New AI Workflows in Remix - from Concept to DApp

Transcript

Hello everyone. Yeah, good afternoon. I am Stefan Tising from the Remix team. We are based in Germany and pretty much uh everywhere in in the world. And today we will be talking about new AI workflows in the Remix uh IDE. As I guess most of you know what is the Remix IDE. The Remix environment, it's the browser-based IDE that you can actually use for deploying in a very quick manner smart contracts and do further stuff without having to install anything on your desktop machine. So, with AI

with significant progress made in the AI environment lately, we are having deep integration that actually let you do quite a lot in in the web 3 environment even without specific knowledge. You can actually achieve a lot of stuff. So, the goal of this talk will be to have uh an interactive uh way of doing stuff together. So, you are pretty much free to ask questions at any time. We will be running some live demo on stuff that are actually live right now on the IDE and then from there on we will be also having a other less

uh at the end of the uh of this workshop a live demo of something I actually grabbed yesterday because yesterday I was attending a nodal meeting here. Uh a team from Switzerland and also based here in France. They were talking about a philanthropic platform for helping children maybe in Cambodia or pretty much everywhere in the world. And they were having some issues for uh raising funds and so on. And then I told myself, "Okay, I can just set up those stuff very quickly in order to show them. This is actually what you can also do in order to show your shareholders without having any deep

technical knowledge." So, with that being said, let us start by going through the overview. So, uh we are now releasing the Remix version two. With the release of the version two, we have significantly uh have significantly improved some stuff that we will go through. And then I will go over the AI workflow. What we have uh integrated so far. And then we will have an interactive live demo where we will go through the philanthropic example that I just stated some some while ago.

And I will give some outlook what is actually to be expected over the coming month or years from our side. So, let us start with first explaining a little bit of AI coding. I'm pretty sure you are all talented in coding and understanding coding. And maybe some of you might not be understanding coding and so on. Actually, AI coding is a nice thing and nice thing also have a a dark side. I mean, with AI coding you can pretty much uh create the app that you have ever wanted to to create, but there

there might be some security risks behind it and also the fact that you might not be understanding exactly the technique the technical stuff that are that were created from the AI and so on. So, with that being said, AI coding is good for pretty much everyone not without any deep review good for production code. It is good for you to kind of create a proof of concept. Sure, okay, the idea that I currently have right now is something that I can turn I can turn it into a technical proof of concept.

Yeah. And the good side of AI coding is that you are interacting with a machine in term of natural language. You are using plain English. You are you can also use a voice in order to uh interact with the AI to create yeah, edge cutting applications and so forth. And I said it is not suitable for production environment unless you made significant review, go through security and other uh architectural review. And the good side is that it is personalized. I mean, back then if you want to change something inside your code, you will have to kind

of con uh uh contact a a programmer or you will have to contact someone with the required expertise to go through your uh your code and do the change and so on. Now you can do it in a lapse of time. And that's actually useful. And this is what actually drives us to kind of integrate that inside Remix because the Remix IDE with combined with the complexity of the web 3 environment is not that very much understandable for everyone. And sometime you don't want to bother with all the security, all the different mechanism that involve web 3 development. You just want to know if

something is achievable, if I can actually do it and then just use plain English or plain German, plain French and then explain to the AI what you want to have and then go from there on to see if that can be manageable. So, what is new in Remix AI version two? In version one we already had some AI integration. And if you recall one year or two years from now we were having AI chatbots kind of arising pretty much from everywhere.

