Cartesi: Build Appchain Rollups with Any Code

Transcript

GMGM. Welcome to what's going to be another fantastic episode of Devantel. So, if you didn't know, Devantel is a 30-minute podcast allowing founders, hackers, and anyone in between the opportunity to come on the podcast and showcase what they built. And today, I'm ecstatic to welcome Joel Garcia, who is the lead devel lead developer advocate at Cartezi. So, if you didn't know, Cartezi enables developers to build app chain roll-ups with any code. That's any programming code. uh while benefiting from the security of Ethereum. So, if you stick around for today's episode, you'll get to meet Joel, learn about Cartezi, and how you can get started with them today. All right, let's get into it.

>> Hey, hey, hey, GM. >> Hey, GMG. How you doing, Jo? >> I'm good. What about you? >> I'm doing well as well. Uh thank you so much for joining us on this Wednesday. Uh I I know it's a little busy uh this time of year. So uh really happy to to to have you on today, man. >> Yeah. Everyone uh in a rush for the holidays, but always good to to have some space to breathe and talk a little bit, you know. >> Exactly. Exactly. >> Thank you for having me. >> Of course. Of course. My pleasure. Um yeah, I I've been digging a little bit into Carteesi and um it looks really really dope. Um the the fact that you can use any almost any programming language to to build smart contracts is a pretty pretty neat concept to me. Um

and I'm sure our audience as well and I I guess before uh we get into all all all those details um would you just like to give a brief introduction about yourself? >> Yeah. Uh sure. My name is Jean as you mentioned uh I'm developer relations lead for Cardi developer advocacy lead. Um, so I've been in working in Cortezi, I think, for a little more than a couple years now. I've been working with crypto a little while before that. Um, and it's been um it's been a real nice journey. Real nice journey. Um, I think uh quick quick things about me. I'm a I love nerdy things. All nerdy things. I'm a tennis player, as you can see from my

background. I love Lord of the Rings, you know. >> [laughter] >> Love it. I love it. >> I think you too, right? So, yeah. Um, that's that's me. That's essentially me. [laughter] >> Awesome, man. Uh, yes. Uh, that I I didn't know I didn't know there was a extra Lord of the Rings element on the show today. Excellent. >> Yeah, exactly. [laughter] Um yeah, I I guess um just curious like uh how did you find yourself um in in the technology sector and and like what kind of led you to Carteesi?

>> Yeah. Um I think since uh small I had this tendency to the computer itself. My father worked with IT and uh he was a developer then he started on managing positions but always very very very connected to technology and so when I went to university I was between technology and architecture and I I just think I don't know technology was had a better market [laughter] than architecture right and I really found myself in in university. I started in um computer engineering. Then I moved

to computer science. I had a very cool um Apple opportunity internship, Apple Academy, Apple developer academy. And there I met some friends who who are my friends to this day and who I work with to this day, which is really really nice. And yeah, that's that's how I started developing. And these friends are actually the ones that brought me to to Kartzi, right? Because we've been uh so close since. And they they've been in crypto for much longer than I do. I mean, they they've been diving in it. I think since 10 years ago or I don't know exactly when, but I knew that some of them um started

working on this project called Carteesi. I always heard about what Cartezi was and I think I think it was around 2023 mid 2023 that I started talking to them to get a better understanding of how they actually did the things and then I found that that's really interesting that's a little different from what everyone else is doing. Okay. Uh I think I think you have a nice path for web three right there. Let's let's try to dive to to dive heading in this thing. [laughter] >> Yeah. Amazing, man. Yeah. Uh I I think you you chose a very fascinating field

to to enter. Um just curious like uh a lot of people uh a lot of people's first exposure um to to blockchain and crypto is through like Bitcoin and Ethereum. Um was that the case with you or or was your first um interest cartees itself and then everything else? >> No, actually my first interest was a while ago uh with Ethereum trying to understand how Ethereum worked. Um I remember there were some classes at university. I really tried to study to better understand it. I found it kind of limited at the time. Um and then in my previous work before Kartazi, I actually had the opportunity to work on a web three project for

green energy tokenization kind of like uh kind of like carbon credit, green energy credit. So I started to have to I I needed to get a better grasp for it and and then I started to see the real potential real potential behind the thing, you know. Uh so it was kind of through Ethereum, but I studied lots of different technologies there to see whatever fits best. And after always finished, then I started thinking about Cardasia and I saw how would it have made my life easier. [laughter] But but [snorts]

