AMA | Behind the Launch: SIWA Open Testnet

In this AMA, Marketing Lead Joules Barragan goes behind the launch of SIWA, our first open testnet. Joined by CEO & Co-founder Sean Ren, AI Product Lead James Costantini, and Blockchain & Protocol Product Lead Jesse Guild, we discuss how SIWA serves as the public gateway to the Sahara blockchain, a purpose-built L1 designed to support a decentralized AI economy. Our team explains the vision behind the platform's full-stack approach, the importance of on-chain data ownership and attribution, and the roadmap for building a transparent, equitable, and accessible future for AI development.

Link: https://x.com/i/spaces/1yoJMopmeldKQ

Transcript

Joules: Okay, welcome everybody and happy launch day. I'm Joules with the Sahara AI team and today marks a major milestone. It's the public launch of SIWA, our very first open testnet, and the starting point for public testing of Sahara AI's on-chain protocols and infrastructure. Now, this isn't just another testnet. It's the first public step towards building the infrastructure for an AI future that's open, equitable, and accessible to everybody. This is just phase one; SIWA will be released in several phases, and we're really excited to go over our whole roadmap with you today. I am joined by our CEO and co-founder, Sean Ren. Our AI product lead, James. And our blockchain and protocol product lead, Jesse. 

For anyone joining us for the first time, Sean, I'd like you to explain what is Sahara AI, and how exactly does SIWA fit into our vision of building this open and equitable infrastructure for AI development?

Sean: Thank you, Joules. Welcome everyone. Sahara’s mission is to accelerate an AI-driven future. We imagine a future where you don't need to know programming languages or use customized apps to interact with information. You can actually just naturally talk to and interact with an AI that can empower efficient and innovative processes of using user-generated content. Sahara AI is trying to give access and resources to all these great opportunities of the AI-driven future by building a full-stack AI-native blockchain platform where everyone on any chain can create, contribute to, and monetize their AIs.

In this whole picture, SIWA serves as the first public gateway to the Sahara blockchain. The Sahara blockchain is basically the foundational infrastructure layer that is purpose-built to allow anyone to do registration, licensing, and monetization of their AI assets, such as datasets and models. This is a process that is entirely transparent and verifiable on-chain using our blockchain protocols. The SIWA testnet will allow developers, contributors, and any other users to explore and validate all these core protocols ahead of the mainnet of our Sahara blockchain.

Joules: Jesse, there seems to be some confusion sometimes in the community with this release of SIWA. What exactly is the Sahara blockchain? Could you go a step deeper and really explain how SIWA connects to the Sahara blockchain and how all of this leads to the mainnet launch?

Jesse: Yeah, that's really important for us to clarify. The Sahara blockchain is the core infrastructure we're building. It's a purpose-built L1 that's designed specifically to support our decentralized AI economy. It's being optimized for registering, licensing, and monetizing AI assets, like datasets, models, and agents. SIWA testnet is our first public instance of this blockchain. It's an EVM-compatible testnet and the first public access point into the Sahara blockchain ecosystem. You can think of it as the initial on-ramp for developers, contributors, and partners to start to test drive the Sahara protocols and help us refine our system ahead of our mainnet launch.

On SIWA, you're able to interact with our global asset registry. You can upload and register datasets, and mint ownership NFTs that represent ownership of data. You can also begin to form the attribution of how you're training AI assets using our tagging system like "trained on" or "derived from" tags.

Joules: So it sounds like we're starting with data. Sean, why start with data? What makes dataset ownership and attribution the logical first step for this ecosystem?

Sean: A good analogy would be to think about the human brain as the AI, and what is powering that? It's food. That's what data is for AI systems. No matter if we're talking about an agent or an application, data powers all of the AI systems. But where do the contributors of this data get in return when their data is used to power AI systems that create all sorts of revenue and benefits? Today, contributors rarely receive credits or compensation.

We want to create an end-to-end user journey where when data is created and accumulated, we have this provenance layer that can be used to trace the ownership and contributions. This trace can be used downstream. When these datasets are used to fine-tune a model to become a powerful, specialized model, or when this data is merged into a retrieval-augmented framework to build an agent, all these applications start getting paid from their usage. That money is supposed to flow back to all of the data contributors. This can only be done in an autonomous and transparent way if we have this ownership and attribution trace, and that's what the Sahara blockchain protocols and network are trying to power. Simply put, if you can record and prove who created the input, you can share the value of the outputs.