And those AI chatbots were not very much aware of the platform where they are actually where they are actually operating. You can of course them with technologies like retrieval augmented system where you go into database, get some data and try to extend the prompt from the AI, but you have no way of knowing exactly what is going on on the platform, what has the user done. And this was it was pretty much, "Okay, just explain me what this stuff is doing and then I will manage the rest." Nowadays the AI has a much more deeper integration that the AI understand exactly what did you click last time,

what did you did what did you do before and what are the foreseeable next step that can be that can be engaged. And these are stuff that are enabled with the MCP integration and assistant the assistant now understand pretty much the entire environment in which you are living or what you're actually doing. And we now also have a front-end builder which is actually a deep a deal breaker. We have seen a a lot of services out there a lot of services out there proposing AI generated web AI generated websites and so on. Now what we are doing, we are bringing the

combination of uh automatic contract general I mean, AI generated contract combined with a front-end with a website which actually communicate with the contract that has been generated in the background with one single prompt which is actually really nice. And we have done also some revamping of the deploy and run for those who have already been on the Remix IDE. And we are also introducing cloud storage. So, you will you won't hear missing any files anymore. Everything is backed up if you wish.

So, talking about MCP integration which is actually really nice to uh to state here is that we have two types of of MCP integration layers. We have internal ones which are pretty much legacy one. And internal ones which are developed by us not by any third platform. And those one are actually living in a browser which are understanding uh pretty much everything inside the inside the Remix IDE environment going from contract deployment, going from what file the user is actually being editing, going uh also covering what are the next steps or what is the user

intent, what is the user wanting to actually perform right now. How can we help the AI understanding what the user is doing so that we can give we can retrieve the best value out of the AI to propose to the to the user. Yeah. And we also have external MCP uh MCP server. And actually they they say a big challenge. I mean, from the browser perspective you cannot just host a MCP MCP server and so on. So, you will have to uh you will have to create in order back-end in order to host it somewhere and go it from there.

And then and then have access to that MCP. So, MCP is not something that is just given out by the uh I would say manufacturer or the guy or the company that actually provide the MCP. It's just a piece of code that you will have to host by yourself somewhere and then have access to to those specific to those specific feature. And this is what we are actually providing for major base I mean major big pro big protocol either protocol or project in the web three environment such as Alchemy, Open Zeppelin, Etherscan, all the other stuff. Those are external MCP

project which are not developed by us but which are actually contributing a lot into smart contract development interaction with the web three platform. And and so on. And upcoming we will be enabling the user to also host his own MCP host his own MCP server as we already do with Ollama. I mean we have a Ollama integration that where you can go full offline still use AI to achieve some workflow and we plan to carry out this analogy to to MCP where you can actually host your

own MCP server either on the cloud or locally in order to achieve specific task if you are technically versatile. So, we have already touched a little bit the points of MCP resources in the sense that now the platform is pretty much aware about what the user does. The platform being aware about what the user does doesn't mean that we as Remix know what is the user is doing. This is just piece of code in order to help the AI. Okay, this this is most probably what the user is actually doing. This is just some

probabilistic model in order to say, okay, this is this might be interesting for the user. If you say, I want to compile a specific project or a specific contract, there's no way that we as Remix we know what you as user but probably in a probabilistic manner the word compile is already matching with the feature that we are providing and therefore compiling will be at the top of the list of actions that the AI might choose to propose to you. And this is the type of resources that we are providing to the AI in order to uh to provide much a seamless integration.

I mean, quite good value to to the user. Are there any question until now? Feel free to ask any question if you have any. Now, let's jump a little bit into the Remix AI assistant. What is actually being what has actually been implemented on the left-hand side you have some conversation some conversation starter and those conversation starter are actually good for someone who actually doesn't know what he wants to do. It's like a really nice kickoff into doing stuff. You can here you can press on Alchemy or The Graph to have some example prompt that you can

actually use in order to query some stuff. I mean, for The Graph example you can query, okay, what was the last transaction above maybe to Ethereum and so on. These are information that of course you can navigate Etherscan, you can go pretty much into two or three web pages until you get the until you get the the answer. Right here you actually have uh you have to you have that in a very simplistic manner. And right now we have a set of model we are supporting Mistral, OpenAI and also there's the frontier coding expert from Anthropic. I mean, the Haiku and

also the uh the Sonnet model. And for now they are they are free for now for our beta user. As we will also see later, there's also this Mistral medium which is also free of use. Nothing is chargeable right now. And we have also the history uh yeah, the history as in any traditional chat environment. Uh actually this is the place where we actually jump into some live demo.

