Create Data Coin and Agent with Lighthouse I Nandit Mehra

[Music] Thank you for having us here and glad to uh join all the attendees. Um so yeah so the today's session is about create data coin and agent with lighthouse. We'll start over by giving an intro about what lighthouse is uh followed by our new product 1 mb.io IO and then my colleague Akash he will take over the screen and then uh walk us through some live examples and workshop which will be really useful for you all to build your hackathon um projects and uh what are we solving here at lighthouse primarily so while building lighthouse we realized that storing important data especially AI and blockchain history long-term is very expensive and especially hard to verify Another thing that we realized that when you want to store your data on decentralized web um there's a lack of security and that just eliminates most of the AI use cases out there. And third is fragmentation like when you are new as a developer or even while you're working on the startup you want uh like right now the solutions that h uh that were existing before lighthouse they were fragmented that we had to set up our own node you had to do storage and then pinning on your own side and then manage your own encryption and that's how like u how lighthouse solves all of this is first of all we have our permanent storage what that allows you to do is you can pay once and store your data long term be it images, videos, text, JSON files, etc.

Uh that could be um data like um NFD images that you are launching for NFT collection. That could be the AI memory and data that's part of your uh AI app. Um and then we have our access control which is powered by threshold cryptography that I'll talk more in a while. Uh and then a unified SDK like through one single Lighthouse Web3 SDK you can um interact with all of these networks store upload and encrypt data onto our uh web uh and for users we have our web app and CLI and SDK for developers as a primary entry point to use Lighthouse. Um and then uh I'll just touch this a bit part u on our permanent storage side we have an onchain smart contract pool of funds.

So when you pay for uh your data uh this onchain endowment pool pays the miners further in filecoin and in USDC and make sure that you don't need to worry about renewing your data long term and your data is stored uh forever. And talking about the encryption and access control part, this is a very interesting technology that we have here which uses threshold cryptography. What that allows you to do is uh every file that you upload to Lighthouse gets its own encryption key and then that encryption key is sharded across five different nodes and you uh as a user you just have to get uh the three out of five nodes to verify that you are the correct user and then the file can be decrypted. So this solves primarily two problems. One is that you don't need to do key management on your own side.

Let's say especially the problem comes on okay when you want to share files how do you give the encryption key to the end user and when you want to retrieve the files how uh does the end user who has the access gets those uh encryption keys and without losing those keys um our infrastructure manages all the keys shards them and then aggregates it for you and the other problem that it also um solves is like a centralized trust problem where Your key is not plainly sitting around on a centralized server as mostly happens in the case of uh the cloud providers but in our case it's encrypted but more on over that it's also sharded across these five uh nodes that we have and then you can token gate um the files that you upload with any onchain logic let's say I can say that um I want only pudgy penguin nfd owners to view file or someone who has thousand a tokens on base or polygon on they can access the file. So you can tie uh the decryption of the file to any onchain logic be it NFT ownership, token ownership, it could be time lock and you could also build logics like let's say you want uh some folks to pay you in certain tokens and then the file to be unlocked. So those kind of uh onchain scenarios can be created with our encryption SDK. So already we are working with some of the top uh web3 and AI teams out there who are using lighthouse for all different bunch of use cases. Just to highlight eternal AI is one app.

Uh what that allows you to do is uh they have open-source models like llama, deepseek, Hermes 4, all of them already stored on lighthouse and uh their uncensorable AI chart system uh can be accessed from anywhere and then you can access and deploy these open-source models but then all of the data is coming from um Lighthouse. Then we have ocean with their data marketplace uh who have already integrated lighthouse for their users. NFT storage is one great use case where a lot of NFT users and customers they are uh using NFT storage to store their NFT permanently and that is uh stored on lighthouse underneath. Uh and then on the deep inside I would like to highlight web hash which has uh tens of thousands of ENS websites. So folks would connect their ENS do uh domain uh to web hash front-end deployment tool like Versil uh and all of that data like tens of thousands of sites are getting uh served from Lighthouse right now.

