Exclusive Interview with Michael, Founder of 0G Labs: How to Unlock the Future of Decentralized AI?
TechFlow Selected TechFlow Selected
Exclusive Interview with Michael, Founder of 0G Labs: How to Unlock the Future of Decentralized AI?
"Our ultimate goal is to attract Web2 AI Builders into Web3."
Navigating Web3 tides with focused insightsSource: Chi_Labs
Host Blake: Today we have Michael, co-founder and CEO of 0G Labs. I'm Blake, host from Chi Labs.
Michael: At 0G Labs, our mission is to make AI a public good—meaning we provide an infrastructure that anyone can contribute to, with transparency, security, and verifiability. This requires a system entirely different from black-box AI. We've built a modular, multi-layer system: an infinitely scalable Layer1 (execution layer), a storage network tailored for AI workloads, a computation network for inference and fine-tuning (verifiable via TEE), a service marketplace, and AI alignment nodes that monitor drift and malicious behavior. Together, this forms a decentralized AI operating system.
Why launch 0G in 2023?
Host Blake: Why did you decide to launch 0G in 2023? What was the spark?
Michael: We saw ChatGPT take off from OpenAI, and we recognized it as a breakthrough moment for AI. For the first time, we could converse with machines naturally and receive human-like responses with near-human effectiveness. That was truly iconic.
But then we started thinking: what happens five to ten years from now? What if large-scale, societal-level use cases are actually driven by AI? Imagine an airport fully operated by AI. That future made us deeply uneasy, because in black-box AI systems, it's hard to know where data comes from, who labeled it, what's happening inside the model, which version of the model you're actually using, or how decisions are being censored.
What pain points does 0G aim to solve in AI and Web3?
Host Blake: What pain points in the AI and Web3 industries is 0G aiming to solve?
Michael: To truly own AI and build "infinite scale" AI on-chain, we need several key infrastructure breakthroughs.
First, performance: How do we create a Layer1 powerful enough to handle the heaviest data throughput use cases? In many cases, AI data centers process hundreds of GBs or even TBs per second. So we must design a high-performance system architecture.
Second, research: We need to figure out how to properly align AI, and understand how AGI should be conceptualized and built in a decentralized context. Will AGI emerge from one giant model, or from the combination of many small models? We lean toward the latter, but that demands significant research investment.
How do you want 0G to be defined in 5 years?
Host Blake: How do you want 0G to be defined five years from now?
Michael: Our ultimate goal is to bring Web2 AI builders into Web3.
When I envision 0G five years from now, I see it as the core hub for all mission-critical AI applications—such as manufacturing facilities, multi-robot systems, airports, logistics systems, and other societal-scale scenarios. In these contexts, 0G would provide blockchain-based security and alignment mechanisms to ensure safe operations. On this front, I want 0G to always be at the forefront, occupying a central position.
Of course, for individual use cases, people can still rely on centralized or edge-device AI. But when it comes to societal-scale, mission-critical applications, 0G must lead—and this progress must be community-driven, ultimately realizing AI as a public good.
What is the core architecture of the decentralized AI operating system?
Michael: As mentioned earlier, 0G's core is a layered structure, with each layer handling different tasks required to build large-scale AI applications.
This typically includes:
● Compute / DePIN Layer: Not owned by us, but integrated with compute resources from partners like Aethir and Akash.
● Software Layer:
○ Layer1 (Execution Layer): Infinitely scalable.
○ Storage Layer.
○ Data Availability Layer.
○ Compute Layer: Supports inference and fine-tuning, using TEE (Trusted Execution Environment) for verifiability.
○ Service Marketplace Layer: Similar to an App Store, open for various services and public datasets.
○ AI Alignment Nodes: Functioning like a "police force," scanning for model drift and harmful behaviors, maintaining system health through slashing penalties.
Each component involves deep architectural design. For example, Layer1 itself uses a modular architecture: modular execution, modular consensus, modular DA. Each layer is highly optimized and infinitely scalable, free from the constraints of traditional blockchain systems. While some L2s may cap at 500 TPS, our sharding approach allows deployment of any number of shards based on application needs, enabling arbitrary TPS scaling. This design philosophy permeates every part of the system.
What are 0G's performance advantages over traditional L1s?
Michael: Our core philosophy is parallelization. Whether it's storage, data throughput, or TPS, everything scales infinitely in parallel—achieving DA throughput of tens to hundreds of GB/s and TPS in the hundreds of thousands.
How does 0G attract a developer ecosystem?
Michael: One of our design principles is minimizing the barrier for Web2 developers to deploy on-chain.
Through modular design, infrastructure components integrate seamlessly. If you use parts of 0G, they work together natively, offering a one-stop experience—developers don't need to piece together solutions from 50 different places. Additionally, the ecosystem itself is a draw. Joining the largest Web3 × AI ecosystem provides competitive advantage and enables interaction within the ecosystem. We also offer extensive support: incentives, mentorship, accelerators, and investment.
Host Blake: So developers get mentorship support?
Michael: Yes, we offer not just mentorship, but also investment and comprehensive ecosystem support.
Are there early signs of a killer app?
Michael: Initially, since Web3 remains finance-oriented, early killer apps will likely be finance-related, such as:
● AI yield-finding agents
● AI looping leverage agents
● Other financial operation support agents
Over time, as the community participates in training and usage—providing compute/data and receiving token compensation—this too will become a killer app direction once infrastructure is ready.
What experiences can ordinary users have in the 0G ecosystem?
