Hey everyone. I'm going to share with you all today. So, in this presentation, uh this is just like a short workshop. I know it's already been like a few days since you guys started. So, it's just a quick refresher on what we have.
And also, I'll give you a quick walk-through on the builder hub. And we'll also be having like a quick office hours so that we can address any questions that you have. Our team will also be joining shortly, so we we could also help you with the office hours. And please do share any questions that you have over there so we can address it together. And this can benefit from the same question as well.
So, before that, I'll just give you a quick dive deep. I think you guys should be much more familiar now starting off with once you since you've started building. So, it would have been a good refresher of the topics that you might have missed during your onboarding to Zero G as well. So, myself Gautam and developer at Zero G. I just this just a quick presentation on what Zero G is and what you can build with Zero G and a quick rundown on our infrastructure.
So. The the biggest problem that we are currently solving is the agent economy where like the trust is the biggest vulnerability. And there are like damages, whether it may be like exploits, whether it may be malicious attacks, or like the trust that we are currently putting in autonomous AI, which is causing a big backlash because the trust system doesn't work as intended as the all the existing AI solutions are completely centralized and that acts as a black box. And you don't get like a full context of what is happening behind. What is the censorship resistance mechanism?
Like how do you even trust the ex- outputs that are given by these AI models. So, there are like a lot of things that are like hidden from the users' endpoint. And it also is a single point of failure when it comes to like a centralized organization. So, in this presentation, I'll just give you like how Zero G is trying to solve this problem using our stack, which is dedicated for on-chain AI. What is Zero G?
Just give me a second. Yes. So, Zero G is the decentralized AI operating system which acts as like a foundation for the largest AI L1 chain. So, this is like a layer one chain for on-chain AI. So, we have a complete stack from like an infinitely scalable chain, which is EVM compatible.
And we have three core pillars on top, which is like the storage, the DA, the compute. And on top of that, we have the service marketplace. So, other than these, we have like few other features like giving agentic ID. So, where you can view your agent identity, which I will dive deep into in the upcoming section. And just moving on with storage and DA.
So, the chain acts like a base layer where all the transactions get settled. It is completely fast. We have achieved block time of like less than one one second. And for the storage, DA, and compute. So, the storage, you might question like why do you need a storage when there are other storage providers in the market?
Why do you need us like to have your own storage into the ecosystem? So, here the thing is that the existing storage providers are really expensive. And also, they they are not meant to handle the throughput of large-scale AI models uploading huge datasets of GBs and TBs of data onto like the uh and giving them like an on-chain verifiability. So, that is not what other chains or like other storage solution providers are intended to do. But here, with Zero G, we have our storage uh dedicated for you to upload large-scale datasets, models, assets, metadata, anything that you might need to train and use that use those models to get like verifiable inferencing.
Next comes like the DA layer. And with the DA layer, it acts like an off-chain data storage where you can just check for like proofs. Later like once uh for example, you're training some datasets and the model is released. And like a lot of people are using that models with verifiable inferencing. If someone wants to verify the history of like the training data, they could use this off-chain storage or like the data availability layer to get access of like the historical data.
Next is like the compute, which is like the cream on top of the cookie that we have. So, with this, you will be able to get verifiable inferencing from AI models. All we have, we call it like a sealed inference. So, the complete execution of like the input-output stream that you do a request for an AI model happens in a TEE environment or a trusted execution environment where like the input-output system is not tampered by any external agents. And now, the training part will be live soon.
Currently, we have like the verifiable inferencing live. And also, we have the fine-tuning live. So, you will be able to fine-tune the existing models that are available on Zero G. And if you want to like upload your own like use your own model, then you would require a H100 or H200 GPU. And using this service marketplace, you will be able to register your AI model and use it as a provider and get like the input-output usage or Currently, we have like open-source models like GLM 5.
1. We just uh had it live today. And other than that, we have like 23.6. We have GLM 5.
We have GPT OS. Uh we have uh 30C. We have uh 27B on the testnet. So, these are like different models. And we have a few other models as well, which you could find on the docs uh to get like the upgrade models list.
