different parts, with different reward dimensions designed.
JT: We will segment different groups: testnet Runners, community activists, social media task participants, Yapper, early OGs, NFT holders. Everyone can get rewards from different dimensions, but the overall philosophy is to reward long-term contributors.
How does the team ensure fairness and enthusiasm for early supporters?
Michael: In addition to airdrops and other community rewards, greater benefits should come from long-term interaction with the ecosystem. For builders: there will be retrospective community rewards, investment, mentors, etc. For ordinary users: by continuously supporting and sharing the mission consensus, the value of tokens accumulated over time will grow.
JT: We give early OGs high weight in Discord, and have also issued One Group Grivity NFTs (free/0.1 ETH mint, peaked at 1.8–1.9 ETH). There are concentrated incentives before and after TGE, and long-term incentives will shift to ecosystem projects. By binding wallet + Twitter + Discord, multi-participants can get multiple rewards.
What role can the community play in designing incentive mechanisms?
Michael: We openly listen to community feedback as part of the governance mechanism. The community can propose new incentive mechanisms and reward methods. We will set up a Security Council to adopt suggestions, and incentive mechanisms can certainly be one of them. The 0g Foundation once quickly adjusted the initial tokenomics based on community feedback, and we held meetings until 4–5 a.m. to complete the revisions.
JT: When designing incentive mechanisms, the community can actually play a big role. As Michael mentioned, community users can propose things they want 0G to do and give suggestions, and the team will adjust and change based on these opinions. For example, we previously released a version of tokenomics, and the community gave a lot of feedback and suggestions. Our team quickly made changes based on these opinions, and the final feedback was very good. This is a relatively important example. I remember we even held meetings overnight, until four or five in the morning, just to respond to the community's voice as quickly as possible.
Key goals for the next 12–24 months
Michael:
Mainly two aspects:
● Build the largest Web3×AI ecosystem: around trends such as robotics, improve key infrastructure, new verticals, and key services, attract top AI dApps and top Agents to build on 0G.
● Benchmark/catch up with centralized black-box AI in infrastructure: strengthen verifiability mechanisms, support model training at any scale, further improve L1 performance, and bring anything from Web2 fully on-chain.
JT: The goal is to make 0G a fully on-chain evolvable AI ecosystem, reaching the level of top Web2 infrastructure in about two years, while maintaining transparency and decentralization.
Industry trend predictions (AI Infra, RWA+AI, stablecoins, etc.)
Michael:
● Stablecoins: Large institutions are already discussing replacing banking rails with stablecoin rails, and in the future, they will merge with AI Agents.
● RWA+AI: For example, tokenized hedge funds + collateralized stablecoin lending + loop strategies, AI Agents can monitor interest rates and risks, and automatically rebalance.
● Robotics: Household and factory robots will become popular, but security and alignment are key, otherwise the consequences of being hacked will be serious.
JT: Now everyone is increasingly using stablecoins, and many institutions have said they will use stablecoins instead of banks for transfers. At the same time, RWA plus AI is also a very promising topic. For example, someone stakes tokens and then borrows stablecoins. At this time, AI can help with risk balancing and position management before they are about to hit the liquidation line. Your previous supplement was already very complete, so I'll just add a little: Michael mentioned the Jackson Hole meeting earlier, which was actually a very important meeting held near Denver, USA last month. Many American institutions and officials discussed how to better integrate traditional banking and the crypto industry, including policy and future development directions. Michael also participated in this meeting as a representative of 0G.