the first contact was through through Ethereum. >> Awesome. And and that this is a great segue. Um yeah, I guess uh what what is the elevator pitch of Carteesi? Um in in your own words, >> I think it's enabling people to do what they are not capable of doing on Ethereum by itself. Okay. It's of course it works through a rollup. So it computes offchain and brings the information on chain afterwards. Um but it empowers the developers to do whatever they want with the technology that has been created since technology since internet is a thing. Right? So we we are bringing the

tools from web two to web three and in the best way possible. Right? It depends on on who we're talking to, but Linux on chain kind of sticks to some people and then web two on web three sticks to some people. It kind of it varies varies a little bit. [laughter] >> Yeah. Yeah. Yeah. Certainly does. Uh and uh so the concept is essentially that um you can use any programming language to create um smart contracts credit.

>> Exactly. Exactly. Wow. >> And not not just the language. This is very important because there are other protocols that allow you to use other languages, >> but the way we do it is running uh an instance of Linux inside the virtual machine that we created. That means you're not like transforming one language in another, right? More technically speaking, you're not transpiling or recompiling or doing any sort of this magic. What you're doing is you actually have a Linux environment. So anything that you can do on Linux, you will be able to do on it. Which means you can also use different libraries. So you want to create a game, you can put the game engine right there.

You're not just using the language, you're using the tools for it. You're working with um AI models, you can put an AI library there. you know you are working with um let's see uh you're working with a very complex financial application that you need to calculate linear regression you can use mathematical libraries that will allow you to do that instead of rewriting the whole thing recreating the whole thing on solidity and doing it from scratch you know we're keeping it simple and just allowing you to do your best with what you already know >> wow that that's that's quite the powerful construct and yeah like when you're describing desing um the environment being like a solely Linux

one. The first thing that popped in my head is like, "Oh, can I run maybe like a local AI model or something like um side by side in the same container?" And it's like, "Oh, you you said it." It's like "Oh okay." >> Exactly. Um >> yeah, >> there are there is a teeny tiny limitation um I would say is that the computing power of the the execution environment is limited, right? Uh as any execution environment. So you you would have to scale to get better performance. Uh when we're talking about AI models, if we're talking about uh general AI, if we're talking about like uh linear regressions, we're talking about decision trees, we're talking about um many different uh uh AI algorithms, AI solutions. You can we have templates for

many of them on Cortezi. Of course, if you want to run a full LLM, a version of chat GPT, that's much more costly because in general, you need big C G G G G G G G G G G G G G G G G G G G GPUs and it's much harder to do. It's something that it's something that every now and then we explore and play around with, but it's deep learning uh deep AI is is harder to to deal with. >> It it it is. Yes. And like the cost of of RAM and GPU these days are is through the roof. So, >> Exactly. [laughter] Exactly. Uh, awesome. Yeah, I uh I'd like to double click a little bit on um some of the concepts uh surrounding uh the Carteesi EVM. Uh starting with uh the Cartezi machine. Uh can you uh give us a

little uh a deep dive into um what what those are? >> Yeah, of course. Uh the Carteesi machine itself is is a is this more powerful machine is this more powerful execution environment. being very technical here uh but trying to be fast to not bore anyone. Um we created uh an emulator that uses the risk 5 uh instruction set. So it operates like a risk 5 processor but we have to make it deterministic because in the root of web tree uh you have to be able to verify the computations. So

deterministic as in for every input that you send given a state you will always have the same output right. So you can actually see what someone processed and actually verify it. So you can so you can attest for the the the truthlessness of the computation, right? [snorts] And yeah, so this is the whole magic and for doing that we have we had to kind of make many many a very complex uh implementation of this risk 5 instruction set. [snorts] But now in the high level explanation, we basically created a virtual emulator that allows you to run anything given it's a

open-source processor and for convenience we put Linux on it for anyone to be able to get ready and start coding on the go. So use a a familiar uh environment. >> Yeah. Yeah. And yeah, that familiarity is is huge. Um I know like a lot of people >> um who do start in like smart contract development for example um they gravitate towards solidity and if you have like some JavaScript or TypeScript um uh knowledge under your belt, it's a little easier to pick up. But then like once you get into the nuances of it and you get the gas golfing and this and that and it's like oh it's not it's not as uh not as straightforward as one

might think. Um but yeah, this this sounds very fascinating and um uh this concept of the u of the Carteesi machines they play in the um they play a role in the in the grander scheme of things, right? Um with the with the actual rollup itself um the Cartezi rollup. Uh do you want to get into that a little bit to Yeah, I I think for that I I would like to share my screen here a little bit to get a little bit in the >> I think in the >> weeds. Yeah. >> Yeah. Also, uh, don't be afraid to nerd out here. Uh, we're we're a bunch of developer developers on this show, so >> All right, >> get as nerdy as you want.

