Hello everyone. Glad to see you all here. I think you're ready for the session to begin. So, this is a workshop from Zero G. I'm Gatin, a senior developer at Zero G, and I'm really excited to be here in Lisbon.
Greetings to everyone who came from Lisbon. Before we start, tell me, which of you is from Lisbon? Just raise your hand if you can. Oh, we have about half the room. That's good.
So, yes. I would like to delve into the topic and this is my first visit to Lisbon. You are very welcoming here. So first of all, I want to thank you for this. So, Zero G is a blockchain for AI blockchain for AI agents.
This presentation is about what Zero G is, how to get started developing with us, what resources might be helpful, and I will answer any questions you may have. I'll take a little time at the end, and we'll also be on the stand today. So please reach out if you have any questions. Let's start with the problem. What is a current problem you are facing?
Before we delve into what Zero G is and what our stack is, remember: we are a stack is, remember: we are a blockchain for AI blockchain for AI agents. You get the ability to perform verified inferencing that is private, and all queries to models are executed in a trusted execution environment (TEE). This means that input and output data are verified, so the information does not reach centralized organizations. This is a quick overview to keep in mind, but I'll go into more detail in the following slides. Regarding the issues with centralized systems that you may have noticed recently: with the release of Fable 5, we saw how access to features was restricted and then reopened.
You're not even sure if it works as powerfully as it did in the original version of Fable original version of Fable . It may be restricted due to government requirements, or it may be closed altogether or changed tomorrow. This is a problem we face in centralized models now. The next point is a subscription, for example, for $200 per month. You don't even know if they will raise the price tomorrow.
For example, I'm from India, and Anthropic just introduced special pricing for India on Claude. So yes, you are not sure about pricing, you are dependent on what this centralized entity offers. Another problem is censorship. For example, you are conducting cybersecurity research, so you may ask questions that seem sensitive, even though you are not actually trying to blow anything up or carry out a large-scale cyberattack, but your request is rejected due to censorship settings in the models. These are problems that you can face every day, and not only you, but everyone who uses centralized AI faces this.
Delving deeper into these problems, the first of them problems, the first of them is opacity and lack of trust. Let's say you are using Opus 4.8 or any other model, you are not sure if you are really getting answers from that particular model. You simply interact with the chat interact with the chat interface, and the model gives you what appears to be Opus 4.8 gives you what appears to be Opus 4.
8 . You don't know if it's being throttled, slowed down, or what's going on "under the hood." This is a big problem that we call the "black box." Zero G provides verification just so you have a general idea. I will provide more details as we speak.
The conclusion you get from Zero G is evidence-based and has data references to it. The data you provide for processing is used only within this channel and is performed in a trusted environment trusted environment . Therefore, your data is not used to train models, and no one but you has access to what is being processed. You could say that your data input and output are protected from third-party interference. Next interference.
Next is binding to the supplier. All you can do is depend on closed APIs, random prices, and the rules are constantly changing. After all, no one reads these rules. You simply click "accept terms" when using one specific API. But with Zero G, you will be able to switch between different providers.
Zero G providers. Zero G is more of a marketplace. You can use similar models from different suppliers, depending on your location. For example, you can find much lower prices in Asia, so you can use this for low latency if you are in that region. And there are many different providers that can be used for the same model.
Next is the lack of property rights. For example, you simply create an agent. You do not own this agent. All you have is a name, a description, like a have is a name, a description, like a set of instructions for an agent. This is what you own.
You do not own the agent itself as such. But, for example, Zero G has something called agentic IDs. This means you will have ownership of your agent. You just...It's like an NFT.
You transfer this NFT, it's as if you are transferring ownership of this agent to another person. This is one of the things to consider when they say you don't have actual ownership of this agent. Then there is this agent. Then there is centralized control and censorship. This is a summary of what I have been talking about up to this point.
That is, several companies can decide who can create, deploy, or access AI. For example, a change in policy can destroy everything. But with 0G, it's a permissive system without any restrictions or "gatekeepers." That's why I called this whole presentation " permissionless blockchain for AI agents." And apps break when prices go up and you just have to change providers.
And changing providers means you have to sacrifice the quality of what you offer to users. For example, you might be using a very good model that was cheap at first , and then you have to change it because they raise the prices and it no longer gives you a good return on investment. But this will affect your app users, right? So it's a big problem when they dynamically raise or change prices. So yes, the 0G modular stack allows you to replace any layer without rewriting the entire application.
So yes, you'll learn more about that in the next slides. And so, this is it, yes, it looks like the platform is charging a lot of fees, and the real benefit for developers is much less. So it's about incentives. And with 0G, since...and I also forgot to mention that we are a modular stack.
So, inferencing is a separate layer, it's like a marketplace, and then we have a chain where there are many computing power providers. That is, there are many people who provide computing power, and there are many users who use inference. So you can get rewarded, for example, if you are a computing power provider, you get an incentive. For example, in the future we are focusing on fine-tuning fine-tuning . So if you help with fine- tuning, you can get rewarded for that too .
So, there are several levels at which a user can receive an incentive if the system is decentralized. So, yes. The change that is happening now, and that you may all have felt, is that open source models are becoming equivalent, or even better at some level, compared to centralized models. For example, Kim K 3, which was recently released, is quite good. For example, I use it every day.
I often write code. So, for example, Kim K 3 produces results that are comparable to Fable 5. And, for example, like GLM 5.2, DeepSeek, there are a lot of open source models coming out now of a certain level. Additionally, with Zero-G you will be able to get private routing.