Joules: Jesse, could you speak a little bit about what makes our protocols different from what's already been attempted in Web3?

Jesse: A lot of the protocols in the space have stopped at the service level, where they allow you to hash a file and upload a metadata pointer. I don't think they go as deep as what we're doing with Sahara, which is creating this connective tissue that allows you to track not only how the data is used but also helping you enforce revenue-sharing policies for downstream reuse for creators and builders. We're creating a full-stack system that handles the registration of data, models, and agents, attribution tagging that allows you to track the provenance of those assets, and a suite of smart contract tools that allow you to enforce licensing terms and also apply royalty policies. We're also creating staking pools, or what we call vaults, that allow you to stake contribution NFTs and earn passive revenue streams. All of that flows back to your account automatically. I'd say the key difference between Sahara and other protocols is really the depth and the enforceability. Tokenization is really just the beginning of what we're trying to achieve here.

Joules: A lot of what we're building has this on-chain component. James, could you tell us a bit about how that works? What's the process for uploading a dataset? What formats do we support? What does a successful submission look like?

James: When you log into our system, you can upload through our developer platform. Right now, the datasets we're supporting are JSON, CSV, or text files. We have a UI that walks the user through the steps of uploading the data and the metadata. We then figure out if it's structured, labeled, diverse, and valuable for training and evaluation.

Joules: What is actually stored on-chain versus off-chain during this registration process?

Jesse: One of the key design decisions of our system is to avoid storing raw data on-chain for a couple of reasons. One is cost and scalability. The other is that we want to allow for selective disclosure so that data can be made private, gated, or selectively licensed. When you're uploading the dataset, it's not the raw data itself that's being written on-chain; that's stored off-chain via our platform. On-chain, we store the metadata, which includes the format, data domain, and licensing intent of the dataset. It also creates a cryptographic hash of that dataset and the attributions that describe the relationships between other datasets. This is all encoded into an ERC-721 token that represents ownership of that particular asset. All of that gets written to the Sahara Global Registry, and that serves as the authoritative on-chain source of truth for all of the AI assets in the ecosystem.

Joules: From a product perspective, James, what kinds of datasets exactly are we hoping contributors will register?

James: We're looking for very high-quality, diverse, and domain-specific datasets. These could be Web2 and Web3 domains. We like to place an emphasis on underrepresented or high-impact areas for our users to be uploading datasets.

Joules: Sean, do you have anything to add to that?

Sean: A good dataset is really customized and sometimes personalized to a specific use case and application. It's often a conversation between the demand side and the supply side. We see the data services platform and the Sahara AI platform as a place where we can bridge people with the needs for good datasets and help them satisfy those requests.

Joules: So SIWA is more than just a one-off release; it's part of this larger phase rollout. Jesse, can you walk us through the full roadmap from Phase One to mainnet?

Jesse: We're going live now with this initial release, which is the UI, dataset registration, on-chain ownership token minting, and the ability to index the dataset metadata on-chain. This is creating a precursor to what we're calling our attribution and royalty graphs. This is about bootstrapping the base layer of provenance on our blockchain and laying the groundwork for monetization.

• Phase Two will release our licensing protocols, which is the ability to license both datasets and agents, and the creation of contribution NFTs that allow you to stake your NFT and have licensing and royalty flows routed to your account.

• After that, we're going to start to open up the participation in the network to be more permissionless. We're going to fully open-source our protocols, and we're also going to have the ability for people to deploy agents on-chain.

• Then we're going to start to build out the functionality around our agent tokenization, so support for the mobile contacts protocol, MCP servers, and multi-agent workflows.

The end game that we're really working towards is a fully composable AI economy. By the end of our initial testnet and leading into mainnet, developers will be able to have full licensing and royalty capabilities for datasets and agents. We are looking at Q3, maybe early Q4, for mainnet, but that's still TBD.

Joules: We also have our broader stack, right? Our data services platform, our AI developer platform, our AI marketplace. Sean, can you give us a high-level overview of that full stack and how those all interoperate with each other?

Sean: I see this whole stack of Sahara AI as two major layers. The underneath is the Sahara infra layer, which consists of the Sahara blockchain and its protocols. It's in charge of all the major features like asset registration, ownership, attribution, licensing configurations, and revenue-sharing mechanisms. This is basically what our SIWA testnet launch is trying to cover.