I hope you can clearly see what is going on. Yeah. As I was explaining some while ago, as you are load the Remix IDE, it looks like this. You have the chatting environment on your right-hand side. You have your conversation starters and the conversation starter actually help you kind of starting any type of conversation that you will that you would like to have. You will see the what they uh Remix is able to do. If you are not pretty much sure about what you are actually wanting to engage, you can just click, okay, what can I do? What are the MCP functionality?

Here's a summary. You can write Solidity contract, you have file management, you have deployment and interaction. And let us pick up some example from this Remix AI workflow. We I have written a very simple workflow prompts that you can use a one-to-one in the IDE in order to deploy a contract and interact with a contract without even having to take a look into a code. And we can do that together.

So, I will start this by starting a fresh session. A fresh chat session. And then I will just copy I will just copy this which is uh Okay, I will just type it. Create a new ballot contract in a in a new file. in a new file.

So, what I will be expecting here is the AI to come in the back to the IDE and then create a new file because we have a file system abstraction and we have a little bit of security. I mean, that's the stuff we we we should always have security for writing files. If you don't want that AI write any file into your workspace, you just deny. But in this case, we want to accept pretty much everything what the AI is proposing to us. And now we are seeing we are seeing a a file has been created create a new contract. Oh, sorry. [snorts]

I did Yeah. Oh yeah yeah. Yeah. And now I can't see anything. A ballot con tract amendment.

You can also be sarcastic to with the AI. And now he's processing in the background. It seems like there was an issue with the code formatting in the ballot contract. Well, this is a showcase issue. Let's repeat the process.

A new in a new file. What is actually happening is that we have a set of tools also in cooperation with Open Zeppelin. I know every time a tool is being executed, the tool name is being written over there. And I know by experience that that tool is a tool from from Open Zeppelin since Open Zeppelin are actually providing secure contract and the contract will be based on the Open Zeppelin

libraries. But it doesn't mean that the contract is safe. So, that's something to be taken with a little bit of caution. As you can see here, we have a new ballot contract. Oh, maybe there was a little bit of collision because in the uh workspace that I already have right now, there was a there was already a ballot contract. But the point is we have a new ballot contract here. You can review it. You can do pretty much everything you want with it, but I don't want to take a look to the code. I just want to say compile that contract.

And then you will have some feedback from the AI saying that the compact that the contract has been compiled and so on. I will say just deploy it on the VM. So Remix is having in a browser a virtual a virtual machine chain for very very quick deployment. Of course, you can also hook this uh to Sepolia as we will see in shortly in the example later.

Okay. I see what is going on. The ballot contract requires constructor argument. Please provide those constructor argument. I don't know about constructor argument. I will just say make some uh ar- ments up. Uh pardon me? I usually copy paste the error message. Oh, the error message. Oh yeah. Uh but in this case, you will rather get an error message from the uh from the terminal. Yeah, we can copy paste or we can just ask the AI to make some to make some arguments up for the

deployment. And hopefully it will make some argument up and then create a deployment. Yeah, as we can see, the contract has been deployed on the Remix VM. And for anyone who wants to read exactly what happened can kind of he can kind of go through the contract and say, "Okay, I want to know exactly the param and so on."

But what we actually want to highlight here is the possibility that it's actually may been made available that you can actually create a contract even if you don't have a deep knowledge what is a constructor and so on. You can you can still have access to the blockchain. And and so on. So we have compiled and we also have deployed and we can also we can also deploy it to to MetaMask. I already have MetaMask here in my environment. I will just ask it to switch my environment environment to

Sepolia Meta to Sepolia MetaMask. What I will be expecting is that it will change the execution environment for those who actually know how to use Remix is that we will have a execution environment here will be browser extension because MetaMask is a browser extension. And then in the background, it will it will call the MetaMask API because this has been done automatically because I already I already was connected uh previously and there was already a a

an approval. If there was no approval, I would get a traditional MetaMask pop-up to actually allow the access to to it. And then I can ask him deploy the same contract on Sepolia. Yeah, and here there is a little bit of caution because I mean Sepolia is a testnet, you won't get any warning. Uh there is a security warning whenever you are trying to deploy a contract to a to a live chain.