Um and yeah during building lighthouse we realized that we wanted to uh give something um to the users more because we were able to analyze and get these insights that um developers and startups are not just using lighthouse to store data but also retrieve it very very frequently. they wanted some way to monetize that data and also uh make sure that um these AI models which are now coming up especially like the ones that big tech is making right now let's say charge GPT Gemini that most of us are already are very aware of they these AI models have already read most of the public data out there like they have already read YouTube read it all the public data out there and the big unlock uh that we see could be possible in uh is Then the private data. So and that private data right now it's siloed. It's trapped in consumer apps. Let's say for us it could be apps like Facebook, Uber, Zomato, Spotify etc.

Like all of the apps that we use on a daily basis. Um our data is trapped in those consumer uh apps and when they are monetizing that data for AI the end user is not benefiting from it. Uh so that's why we ended up building this new platform called as 1 MB.io IO which essentially allows you to not just store and encrypt data on Lighthouse but now you can launch uh what we call as these data coins and data agents and reward end consumers and even AI agents to aggregate launch um and bring this data together and like the folks on the other end of the spectrum who want to access the data they could pay you uh in these data coins uh you can tokenate these data coins with um the encryption logic that I shared about uh and then build interesting kind of cool applications. So how this data tokenization platform works is first of all um you create a data coin um you can deploy it on one of these test networks like Ethereum Sapoleia or on public networks like polygon base or world chain as well and more of that Akash will show you in a while and then you can easily upload and encrypt the data and then the data coin creator could set up logic to mint data coins and reward the data contributors and then they can be easily be traded in a DEX like we already have an integration with unis swap uh which is live.

So how the architecture overall looks like now is on the base layer we have lighthouse with our perpetual storage that you can pay once store data forever on decentralized web along with our encryption and token getting infrastructure and on top of that in the middle layer we have 1 mb.io IO which is our latest product where you can launch data coins, monetize users uh to contribute their data and then um like trade them easily on a DEX that we have already built for you. Then we see that more than hundreds of different variety of data sets could potentially be created. For example, folks could create uh data coins where um you could reward people with bringing in their chat GPT or Claudia data uh using a concept called a ZKTLS um or bring users data from consumer apps like Amazon uh royalty rewards like Starbucks, Marriott etc. And then users can bring in their data without uh revealing let's say their private information like name, phone number, email address uh and then all of that data aggregates together into a data coin that you created you can earn from day one.

But more importantly it can be traded and it could be sold out in a marketplace. And this is happening like right in front of us. If you see like big tech is already striking hundreds of millions of dollars of data monetization and content deals and this market of data monetization is just growing uh right in front of us and um in web 3 as well with crypto rails we are seeing projects like hivemapper helium etc building economies where at the end of the day they have a way for an end user consumer a soloreneur I would say to contribute their data and earn from the network and we just want to um change this game with uh Lighthouse and 1 MB where you can launch these data coins and rather than the current scenario where 0% of that benefit is going to the end users now users could actually have a way to monetize uh their data. So what people are already building using 1 MB.io uh right now and you could also take inspiration from some of these apps.

uh the first two on uh I think the ride sharing dow is some someone is already building that and even the food delivery dowo. Uh so just to highlight for this hackathon we have two tracks one is um that you build best consumer data coin and another is best data agent created. So on the best consumer data coin just as an example um someone previously created an app where it lets uh users to connect their Uber accounts bring in data through a privacy manner with ZKTLS using a protocol like reclaim protocol or opacity um and then all the contributors who are sharing the right history with this with this sharing dow they rewarded in some UBZ token that the data coin creator ated in another example which the app is called Pentry points which is a food delivery DAO. Uh their users are giving in their Zumato data and getting rewarded for um giving in their data with these data coins and just to highlight uh some of the uh data agent ideas that are being created. uh miles is something that my team member Akash will demo you today where essentially uh now the thing is not just you can have consumers given their data um and then reward them with data tokens but more importantly you can have AI agents itself produce synthetic data useful synthetic data just imagine let's say if I have the data from hive mapper about uh US roads that I know that okay this is how uh traffic is moving there and I just give it some prompts that maybe in India uh some areas which is uh not so great have some potholes or have some livestock moving around not just not to generalize but sometimes that happens uh but then I can use the data from a real uh road in some other country and give it some prompts and get AI models to generate um some synthetic data which I would verify uh and then uh produce more and more data over a period of time.