On top of the blockchain and protocol layer, we are also actively building what we call the AI execution protocol layer, which is chain-agnostic. The purpose of this protocol is to bridge the off-chain AI compute, such as model inferences, model storage, and model training, and the verifications and proof of those execution steps with the on-chain components. We generate those proofs in a very compact form and then push those proofs onto the blockchain so we have the entire trace of the contributions of what happened for the AI execution.

We are building on top of this a layer that we call the application layer, where all of the user-facing products you guys have been playing with happen. Sahara Platform is building a suite of flagship applications on this layer, including the Data Services Platform, the AI Developer Platform, and the AI Marketplace. There are very organic interactions among these three products. For example, if you're a developer coming to the AI developer platform to orchestrate an agent, you might go to the marketplace to purchase some dataset access. If you don't find anything you like, you might create a project on the Data Services Platform and get the datasets you want. Then you come back to the rest of the user journey for your agent orchestration. You grab the data, you grab models from the marketplace, or you upload models from your local machine, and then you compile the agent. Now, with the agent compiled, you can specify what the ownership composition looks like and what the revenue-sharing mechanisms should be running. You can publish this on the marketplace and start monetizing your agent.

That's how all of these platform products connect together. We are also trying to welcome our ecosystem partners and participants to help build more of the important tools and applications that fit into this whole user journey. We need AI-powered apps and agents, more foundational toolkits, and even DeFi apps that power our protocols to unlock new financial primitives on the application layer. This is an overview of some of the flagship applications we want to bootstrap the ecosystem. More importantly, we are welcoming everyone to come to build tools, applications, agents, and DeFi infrastructures to power the entire ecosystem. 

Joules: Sean, you bring up a really good point too that the entire platform is chain-agnostic by design, right? Anybody on any chain, Web2 or Web3, can build using our platform. Why was this important when you were developing the platform, and what does this mean in practice for AI development?

Sean: That's a great question, and really something I want to highlight. What that means, technically, is that developers coming to the Sahara platform can build on any Layer 1s or off-chain, like Web2 contexts, by just plugging in our APIs and coding our protocols. The context today is that we have a lot of great developers sitting on different chain ecosystems—friends coming from the EVM chains, friends coming from Solana, friends coming from other Layer 2s—and they've built great agents and assets already. In the process of improving those AI assets, they might want to get newer versions of datasets from Sahara's AI Marketplace. They might want to use the better agent orchestration tools that are on our AI developer platform. They might need to create new datasets using our platform. We don't want to give them any hassle of doing that.

So, we are providing the best user experience possible to just help them streamline the whole lifecycle of getting the right AI resources and assets, using the tools to do better AI development, and trying to make this process as frictionless as possible. At a high level, it's called chain-agnostic, but at the user experience and UI level, we don't want to slow them down by interacting with Sahara blockchains or the protocols. We're achieving interoperability to expand the access and lower the adoption frictions. The general principle is that we are not trying to compete with other Layer 1s or even Web2 AI infrastructure tools; we are trying to support developers everywhere.

Joules: Awesome. Speaking of supporting developers, we launched with 40 new partners across AI, Web3, cloud, and research. And if you're building something cool and you want to join our ecosystem, we've launched the Sahara Incubator to take your projects from MVP to market. Can you speak a little bit more about what we're doing with the incubator and what it offers for developers and founders?

Sean: The Incubator program supports all the early-stage AI projects building on the Sahara ecosystem. The concrete way we provide this support is through go-to-market support, engineering help with our great designers and engineers, and mentoring about both go-to-market strategy and product development. You will also have access to our great array of investors who can give you lots of insights about the market and your product. The applications for the incubation program are closing in roughly a month, so make sure you apply if you're interested. I think the website is saharalabs.ai/incubator.

Joules: We have a question from the audience: "I'm a little confused. Do we understand who will use this data? Who am I trying to provide useful data for? Developers, users, your AI system?"

Sean: I would say developers are definitely a major user group for consuming the datasets. They have brilliant ideas about what kind of AI or agents they are trying to build, and they don't have access to the best datasets for powering that AI yet. So they will basically come to the Data Services Platform, initiate a project, and anyone can see that project's description. They can try a qualification task, and if they pass, they become part of the contributors of the dataset and, down the road, part of the shareholders of the datasets. What's really exciting about Sahara is that when the datasets are continuously used by hundreds, or thousands of developers, you also become a shareholder of those AIs and those agents. When those agents and AIs start monetizing and getting revenue flows, you will get your own share as well. That's the ultimate sort of equitable and transparent monetization flow that we are working hard to make happen on our product.