There will be there will be some warning from the AI because that can you can end up into the AI parroting and deploying the same contract maybe 200 time. It won't happen, but you can never >> [snorts] >> you can never discard this uh this eventuality. And now we just accept here to deploy the contract and it going it's going through the deployment infrastructure of the Sepolia network and everything is being done in the background until we have Yeah exactly.

We have the confirmation that the contract has been deployed on Sepolia. We can also navigate uh I can close MetaMask. We can also navigate on Etherscan for that specific transaction. I don't even know which one. I think that was the second one. Huh? Yeah, there's no there's no direct link. But I can copy the contract address and Sepolia.

Just to make sure. Yeah, it says it is a contract. There is no interaction now with a contract and so on. So this was a very super duper vanilla integration of I mean showcase of AI usage for contract deployment and uh functionalities. Yeah. What you can actually do with the Remix AI in order to achieve some minor minor stuff.

Yeah, we have gone through the compile and depart and and deployment workflow using the AI, which is fairly easy. You just use natural language. You can also use audio prompt if you don't feel like you want to type on the keyboard. And your audio will first been transcribed to plain English. Now, I think we only support English. You always never know with those model what kind what other languages are still supported and so on. Maybe if you speak French, it will also get supported, but most likely you won't have the same accuracy as if you happen to speak English.

Yeah. And of course, we also have advanced prompt. And in advanced prompt, you actually want to achieve something much more bigger than just a simple contract. You want to you want to create something new. And by creating something new, I means like creating a entire new workspace, entire new workspace with a new project. Let's say, "Okay, I want to kick I want to kick off a project and I don't know where to start. I don't know where to start up." I just come there and a Remix in the Remix environment, I say, "Okay, what I want to start a new workspace." And I just describe what I actually this

workspace will be able to do. Let's just say I just copy this first one, which is an example a practical example. I want to create an ERC-20 token and explain with comment in a contract what I will be expecting from this workflow is that I get a new workspace generated with a set of contract. I don't know how many contract. I just need exactly a ERC-20 contract being generated and also with pretty much all the description of every function, what that function is actually doing, what is the input parameter, what is the output parameter

without me specifying anything. So the AI come up with a structure and when that structure is done uh when that structure is done, it creates contracts and uh one uh very specific part of this um of this process is that the contract that is generated is being checked to be uh is being checked in in the back end to uh not to not be having any vulnerability. I mean, static check with Slither and other stuff.

And also it has been made sure that the contract is compilable. I mean, when we get the payload from the AI, we pretty much verify that the contract is compilable so that the user won't have to tweak it anymore with other with other stuff. There might be some dynamic error as we have seen earlier with the uh with the deployment where we might need some uh some complex argument for the for the constructor. Now we have our commentated token. So what the AI actually understood is that we want to have an ERC-20 token and he decide to call that token commentated

token. Now you can see what uh power each word is actually having in the prompt that you actually have that you write. What the result you will get from the AI will strongly depend on how you formulate your prompt. I mean, if you are precise enough to give uh a lot of pinpoint to the AI, the AI will come back with a very solid answer to you. And if you just give some middle quality prompt, you will have something similar that I mean, no one will ever write a comment that token.

Ping me if someone does that. Yeah, as you can see, we end up having a project with exactly that token ESC token that we uh uh that we requested and everything is commented. For some For someone who actually wants to learn about Solidity without even getting that much deep inside the code, that is already very valuable because every function that has been generated has a uh it has a comment that he can actually relate to.

Yeah, and then I go back. We also have introduced the newest technology or the newest stuff going on with uh Open Claw, Code Claw, Open AI, Code X. on the skills. Now, in Remix AI, you can actually load a specific skill a specific skill set in order to achieve a very specific function. Now, we might ask the AI. A good thing to always do is to ask the AI what he can actually do.