Moving forward, uh just give me a second. Yeah. These are the problems that are you are facing the existing uh the AI today. So, which I just gave a quick brief before. So, it's like a complete black box.
You can't get the data or the models. And with vendor lock-in, apps depend on closed APIs for pricing. So, you have like no idea what was actually inferencing cost and how much you are being charged for the same thing. But Zero G provides like open decentralized storage and compute markets that has no single vendor that controls your complete infrastructure. Next comes the ownership.
So, users don't own their agents or our open system. It is owned by the centralized organization. But here, with the verifiable AI like attestations that you receive, you'd have the ownership and accountability that that model performed as intended. And that also gives much more confidence for like the DApp users that you build using this AI inference. So, Zero G supports on-chain AI agents, NFTs, which we have recently rebranded to as agentic IDs.
As I mentioned, I'll just dive deep into the upcoming slide on this topic. And centralized control and censorship is another big issue when it comes to like centralized AI. This comes into incentives. Yes, this is the one thing like everyone are aware of like what the platform charges and what the builders will be able to like get as an incentive for like contributing to like whether it may be like training these models or whether it may be providing your own GPUs for like the computation. So, there's not a proper structure for incentives.
With Zero G, as I mentioned, we have like the four core components: the chain, compute, storage, and DA. That's it. I'll just keep just keep that in mind. And let's pretty start with the Zero G storage. So, just in this, so we have like an inbuilt block layer storage.
The other storage that we have, which we saw a lot of people having questions in the hackathon, is the KV storage. For KV storage, you can actually need to run your own nodes. But for other 51 for KV storage, you will be able to achieve it with the existing storage, the large storage that is available. And for that, you don't need to run any nodes. You could just start with the SDKs and starter kits that we have.
And with the upload of function, there's no requirement for flow contracts that you previously used to have. So, with these latest updates in the SDKs, which I will solve clear a lot of questions that you had in the office hours. So, with these latest version of SDKs, it's all about using the upload of function and download of function. And this is how the process goes. So, when you upload your data, so that data is split into multiple chunks using a a system called as erasure coding.
And that chunks of data are distributed into the decentralized nodes. And once these are uh like distributed, so that is one part that is like the data storage. And the other lane of data publishing happens like simultaneously. So, that what happens is when these nodes receive these data chunks, it signs that signs that it has received this data. And that signed attestation is stored on the Zero G contract.
So, that is like that forms up like a Merkle tree. So, the Merkle rate gives you the complete system of like how the data chunks are distributed in between these nodes and you will return back from zero G storage network. All you All you All that is returned back is like a root hash. So, with this root hash you will be able to like download the data uh back from the network. So, I currently we have uh recently launched uh encryption and you'll be able to get our encryption uh you you'll be able to encrypt your data before you're uploading it to zero G storage.
And this is a quick cost comparison that you could see with AWS [clears throat] S3 and with zero G storage and you could see that you'll be able to achieve like 80 to 90% of cost saving even with giving in the decentralization aspect into it along with the verifiable performance, tamper-proof, that is like a decentralized redundancy. So, for example, if node A is having chunk A, there will be a there will be another uh copy of that chunk in one other node. So, there will be a redundancy in uh it will avoid redundancy and you will also have like a fallback mechanism. So, even if a few nodes fail, there will be like an alternative node from which that some data chunk can be fetched. So, uh all the like elaborate details are available on the documentation.
So, please do check it out if you're interested to learn much more technical details before you start building. Uh yes, this is GLM 5 that is live and GLM 5.1 is live today. So, yeah, this is like the 50 and the 30. You can you can see KML proofs of like the that ZKML proof that output that you received was completely verified.
And uh next goes to like the zero G DA and chain. So, this is the DA throughput we have achieved of like 50 GBPS. Our chain TPS is 11K and for the block finality time we have given achieved like a fine grain release uh second block time which gives you the latency similar to you utilizing like centralized sources. And this is one interesting thing that you can utilize to build which is called as agent ID. So, with agent ID or like the standard that our team has developed which like the ERC-7857 uh is that you will be able to give like a own like an identity to your AI agents.