So let's say now a new mapping or let's say self-driving company has to come what they what they have to do is potentially like a startup or a developer they don't need to run like their CPU or GPU over uh like one month two month three months period some app like miles would already exist but terabytes of data with CPU GPU running for like tens of days with already ready synthetic data that could be sold to uh 10 such uh like startups and companies. And there are more examples like prom farm as well that could bring in and reward not just the end consumers uh like in case of consumer apps but also uh AI agents itself um to produce synthetic data that could be consumed by uh AI apps because if you see like with Alibaba uh who launched their like new model recently one of the core contributors of their success and the benchmark over chat GPT was they heavily relied on synthetic data uh and I think uh like our platform 1 MB could be really useful on unlocking rewarding uh and launching these data coins to uh bring in consumer data together and also um launching these AI agents which are able to produce really uh useful synthetic data coins uh with crypto rails. Cool. Awesome. I think I'll stop sharing my screen here and I'll invite my team member Akash to maybe share his screen and let's go through some uh live demos to give folks idea on what can be built.

Yeah, sure. Thank you. Uh I'll just be showing you how to like from the start how to upload the data on the lighthouse. and I'll walk you through the uh data fine making process and then we'll see a demo application uh that is miles that has just talked about for seeing how the data data point is being used here. So this is the files tab.

The all the process start here. You can just login with the metam mask and then you can just click on upload new and you can upload a file like that. Okay. There's also an option to upload encrypted file. You can just turn on the encryption and then you can just upload another file.

So what happens with the encryption is that uh your file is being first encrypted on the browser first and then it's being pushed to the IPFS. So your uh private data doesn't leave your browser. So if you want confirm it will take some time to encrypt and then uh threshold cryptography kicks in and your encryption key is being sharded into multiple charts and then it's stored on the lighthouse decentralized nodes. Okay. So the file has been uploaded successfully.

Now you can create an API key for your programmatic use. Uh let me create another API key for e online. So we don't store your API keys. So uh it's better you copy the copy the API key and store it somewhere safe. Now let's just uh walk through the docs to see the programmatic use.

Uh first is the upload file. You can uh use our JS SDK as well as the Python SDK to upload files encryptedly or uh non- encrypted way to the lighthouse. This is uh this is the way to upload the file non- encryptedly. You can use lighthouse or upload and then give the path and the API key. That's it.

You just need to give these two inputs and your data will be stored on the lighthouse. Then the next one is the upload encrypted. So you just need to call the lighter.load encrypted pass the phonic parameters and then your file will be uploaded encryptedly. You also have encryption features such as share file and revoke file.

So share file gives your file access uh it's it's gives any wallet address that you want to provide access to your private file. It gives it gives that user the access to the file and then you can also revoke the access if you like to revoke. Uh as Andre talked about the token getting conditions. We can also apply token conditions with the help of this documentation and you can also check the condition with the help of check access condition.get access condition.

Uh this page shows you all the supported networks. We support all the major chains and all the test nets and yeah uh now let's get to the data coin part. So this is the docs that that you can follow to create a data coin. We'll just head over to 1.io and I'll I'll show you a sample data coin creation flow.

Let's first connect my wallet. It's already connected. Uh I click on data uh create data coin. Let me give it a name. So as a convention, we keep the data agents that we create.