Michael: I mentioned some earlier—mentorship, marketing, and investment. Our goal is to rapidly build a large-scale ecosystem, attracting top builders to deploy their models, data, and compute. This requires heavy investment across all team functions. We see ourselves as a provider of competitive advantage—anyone joining 0g should benefit directly.
JT: Beyond typical public chain features, as Michael mentioned—like investment—we offer many extended capabilities. 0G’s overall advantage lies in providing not just on-chain functionality, but also compute, storage, and full-stack support for on-chain AI. Users or projects can access diverse use cases on the 0G platform, deploy fully decentralized AI applications, and so can ordinary users—something most other chains cannot achieve.
How to balance airdrop scale with long-term user retention?
Michael: This is a very tricky issue. On one hand, we need to recognize community contributions—like testnet participation or promotional support. On the other, we don’t want people joining just for short-term rewards and leaving after mainnet launch.
Our solution: Make part of the airdrop subject to vesting—for example, Yapper allocations or certain NFT rewards. This keeps people engaged with the ecosystem over a longer period.
We maintain a long-term mindset—because our mission is to make AI a public good. This isn't a one- or two-year project, but a long-term endeavor requiring sustained community effort and deep involvement.
JT: We launched an airdrop check and registration tool, giving one-time incentives to content creators like Yappers, while offering ongoing rewards via the Infofy platform—but only unlocked through continued contribution.
How will 0G avoid FUD similar to Movement or RedStone?
Michael: It's hard to completely avoid FUD. Everyone wants maximum airdrops, and no matter what we give, some will feel it's insufficient—that's the gap between perception and reality.
We've learned key lessons: Separating mainnet launch from airdrop/listing harms the community and creates unnecessary FUD. Distinguishing real/long-term users from Sybil attackers is difficult, but we're investing heavily in detection and filtering. We’ve applied these lessons to current decisions—splitting airdrops into different categories: social tasks/community rewards vs. testnet participation. These groups have different mindsets, so reward dimensions are carefully designed over time to attract more long-term participants.
For testnet participants: We hope they deploy on mainnet too, since mainnet metrics are real. Mainnet participation unlocks eligibility for community rewards—designed with a long-term mindset. Identifying real users vs. Sybils remains challenging, but we continue improving. Airdrops will be split, with multiple reward dimensions thoughtfully designed.
JT: We segment users: testnet runners, active community members, social media task participants, Yappers, early OGs, NFT holders. People earn rewards across different dimensions, but the core principle is rewarding long-term contributors.
How does the team ensure fairness and motivation for early supporters?
Michael: Beyond airdrops and community rewards, greater returns should come from long-term ecosystem engagement. Builders receive retroactive community rewards, investment, mentorship; regular users gain value over time through sustained support and shared mission alignment.
JT: We give higher weight to early OGs in Discord, and issued One Group Grivity NFTs (free or 0.1 ETH mint, later peaking at 1.8–1.9 ETH). Pre- and post-TGE has concentrated incentives, while long-term incentives shift to ecosystem projects. Wallet + Twitter + Discord binding allows multi-role participants to receive multiple rewards.
What role can the community play in designing incentive mechanisms?
Michael: We openly listen to community feedback as part of governance. The community can propose new incentive mechanisms and reward models. We'll establish a Security Council to adopt suggestions—this includes incentive design. The 0g Foundation previously adjusted its initial tokenomics quickly based on community feedback—we worked meetings until 4–5 AM to finalize changes.
JT: The community can play a major role in shaping incentive mechanisms. As Michael said, users can suggest what they'd like 0G to do, and the team adjusts accordingly. For example, after releasing an initial tokenomics model, we received extensive community feedback and rapidly revised it—the final result was well-received. That was a significant case. We even held overnight meetings, going until 4–5 AM, just to respond swiftly to community voices.
Key goals for the next 12–24 months
Michael:
Two main areas:
● Build the largest Web3 × AI ecosystem: Focus on trends like robotics, strengthen key infrastructure, explore new verticals and critical services, attract top-tier AI dApps and Agents to build on 0G.
● Match or surpass centralized black-box AI in infrastructure: Enhance verifiable mechanisms, support training of arbitrarily large models, further improve L1 performance, enable full on-chain migration of any Web2 component.
JT: Our goal is to make 0G a fully on-chain, evolvable AI ecosystem, reaching tier-one Web2 infrastructure standards within about two years—while maintaining transparency and decentralization.
Outlook on industry trends (AI Infra, RWA+AI, stablecoins, etc.)
Michael:
● Stablecoins: Major institutions are already discussing using stablecoin rails instead of banking rails. They will eventually integrate with AI Agents.
● RWA+AI: E.g., tokenized hedge funds + collateralized stablecoin borrowing + looping strategies, where AI Agents monitor interest rates and risks, auto-rebalancing portfolios.
● Robotics: Household and industrial robots will become widespread, but safety and alignment are critical—otherwise, hacking incidents could have severe consequences.
JT: Stablecoin usage is growing rapidly, with many institutions stating they’ll use stablecoins to replace bank transfers. Meanwhile, RWA + AI is a highly promising area. For instance, someone stakes tokens and borrows stablecoins; AI can help manage risk and rebalance positions before nearing liquidation. Your addition was already comprehensive. Let me add one point: Michael mentioned the Jackson Hole conference—it was an important event last month near Denver, USA, where many U.S. institutions and officials discussed integrating traditional finance with crypto, both policy-wise and for future development. Michael attended as a representative of 0G.
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