So, I'll just give you a quick down brief. So, why ERC-721 and 1155 falls short for AI? Because it's static metadata, you can't like that's incomplete transfer a transfer. So, just the ownership moves, there's no like connection between an agent and the NFT and that doesn't mean that when you're transferring an NFT, uh here with agent ID or like this ERC-7857 if you transfer an NFT, it means that you're transferring the ownership of the agent to the other person. Here with agent ID, you'll be able to encrypt the model data and other details of the model that is being used.
So, re-encryption on transfer. So, this is what I mentioned. That's like a T oracle that re-keys the metadata for the new owner and there's like T and ZK proofs. There's like 100 delegated users per token. So, without transferring like the NFTs, you can also like delegate like access to different users who can utilize that NFT without like completely transferring the ownership to that person.
So, yeah, these are like a few points of like how agent ID solves this problem. You could use this in your application when you can use this as an identity to your agents. So, this gives like a quick details on like how you can do this flow and uh Yes, this is a quick brief on our Aristotle mainnet uh as well. So, this is live since September 2025. Aristotle mainnet is uh like the main chain and for like the testnet we have Galileo.
And we have 3.5 million dollars raised from day one investors. We have 400 integration of our full stack and also I'm proud to mention that we have the largest ecosystem for on-chain AI projects and uh please feel free to let us know if you need any support from other ecosystem projects and we can help you get connected and we want you to grow as an ecosystem and build for on-chain AI, then you can get educated with zero G, you can understand how we can do it efficiently with zero G and you can utilize all the ecosystem partners' product and uh just start building with and make AI trustless and also decentralized. And one other thing that I would like to introduce is we have a ecosystem grant program called build on zero G and we have the 2.0 version that is currently live.
So, what you can do is once you're done building in this hackathon and you're like looking forward to get personalized assistance to scale the product for you to like launch in the market, whether it may be go to market support, whether it may be like uh development support, integration support, financial support, we will be able to like provide the necessary support for you post the hackathon. So, please I'll share the links once you're like once there's like the details for the hackathon. So, otherwise you could also check now it's build.0gfoundation.ai.
So, that is a good place for for us to like give like personalized support for you to build on top of zero G. And this is like the alignment nodes that you might have seen when I showed the architecture diagram. So, these are like the nodes that we distributed over a node sale to our community members. So, this gives uh access all the power to users to uh like have ownership or like be part of the governance of zero G's uh network in the future. So, yeah, this uh for like the development stage this has no impact right now, but please do keep in mind about like the alignment nodes that we have and you could even like have this to be part of the governance of the zero G chain.
And uh yeah, just give a quick dive deep on the price pool. Uh we have like two tracks it's halfway and you might have all planned what you're planning to build. So, by just giving a quick rundown, first one is 7.5K dollars, second one is also 7.5K dollars.
The first is for the framework and tooling, the second is for like the agents you build with the existing frameworks and how you leverage zero G with it. So, that is the core difference, but both through gives you like a complete generic space for you to build any idea that you have with any uh domain expertise. So, you'll be able to build that uh and And uh here are the ideas that were shared with these tracks. So, these are just like for uh a small hint or like the idea bank for you to get started, but by this time you might have explored on few other ideas as well and we got a few other interesting ideas that uh builders validated with us which seemed pretty cool as well. So, we are really looking forward to see you execute and build this and get this idea live.
So, can't wait to see the projects out and anytime if you need support with any MVP that you've completed and you want us to help get reviewed. Since it's an online hackathon, please do DM us with the demo link and we can review it as a team and we can give you some feedbacks before you go into the final demos. And yes, I would like to conclude by saying that uh this is the link to the zero G's build.com and this is one important thing most of you would have joined by now, but if you haven't, please do join the support TG channel. So, here you can get necessary support, whether it may be testnet tokens or whether if you have you might have any questions.
Uh also to know what any other small office hours that we might connect in future before the hackathon ends. So, please do join this TG chat and if you want to get in touch with me, it's and at NT both on X and Telegram. So, please do DM me with any queries that you have as well. Uh so, thank you so much everyone. Uh it was it was a pleasure giving this presentation to you all and I'm looking forward to jump to the office hours section right now.
Automatic transcript — names and jargon may be misspelled.