We might ask him, "What skill do you have?" Yeah, as I was saying, a good thing to explore the AI is to have a a sort of conversation with the AI like first having acknowledgement of what is the AI capable to do because it doesn't make sense to ask the AI "Yeah, use the Solidity skill to write a contract Solidity." If the AI doesn't have an implementation of that Solidity skill, it will start making some stuff up because the AI

is the piece of knowledge or a probabilistic model that thinks he knows pretty much everything without any constraint. He will pinpoint something from his memory and then use that for generation, which is not also wrong. Okay, here we have a set of skills that are actually load inside the Remix environment. It is up to the user to kind of discover those skills. Now, those skills are not load dynamically based on the user request. This is also something that we might work on, but it's up to the user to know exactly what skills he actually wants to to use to perform a specific task.

And yeah, security skills. Let us say, now you know the AI is having a specific security skills. We ask the AI to review this contract using the uh security skills. Using Yeah, using the security skill, analyze the commented token.

What this will do, it will trigger uh from the AI it will trigger a workflow where first there is exactly the security skill being loaded, which is a set of a set of instruction that the AI will follow while reviewing a specific contract and this set of instruction is written in plain English like "Okay when scanning a contract, maybe you look for the specific latest uh uh maybe zero-day exploits. If that I mean, if that uh if that skill where is hosted somewhere online, it might be updated in like 2 seconds and with the zero skill and the

that skill will be load in the AI and they are specific they are specific uh they are specific instruction that will be that will be executed. Huh? I don't know what happened. We'll just repeat. Oh, it was still working. I'm sorry. The security analysis contract has been saved to >> [clears throat] >> Okay.

Um I don't know. Contract security analysis. I don't see anything. Okay, it's saying that it has been saved. Let's give it another try. This might be an edge case where the AI is parroting a little bit. I mean, from the workflow, it is very it is matching what we would be expecting from a skill to be done. We first load the skill and then we load the contract and then we do the

analysis. Analysis contract has been The analysis has been saved in security empty token. I don't know why. Well, this is a case where writing files is actually failing. Give me that security on on the the These are in plain text.

Yeah. Okay, actually we might be having a a tiny issue either with the model or or the tool, but I want to highlight here that we are using a low-end tier model, so there might be a little bit of parroting somewhere. It is actually dedicated for very simple task. And yeah, as you can see, we use the skill and there is uh there's a report about what can be done. There is a issue, there is a fixed solution, which we can actually address in in subsequent message. We can actually

write here, "Fix all these proposition uh fix all the proposition you did above." And then the AI will go through the over the contract and then perform all those fix for us and then we will have a We can also we can re-perform that uh security analysis. After re-performing that security analysis, we will be helping to have a much more flawless contract that that is audited and that you can use in to deploy somewhere.

Okay, and then we also have querying token and nodes in specific uh address holder. I mean, you can also play a little bit of kind of querying what actually Vitalik did last night if he kind of sent some tokens around the globe. You can you can query that with the with the with the AI integration when when our when our date. And maybe you can construct a business case based on that or you can create a contract which is doing some specific stuff based on that.

Yeah, we went through generating a workspace where you actually generate a new workspace with contracts and depending on the description that you are actually providing. And and so and so forth. And whenever you generate a workspace, there is a new workspace with your dedicated code and you can be pretty much sure that every time you generate a new workspace, you will get a contract which is compilable. You don't have to fix around in the contract. You will yeah, I mean, you can extend the contract just by asking also the AI where the AI interacting with it doesn't guarantee it to be compilable or to be static analyzed.

Yeah, now we come to the to the second part of our workshop, which is about app creation. We call this Quick App. Quick App, why? Because it's quick. We can really create uh decentralized application in a very in a very quick manner and this has been done by imple by extending our already available Quick App version version one uh by integrating AI.