>> I'll do it then. [laughter] >> Okay, so talking about Cardazzi, u one of the coolest things that I want to show you later here is that you can actually do with Cardazzi. And one of our our uh spotlights is running Doom on chain, right? So, everyone runs Doom everywhere. People ran Dune on Minecraft. People ran Dun on Notepad++. And people were I think it was 2023 doubting like, "Hey, how come no one ran Dune on Ethereum yet?" And then this guy who's a developer for Karti who's crazy good developer called Eduardo. Um if you if you ever play Tibia, you might have heard of him. But um his uh he he

actually managed to record the whole gameplay of Doom and replay it on chain to verify the gameplay of person of people. So anyone can actually run a game of Doom and verify it on chain. So you can attest for your game plays. This is very cool for speedruns and other things. Um, so we we ended up seeing like this is just a I mean we we ended up seeing that this is just a a tip of the iceberg. We there there is even the creation of this whole other solution that I will show later called Rives for gaming but it's not just about gaming. Um so let's talk about decarti machine and decarti

roll-ups right we mentioned the cartesi machine it's an execution environment and it's deterministic right so for every input that we give we have to get a specific output and that should be really really simple it's should be obvious and it should be just that but when you're picking your execution environment for whatever you're developer you have to consider so many things like what you're optimizing for um what kinds of deps people are going to be running. You're going to have to think um what kind of applic uh languages your developers are going to be using, what kind of resources they leverage uh will they need introspection

um containerization. Uh is there is there a need for instant composibility? Right? So you run your app, you need every other app to be able to see the result of it immediately. Can you tolerate fast finality? Um, so we created this environment which for us at least it sounds like uh right now it is the best execution environment for web tree. We call it B. So this is our mascot. It's a B. Um, and we we think it's so good, right, as I mentioned, because of the the things that I mentioned, Linux and more power in the CPU than the EVM or other environments can provide still. Uh, right, it does

not use GPUs, just CPUs. So, it's limited in that sense, but it's far far better than than any other option that is out there right now. So I like to to think of the the advantages of Linux not only that you can run everything that has been made so far but you are also going to be able to run everything that will be made in the future. So think of like new libs uh new tools uh new gaming engines new AI tools new mathematical tools everything that's going to be built that will run on Linux will run on this execution environment. So that's why

it's so good, right? Uh there's also the difference of uh do do you know that uh story about the space race where people were just uh trying to develop this amazing pen that was able to actually write in outer space and >> so they had to create this special pen that projects the ink with gravity because it wouldn't stick to the paper and then the other team came and just use the pencil which make it made it so much easier. So I kind of like this analogy. You know EVM tooling is very flexible in the sense that you can have like a plier and you can use it for so

much and you can even hammer things with a plier, right? But there are specialized things that give you better and more specific functionalities. So uh we try to gather all this tooling. So you can use AI libraries like scikitlearn, pytorch, you can use databases like SQLite on your applications. You can develop with any language as we mentioned and this is just one of the advantages of the solution. The other one is the way we structure it. It's we structure it through roll-ups and it has a huge advantage because on rollups you're computing things offchain. So you're not competing for block space and you're not computing you're not competing for the

execution environment of the EVM. So you do not uh you do not make uh gas you do not have gas prices so high you do not have all the issues that you have to overoptimize on when you're developing on solidity on a general purpose chain right so a transaction arrives on Ethereum the Carteesi app chain will get these inputs it will process whatever and it will settle back on Ethereum um if we think about how multiple apps work. Let's say we have a gaming uh app chain on one side and a and a uh machine learning PyTorch app on the

other side. And there are many validator nodes for each, but they each of them only validate the applications that interest them. They don't have to validate everything. So they're not kind of stuck in the general purpose environment and they work separately and one does not compete with the other. So roll-ups allows us to scale like this. Application specific roll-ups allows us to scale like this. Right? Of course, you can also play around it with other protocols. So you can use a different sequencer, a difference data availability layer. So you have different performance, faster confirmation times and you can mix and match however you want. Right? So these are the two key things for Cartisia roll-ups. app specific roll-ups and the