Joules: We have a lot of people asking if users will get paid for providing datasets. Can you speak a little to the difference between providing a dataset and completing data tasks on the Data Services Platform?

Sean: We're seeing multiple ways for people to monetize their knowledge and efforts through data. One prominent route is completing data projects on the Data Services Platform. In that process, people get paid a certain rate for each data point they complete, and on top of that, they may also get revenue shares from that dataset that is continuously going to help to monetize their effort. Now, on the other route, you can upload what you think is a valuable dataset through the current SIWA Open Testnet user journey. Those datasets will be examined in an automatic way and, once qualified, will be published on our AI marketplace down the road. Once your datasets become visible on the AI Marketplace, other developers or users will come to explore your datasets and purchase the license access, which is another way you can monetize your data.

Joules: Another question we keep getting is: is the SIWA testnet just for developers? How can non-technical community members interact with SIWA?

Jesse: That's a good question. It's definitely focused on developers, but you don't have to be a developer to start to interact with SIWA. Our development portal, much like the Data Services Platform, is geared towards an audience that is interested in AI. So anybody can really go in and upload a dataset and mint an ownership NFT. We've created a really smooth user experience, so we've abstracted a lot of the complexity away from that. Even once we have more of the development tools built into the AI platform, things like our agent builder are going to be extremely intuitive. So we're trying to make this as accessible as we can to a semi-technical and non-technical audience. But once SIWA goes open source and the protocols are fully open source, we're going to be really focused on bringing on more Web3 developers to start to interact with our blockchain protocols and smart contracts.

Joules: How does Sahara AI ensure data contributors actually receive fair credit and compensation in their decentralized AI development?

Sean: We are recording everyone's contributions on the Data Services Platform on-chain. If you are a participant in finishing a dataset—let's say you finish 5% of the data points—then imagine you are taking a proportional percentage of the revenue share of that dataset when it's going to be listed on the AI marketplace. When other people purchase licensing access to the dataset, either in a pay-as-you-go manner or a monthly subscription, the dataset gets paid, which means you also get paid. That's a big route of monetization. Everything is recorded on-chain, and everyone can audit it. What we need is just to set the rule when you start configuring the licensing and revenue-sharing mechanisms of the datasets, and those features are coming up in the next release.

Joules: Are there any mechanisms in place to discourage uploading low-quality or scraped data?

James: Eventually there will be. At this point, it's a lot of human moderators, but eventually, at some point, we will use both AI and traditional moderation techniques. Things like contributor scoring, moderation tools, and reputation systems will be added to the platform. These are issues that are also prevalent in Web2, and we're going to use some combination of ecosystem maturity and reputation and moderation in combination with AI tools as time goes on.

Joules: Excellent and I think that puts us at time. We will have additional AMAs throughout the week so keep a lookout for those. I'll be posting more information about that later today and thank you to everybody else who has come. Do you have any last words? Sean, James, Jesse, any shout outs you want to do? 

Sean: Yeah, definitely shout out to the team for really making this launch happen. It's a collective effort from our AI team, our blockchain team and our marketing team for sure, ecosystem teams and I also want to really express my appreciation to our users, our communities and this is the one of the first opportunities that we really open up our product to the public. Anyone can experience what it means by claiming the ownership of your data assets on our SIWA testnet and also familiarize with some of the features that's coming up on our AI developer platforms. So hope to see you giving more feedback and comments about our products and really excited to have you in our ecosystem. 

Joules: Awesome. Thank you everybody. We weren't able to answer all the questions today, but I will go in and I'll respond to a few of them as well throughout the day that I see. So if you have any other questions just comment below and I will get to them. If you haven't already, definitely visit our developer platform. It's all over our website but it's app.saharalabs.ai/developer-platform. We might come up with an easier way to access that, but that's what it is right now and then just upload your dataset, mint your on-chain ownership token and be a part of shaping the next phase of decentralized AI infrastructure. Thank you everybody for coming. 

Sean: Thank you everyone. 

Jesse: Thanks everybody. 

James: Thanks guys.