We start them we start their name with synth uh synth dash maybe let's just say fraud for fraud fraudulent data agent or uh if you want to upload a real data dh for for the real users you can use like any name like pop pop pop rewards or maybe find points whatever you want let's give it let's give it synth broad uh for the name let me give s n r sr You can upload a icon on a banner image. I'm skipping it for now. You can now then give any website. Then you can upload the all the social information here. Let's skip it for now.

Now the main part you have to select the chain. We currently support four major chains. Uh first one is the seapia test net and then the base polygon and the v chain. Let's use seapia for now. Okay.

So give you so we give you already pre-filled data which you can change and also if you want you can go with these with these pre-filled ones also. Let's keep the vesting period as 12 months. And for the locking asset I'll choose I'm choosing USDC for now. Uh let's just say 5 USDC because I have low coins on my destinate wallet. Okay.

So I have to run submit data. It's asking me to confirm the transaction. So a live data point is being created on the test net right now. Let's just wait for a few more seconds. Yeah, that's it.

So this is the token that I have just created. You can see all the token information here. Okay. Now let's just move to the sample application part. So this is the miles that I have created.

Uh my is a AI augmentation engine that helps uh that helps you create a synthetic data on the top of real world AI data. So let's log in and see the entire flow and click login to metamask. So what let's so what mice lets you do is mice lets you uh upload any real world image or any real world maps data and it augments data on the basis of the on the basis of the fields that you select right here and it augments the image on the basis of that and generates a data set with the AI with a synthetic data set so let's choose a heavy rain for here and I select maybe okay so my CPU have generated these augmented uh images for the data set and uh the mice will be rewarding me one data coin for the uploading of a single image. So done. I have got the data coin.

So these are the transactions. Here's the rewards page also. Here I can see all the rewards and if you if you want uh if you want to explore the data sets, this is the page where the users can go and like explore the data sets. If you want to like purchase a data set, you can ask the uh creator of the data set to maybe allow you and you can purchase the data set from here. This is the another application that we have created that's a pantry point that allows your zomato orders to be approved in a privacy preserving way using zls and lets you reward on the basis of that.

Uh maybe I'll just give a walk through here also. So this is the main dashboard where you can add the data. So all the data is being currently provided through the reclaim protocol. So if you scan this QR code, you'll be redirected to a page where you can log in with log into your uh Zomato application and then uh we'll fetch all the orders with the privacy preserving way and you will and we'll give you rewards on the basis of that. So yeah, that's pretty much it.

I think that's the main examples.

Um cool. Awesome. Um that sounds good and uh happy to answer any question if we have some time for that.

Sweet just jumping in here for Nand and Akash. Um yeah folks uh we are open for Q&A. Um and I see someone has raised their hand. Rome reactor. Uh uh Rome reactor do you want to unmute yourself and shoot your questions?

Hey. Hey guys. Um can you hear me fine? Yep. Loud and clear.

Okay. Um, yeah. So, we're creating basically an AI project with ASI where the U agents are going to be able to it's a protocol for the U agents to pay each other. So, like think of it like you ask Siri to book you a restaurant reservation and then Siri goes talks to other U agents that are restaurant agents and then they're able to finalize a transaction and book it. And we're storing the data in Lighthouse.

some of the data. Now, my question is to you guys um uh to be eligible for the uh bounties uh is it just uh fine just to store the data in the lighthouse or do we need to create like a the data coin specifically? Do we need it or um how would you guys advise uh just so we would be eligible for those two bounties because I know it was centered around the data coin and data.

Yeah, absolutely. uh first of all I think it's a very interesting use case that you're building around agents communicating with each other and having that commerce element along with it as well. Um so for us um like the two bounties are around creating the best uh data coin agent and also uh like best data coin and best data agent. So maybe in your case uh what you could potentially do is while you store data on lighthouse and then uh that data could potentially be encrypted uh if that data is useful and then those agents could pay uh in those data coins to each other. So yeah like we are looking for not just uh storing data on lighthouse but go one step further on launching a data coin on 1 mb.

io. IO and that being like a important element of your overall hack. So obviously you could use some uh things that ASI already have again that's always good to do like combine the best of everyone uh but then not just store on lighthouse maybe implement some data coin uh logic there that could be used to token get the data pay for that data that could be I think one step further um useful. So whoever does this does that best would have like a good chance of winning than just storing on lighthouse.