And now we have also several options. You can publish your app on different on different uh on different blockchain. You can even have a a blockchain mini mini app and the source code of your front-end application, you can also host it somewhere uh either on Limo or on also on IPFS. And yeah, let's create the philanthropic project we I was talking earlier uh as I started this uh this workshop.

Oh, do do. I will push this a little bit. I cannot see anything. I will start this I will start this by creating a new chat session and then create a new workspace. And in creating a new workspace, I will paste in here the pre-prepared prompt I I crafted yesterday, which is create a create a contract for supporting a philanthropic project that helps family

and kids. It should be uh it should issue some initial tokens to user registering for the first time through direct minting. This is a little bit of web three speaking in order to kind of restrict their generation space uh the generation space for contract. And anyone can register themself. A user could uh a user could then spend his token to a special project. The contract owner should be should be able to update a list of project and also close them manually or automatically after a specific uh time has uh has gone.

And also add comment to the to the function so that I will be able to maybe uh explain it to someone else or uh grabs the idea behind it. And then I hit generate. And I am using here the workspace the new workspace generation uh feature because I want to have an atomic solution. I want a solution that kind of match my intent and also want a solution that is compatible, that works, that only works out of the box without me having to interact with anything at all.

And yeah. Now, we wait a little bit while it is generating. You always have some feedback here. What is actually going on in the background? We are compiling the contract. And if it fails, they say I mean the agent will reiterate reiterate the process. Reiterate the process in sense in the sense that the same prompt will be sent again to the to the AI. And then we will be uh yeah.

That was the case. That the contract was not compatible. And then we send the prompt to the AI to for another generation. In that time I'm drinking a little bit of water. Well, that takes way too much time. Mhm.

Let us give it one more try. Otherwise, we will maybe reload the page and start from scratch again. Mhm. Yeah. There might be there might be an issue. I always have to check maybe the issue can be related with the internet connection.

Which is optimally not the case here. Just a little bit of patience and then we will be we will be there soon.

Yeah. Exactly. And here we have our philanthropic token, which is an ERC-20 with a I guess with the initial amount of with the initial amount of token by user, which is 1,000, which is really good. I mean, by experience I already know when I'm looking at function minting and so on, I already know that there's a process of issuing some tokens to some users and so on. But this is not what is actually uh good for us.

I mean, we can go through the process of asking the AI to compile and deploy it everything and so on. We have already gone through that. We don't want to do it anymore. So, we will just do it manually to speed up the process. I compile the project. The project is compatible. And now I want to deploy this project on on Sepolia because uh it will be much more easier to interact with and also for you. I select the browser. We have Sepolia. I have this specific account. It is also the account from which I am logged in.

And then I will just deploy the philanthropic token, which is compiled here. I press deploy. I expect MetaMask to send to create a transaction, which I confirm. >> [clears throat] >> Okay. I wait a little bit and then voilà. I have a contract, which is deployed on Sepolia. The good thing is now, in order to start Quick DApp, there's this treat up here right to your contract. You just go over there and then to create a DApp.

In creating a DApp, when you press that, you will also have a text field where you actually describe how your decentralized application will look like. And this decentralized application is just a website, which will be reflect reflecting what is inside the contract. What is done here is that uh we are taking the contract source code and also the contract ABI, so we know exactly how to connect those and then we connect it to traditional traditional website uh uh I mean, traditional website front end generation that we already know. And for that purpose already also have a a prompt for

that. Which is a little bit long because we want to have something nice. Which I will just paste in there and then read again through it. So, I want to create a DApp for this philanthropic project that I already created the contract for. Where user can participate to some blockchain web three donation. So, I want some user to to donate some tokens. In real world case, that will be donating money. Once the user connects to a connects to the platform with its wallet, there should be a mintable button for initial minting process in order to get the philanthropic token.

Actually, the token you want to uh you want to uh contribute with. That will that will contribute to the philanthropic project. Okay. So, see register user from the ABI to make sure that stuff are aligned. Make it much more modernistic to look at to look make it modern to to have a really nice UI. As a contract owner, I want to be able to push new contract. I want to stop ongoing ones. I want to I want the AI to make all contract function available to those with access.