Linux environment that allows you to do anything without competition. So, we've built some things uh that were very cool. I mentioned Doom already, but we built a game called Bubble Wars, which is very fun that you pay gas to move your bubble and if it hits another bubble and it's bigger than it, it actually consumes consumes it and so you get the tokens that the other guy have, right? So whenever you move your bubble, it's trying to eat your other bubbles by paying tokens, but you can always get more tokens because you're eating someone else. So it becomes a kind of like a crazy DGEN massive multiplayer game. [laughter]

Um, have you ever played Sim City? I think >> yeah, >> everyone played Sim City. We actually managed to put it uh on chain as well, so people can create their own cities. and the the you deposit some tokens to start playing and if your city makes money, you make money. If your city loses money, you lose money. That's that's also very cool and it's actually a real use case. We have some social applications as well. um co-writing where you can actually share the share the revenue um with different uh proportions and this calculation can only be made of course because of the complex computations that the machine can can process or even u a drawing

canvas where you can actually check for the steps in the drawing of the person. So you can actually see that the NFTts that you're minting were created in the platform. you can prove the creation of the NFT not only not only mint it but mint it with the whole history of the drawing which is very fun as well. uh and of course some machine learning templates as I mentioned biometrics classifier Iris classifier which are very common data sets Lilium that a friend of mine built which is a for carbon credit sol it's a carbon credit solution that uses AI to verify forests which is very nice and you can see much more solutions on rollup lab um I will

show uh one um real quick here this is called ries right it's Uh, it's what came from Doom that I showed and now it has so many more games that you can play. This is my favorite one. It's called Blade Bomber. So, I will just do a small play through of it here. And you can see the kind of complex games that you can that you can run here. So I am an adventure on a dungeon right now actually killing monsters and people are able in this application to create their own games and sell their

own games. And all of these games run in a template that run in this uh emulator for for rides and you can then submit your playthroughs and it's all verified. So speedruns, contests, lots of things can happen here. And just to show how easy it is to to develop here, I have this playground application which is basically a template of it's a website tutorial for you to learn how to develop with Cartezi. Kind of like cryptozombies is for for Ethereum.

And you can see here uh let me do some zoom in zoom in just for people to be able to read better if they are on their screen. And you can see here that there is an explanation of like this is where your inputs come from. So whatever you send to Ethereum will hit this handle advanced function right here. Right? So let me just decode the data that is coming from from Ethereum. So I'm going to do a hex to string on the data payload and I will console.log the decoded variable. Um and I can just run it and I can actually

do a console log of whatever I whatever I I decoded. Right? If I want to do other things, if I wanted to do uh let a = 2 + 4 and just console log it as well, I can so I can just play around however I want inside Cartezi and I can do whatever I want here. I can run a full game engine like you just saw with RES or I can run an AI model. I can run complex mathematical computations. I can run a full um uh marketplace uh whatever whatever you want to do really. It's you're you're free to to do whatever you

want. And as you can see, I'm just coding here with uh JavaScript. So, it's really that easy and that immersive. You can just import lips, import tools, and feel free to do it. Uh and that's it. That's that's the the the showcase, the speedrun on on Carteesi. uh what it is uh a good example and how you can develop with it. >> That's that's amazing, man. Like uh I'm not going to lie. Uh and every time that you're um for example, like you're working with with that um example um tutorial, uh when you run the code, do you have to pay any any gas um to have that code execute or is that all like

locally sandboxed? >> For the tutorial is completely sandboxed. Uh the way it's running now, it's emulating the Carteesi machine on the web browser. We can actually put the whole Cartezi machine in the web browser. Um but it's just an emulated version for now. [laughter] >> Yeah. Yeah, for sure. For sure. And um yeah, I guess uh you showcase people like all all these um amazing capabilities of Cartezi. Are there any SDKs um frameworks available for developers to easily get up and running with?

>> There is. We actually developed um the the CLI team developed an amazing CLI for Cartezi where you can bootstrap different uh apps from templates. So you can already have a head start just like the template I showed in JavaScript. You can have a head start with Python, Ruby, C++ Java um TypeScript uh, Rust, Go. I I don't know if I mentioned them already or am I already repeating, uh, but so many different languages that are already available for you. And it allow this SDK allows you to run your apps locally. uh you can you can uh uh send inputs, you can emulate sending inputs. You have the

whole environment running on your machine based on Foundry, right? And but it's all ready to go. It's just a simple comment like Cartezi create, you create a template, Cartesi run, you will run your app and then when it's everything ready, you can just do a Carti deploy. And there are high level frameworks for a few languages as well already. We have high level frameworks for Rust, Go, uh JavaScript, TypeScript, uh Python. So they they make your experience so much easier, right? Uh they already handle inputs, outputs, asset deposits, lots of different things.