Okay. I I see. So just uh just to clarify um so we could be using the uh ASI like the the meta that they have with uh with dark met or whatever to store some of the data there that the agents need to just query. But then like we can create a data coin with Lighthouse and uh let's say like a partner wants to look at our protocol and get all the data for like all the order statuses and whatnot. We could encrypt that data with uh Lighthouse data coin and then to access that data um we could create like a dashboard of some form to access that data for for clients and then they would with that data coin with that access to the data coin they can retrieve that data.

Correct.

Yeah. that uh that makes sense. Obviously like the end idea has to make sense as well. Uh like some of the ideas that we just demoed you was the one uh the first one was uh AI itself generating more important valuable synthetic data and that having some end use case and the other one was where uh an end user is able to reward uh consumers to bring in the apps data that they are already using. So I could reward let's say Uber, Amazon users to bring in their individual data and then reward them with this uh data coin.

So uh while what you are saying obviously makes sense you could do that. Uh but just make sure like it just fits into like a end flow on why uh the data coin is there and what that data coin is unlocking. Um I think that would just give you a edge.

Yeah.

Got it. Got it. Thank you so much um push that information to the team. Hopefully we get something exciting for you guys. Share it.

Yeah. And there's another question in the chat that I can see from Lee was asking were the rain images generated by an AI model from lighthouse. So um we don't provide like AI model at lighthouse. What that example did and happy to share uh that materials in our discord channel at globals's discord. uh was that you could bring any AI model like that could be like something that you are running locally or in that case it was actually deployed on a cloud uh like a simple basic AI model or it could be like you could link it to chat GPT or any other open u like APIs that you could call and use that to generate rain images.

Um yeah and then the other question is how does the DAO verify the data? So again like right now we are very liberal on uh verifying the data as well. So um first of all it's up to the end developer or the end startup who is using uh 1 MB and lighthouse. So for example in the uh Zomato case which is a food delivery app what's happening is you could use a technology like ZKTLS which you can get from like reclaim protocol or opacity and uh that ZKTLS proofs can be verified um from those platforms and then you could u like reward the end user. So right now we don't have like a specific way of verifying the data.

It's up to the end user and that's how we want it to be like when you are the creator of the data coin you set up the logic how you want to verify long-term the markets will decide let's say if uh you are verifying the data data right you are then aggregating like a good valuable data set and then that will reflect into like will people would want to use that data coin or not but yeah this could be like something we build into into the future where we have some E mechanism to verify the data. Uh but that's still in process from our side.

Awesome. Thanks Nand for answering. Um one last check if anyone has any questions.

Yeah.

Before we leave the session.

Yeah, go for it. Aish. I was asking that uh like uh does this lighthouse is compatible with any uh aentic framework or like is it like more compatible with any particular framework which I should build on or does lighthouse provide uh its own framework? Um yeah so on our storage and encryption we have JavaScript Python uh and our APIs that could be used easily to store upload encrypt data on on the 1 MB.io IO like we just like really built this platform three weeks back so we don't have like a live integration with uh some existing AI frameworks but those are coming soon and there are some community members building that so it's quite open to you right now so if you have to work with some open frameworks be it let's say MCP or Eliza OS or whatever uh makes like the end app useful uh you can do that and if there's anything you want from our side uh that you don't have uh you can ask in discord and we can see if we can provide that to you.

Yeah, I I I was thinking to integrate like the open AI SDK uh framework so I guess that might be compatible if Yeah.

Yeah. Yeah. You could uh do that but we don't have any concrete requirements around um how you do that. [Music]