I mean, there might be users uh that I want to relegate times I don't want to host any project anymore. I would like to say, "Okay, this user, you have contributed a lot. Just go ahead and take over this project and so on." This is what this prompt is. I mean, we can go further, but I will hit generate. And what this does is it is similar to it is similar to what we already know from the new workspace. This one will generate for you a React a React front end uh application.

And that React front end application will be will be React here in your browser, so you can also use the AI in order to uh Yeah, you can use the AI in order to go through uh through the code review or change specific stuff. If you don't like it, I mean, if you have the design and say, "Okay, that looks maybe too bright." You can just ask the AI, "Okay, make it a little bit dark." And so on. I think you get the idea what without getting too much uh specific into into deep into details.

You can you can change everything you want as as you desire. And also one good thing is that uh the the space for the React app is very different from the space uh from the from the workspace for your contract, so there is no so there is no interaction, there is no confusion. I mean, the front end app is based on your contract, but where you are actually developing your contract is separated from where you are developing your front end application, which is something you always want to have. You don't want to mix them all up.

Now it's taking a little bit of time, still cooking in the background. Cooking, and then I hope we will end up with with a nice front end project that we can publish. Yeah, and then we are done. And then once it is done, you can see up there Sepolia. That means the contract has been deployed on Sepolia and all the interaction you are having is on Sepolia. And here you have your philanthropic platform. You can click here deploy to IPFS in order to have a Yeah, an external link. And we can also register it to be a Remix DApp

application, and then I will give it uh filler. Okay, then register subdomain. I hope this goes really quick. As we only have less than 5 minutes. And then I can click to this link. Let's see which one was quicker. Yeah. I think the IPFS was much more quicker. Now there's no project, you can connect your wallet to it.

I will just connect my MetaMask. I am already connected, and I can register and mint my token. As you can see, it creates a it creates a MetaMask transaction where I now get 1,000 tokens. And so on. And I can also create a new project. Exactly, after the minting process, you can see my account balance has been updated. And now I can create a project. I can say this project I want to test. And now you can also specify the bene- the beneficiary of the

uh of the philanthropic project and the target amount also. Let's put this to 100. Huh? What did I do wrong? Let's create again. Oh, I think Yeah, I think I guess I know what the typical problem is. Yeah, I mean, nevertheless, we do not have enough time to go through through this anymore, but I think you guys all get the point of uh these philanthropic tokens being created.

This is this is actually the power of the new of the new tool that you actually get in a in Remix AI. Whenever you encounter such a such a problem, you can you can always return you can always return to the to the Remix to the Remix AI and then explain the problem here, and then this front app application will be updated until you will arrive to the point where you say, "Okay, this this front end application is now reflecting my proof of concept that I can actually uh I show to uh my stakeholder, and then they will have a much more better understanding of what I am trying to

build without going through several months of development and and so and so on. Since the time is very short or already almost up, is there or are there any question that I can quickly answer? Yes please. Uh yeah, can you please say? Um we came a bit I came a bit late, so I'm sorry about it, but which model is below the Remix uh AI? Yeah, we are using a set of models for different task.

Now we are using the Mistral medium as a default model that you can always get, I mean for for free. Yeah. For specific for specific uh deep AI uh requests, it might not be sufficient. I guess you already know that's the main reason why OpenAI and Claude they have a pricing tier which is much more higher for this big model. And actually, if you have a solid workflow, it would be much more better to use those model because you can definitely feel the difference.

Yeah basically. >> uh rate limits on this? Like if you can iterate and iterate again and again? Yeah, uh there is there is a rate limit. You cannot iterate 24/7. And or do you have any number like if people are using it for main deployment or anything? Like I mean, the this is a very very new concept. This is a very new tools, and we are actually giving everything out in a beta. I mean, there is absolutely no restriction, no paywall. You just register as a beta user. I will encourage pretty much everyone to go on beta and then try these features all for free.

Are there any further questions? And on note, I thank you very much for taking all this time. >> [applause]