For example, TEE TLS routing is used. So, the input and output, as I mentioned, is the provider, and you are the the provider, and you are the user. And when you interact with the model, it happens in reliable hardware. So you will also get a certification to know that the result you get was truly generated in the TEE environment. So, yes.
I talked a lot, I repeated TEE many times. This is a key part of the entire architecture. This is what I want to emphasize. Yes. So, before I delve into the architecture, is it clear now what problems you're facing?
Are we ready to move on, guys? Yes. Perfectly. So this is the architecture, as I mentioned, we are modular. So you can use individual stacks in your application or use the full stack in your application.
Let's start with blockchain. We are infinitely scalable. This means that depending on usage, we are currently processing 11,000 transactions per second and block finality is less than 1 second. And yes, we have storage. This is a big part of our stack that I haven't emphasized yet, namely the data warehouse.
For example, agents need memory. Models need weights and datasets. And when you create an application, users will need, for example, a good place to store your data. So, if you, for example, want to store assets, datasets, models, and vectors, we have decentralized storage. We also have the Data Availability (DA) layer, the compute power I mentioned, and fine-tuning and RAGs that will be launched on our platform soon.
In addition, we have a marketplace. There, providers register their computing resources, and users can use inference from these providers. And we also have what we call " call " alignment." We sold the matching nodes at the end of 2024, in December. These were nodes, we sold about 92,000 units.
So our community has ownership to make sure the models work without any external interference. So, they have some control over this process. This is what we gave to users— gave to users— ownership. So yes, this can be understood as trust being a vulnerable spot. The market is growing.
Trust is the main issue here. And yes, this is the same thing I was already talking about. We have blockchain, compute, and storage. Agentic ID—this is what I mentioned, it's an NFT tied to your model or your agent. You will be able to transfer it or have sole ownership of it.
Regarding Zero G calculations, one model worth looking at is GLM 5.2, which has 744 billion parameters. And she also has TEE confirmation. There is ZKML for cryptographic proof, which is stored and calculated on the Zero G blockchain. Regarding the Zero G storage, the main thing I mentioned earlier is cost savings.
This is about 80-90% about 80-90% cheaper than AWS S3 buckets. It's very cheap and I strongly advise you to try it out at a hackathon to see how convenient and profitable it is compared to AWS. Or any other centralized storage providers. Next, on the Zero G blockchain, we have excellent throughput for DA. Our network speed is 11,000 transactions per second.
We are effectively discontinuing support for DA, but you will still have access to it; just wanted to warn you that we are turning this direction around. The blockchain has a speed of 11 thousand TPS, and the block finality is less than one second. So, yes, agents need ownership. This is an ERC standard, it's called 7857, it's the agent identifier I mentioned. So, yes.
What products can be used on Zero G? So, we had the stack up to this point, and you would have to hardcode everything , but to simplify the task, we created a project. It's called pc.0g.ai.
This is the so-called personal computer. Here you can get API can get API keys compatible with OpenAI, which have a router function that allows you to use any number of to use any number of different models on Zero G. I think we're short on time, so I'll just speed things up a bit, but if you have any questions, as you know, we'll be here at the booth. Next up at the booth. Next up is the Zero G app.
It's a nice app where you can create your own apps. It's called app.0g.ai. Here we have Genome, we have a platform for launching agents.
You can run many different agents. We have a chat function. You just need to deposit tokens and you will be able to interact with any of the models we have. Chat, Genome, and agent launch pad. Next we have the main network.
We are well funded. We have $325 million and really great integrations. We have over 400 ecosystem partners, and we have 350, even less have 350, even less . We only go to hackathons. We have over 1,000 projects built on our foundation, and we also provide grants to developers who create products on 0G.
These are the notes on the agreement, and let's move on to the main part. These are different directions. We are offering a reward of $15,000. $6,000 will get you the best AI will get you the best AI product you can build on 0G. These can be agents These can be agents , applications, and I have provided some examples in the list of directions that can be found on the site.
The next direction is next direction is infrastructure. These are products that use artificial intelligence, such artificial intelligence, such as AI agents, and they are the infrastructure that people can use to deploy agents. For example, it could be SDKs, frameworks, and anything that people can use to build agents. So this falls under the infrastructure direction. And this direction, as you all know, was introduced this time at ETHGlobal—it's continuity.
You will be able to continue developing your existing project, and we want to be generous in this regard. So, we made sure to offer $4,500 for people who continue to develop on Zero G. It could also be could also be any of the projects you've created before. So, yes. This is a brief summary of the directions.
And one important thing: please join using the last QR code. This is a Telegram group chat for this hackathon where you can get test tokens. We've added our promo code there so you can also get a generous amount of test tokens. I just want to be generous to everyone here. So, please join the Telegram group.
If you have any questions, want to talk to us, or meet other people building at Zero G, you can do so. So, please join using the latest QR code. Yes, and thank you very much for being a and thank you very much for being a great audience. Before we invest time, do you have any questions? I think we have time for two or three questions, so yeah.
Yes, please. Uh, yeah, I think that's what I was referring to. We posted this on our website. Uh, let me find this. Uh, looks like I need to format this better .
So, you can just...These are a few ideas that we've listed here, so you can review them, as well as the qualification requirements for each of the different areas. They will be slightly different, as for one direction we may require proof of Zero G computing power usage, while for another it will be an architectural diagram or simply a workflow of your infrastructure to make it easier for people trying to build or use your SDK. So you can just look at these examples, and it applies to any projects you've built on Zero G before. Just let me know if you have any questions at the booth here.
Thank you. Anyone else? Perfectly. Great. Thank you very much for being a great audience.
Happy hacking, and let me know if you have any questions. Thank you very much.
Automatic transcript — names and jargon may be misspelled.