>> Awesome. Awesome. Uh let's see. I want to see if there's any questions from from the crowd. Uh let's see. Uh Lucas asks, "Are there any plans to add support for Java in the CLI? If not, is it something that uh he could do himself uh somewhat easily? He's a Java dev." >> It's something that you could do it yourself, but it's already supported on the CLI. It's it's very easy to do. You can just get a template and see how it works. It just uses a a small lib and you can basically copy the container and just adapt to Java. But it's for Java, it's already ready. But we're looking for someone to to make a Java high level framework as well. So that could be a an

awesome uh an awesome thing to see. >> Awesome. Uh great question, great answer. Uh Prince uh Nanso asks, "What are the possibilities with the Cart Cartis machine?" I guess um beyond what you showcased um what are some uh other applications or or things that you think maybe uh be cool for or um a car testing machine to run? >> Oh man, there was this there was this project that was happening where it was an automated market maker. Um I really wanted to see something like that coming up again, right? Because with the complex computations that you can do, you can play around with with assets in

a to a different level to a different granularity. There was a project that we saw in a hackathon that it it was really cool as well which was a high complexity governance solution. So you could have like quadratic voting and different uh different ways for votes that wouldn't would be possible on solidity but would be so much harder to do and so much more expensive to do. So yeah, I think I think those are two that are on the top of my mind that I really wanted to see. And uh an MMO game, I would really want to see that happening

as well. [laughter] >> That would be cool. That would be cool. >> Uh maybe it's it maybe it's something you build, Prince Nanto. There you go. [laughter] >> Yeah. >> Um I mean Yeah. I mean the holidays um some people uh like to build uh even though you should technically really rest, but yeah, why not? Why not uh take a stab at it. Um yeah, and as as we're kind of approaching time here, uh just uh got a couple more questions for you uh before we let you go. Um can you share um any uh anything uh in the in the road roadmap for Carteesi perhaps coming up in uh 2026? Anything exciting you can reveal?

>> Yeah. Yeah, we we are developing the new version of the SDK the node v2. It's going to have a much more interesting developer experience. First thing is now it's going to include our fraud proofs algorithm. So your computation will actually be proven, right? uh it's going to be verified right through our fraud proofs which is very nice. So you do inherit uh Ethereum security and it does come with lots of uh new convenience features that are really cool. So before we used uh GraphQL and um JSON RPC and now we're kind of unified in JSON RPC.

So it's more straightforward. uh web free developers have a better understanding of it. They're quick to get on the get-go, which is really nice as well. And yeah, some some things that are going to be able to people are going to be able to follow their computations like they're going to get instant confirmations on their on their applications, but they will also be able to follow up the disputes from our fraud groups, which is really cool. And oh, and there are some apps that I'm really looking forward people are discussing some apps to build and I'm looking really forward to see them coming to life. [laughter] >> Amazing. Yeah, sounds like a lot of

exciting stuff and uh yeah uh we'll we'll be on the lookout for all that. And um for the developers uh and audience members here who are interested in getting started with Cartezi, um what's the best way for them to do that? Um where would you recommend they start? I think they should always um take a look at our discord. We are an open source project and we are open source to the level that even our discussions are in the open right so we have a whole uh R&D section on discord that anyone can follow where we actually discuss the next steps we discuss how we're going to do things and what are the best strategies and best approaches right um so follow us on our discord and

participate and engage on the discussions because that's where the things happen and of course our our ex is also very active uh thanks for our awesome marketing uh marketing team. Then yeah, I think those are the two two key places. >> Excellent. Excellent. Yes. And we'll have that information for you um in the description below of uh wherever you're uh watching and listening to this podcast today. And uh with that, just want to thank you so much. Uh it was a it was a pleasure having you on the show uh learning all about Carteesi. It it's really really really cool. Like um I'm a nerd myself and I was kind of nerding out when you when you started the show Doom on Chain like oh my gosh.

[laughter] >> Yeah. Yeah. >> That's insane. >> Yeah. Yeah. Exactly. Uh but uh yeah, I just uh want to wish everybody a very happy Wednesday. Uh h happy rest of the week. Happy holidays for for folks. And uh yeah, we'll catch you back here for another great episode of Devantel on Friday. Um but yeah, uh Gerald, thank you again. I really appreciate you taking the time out of your day to join us today. >> Thank you so much for having me here. Um this was awesome and yeah, uh happy holidays to everyone. [laughter] >> Awesome. Awesome. Uh all right, all um have a good one. Cheers.