Cold Open And Show Setup

Joao Garcia 0:00

So it stops being about hey, what can I create on chain that would be useful? And it starts being, hey, I would love to bring this to chain. It was hard before, now it's being made easy.

Joeri Billast 0:15

Hello everyone, and welcome to the Web3 CMO Stories Podcast. My name is Joeri Bilast, I'm your podcast host. And today I'm excited to be joined by Joao. Hey Joao, how are you? Hey, I'm good. It's a nice day. Very nice day here in Real. How's everything there? All good. I would have said Bombia, but most of my listeners know they expect an English podcast. I'm learning Portuguese at the moment. Guys, if you're now wondering, who is Joao? Well, actually, two years ago, I had Cartesy co-founder Eric de Mora on the podcast to explore Cartesi's model lab blockchain vision. And today I'm excited to go one layer deeper with Joao. Joao Garcia, Devrel lead at Cartesi, and let's focus on some execution environments, DeFi, hidden computational limits, and how to explain these complex ideas to different audiences.

Making Web3 Understandable

Joeri Billast 1:14

So, Joao, you recently launched three levels deep. Where you explain Cartesi concepts to a kid, an engineering student, and a PhD. What has that taught you about making complex Web3 IDs understandable without dumping them down?

Joao Garcia 1:33

It's actually very challenging to make them understandable. Honestly, most of the time I'm discussing with people, and even when people just ask what do I do, it's hard to explain because they have to have all this context in the background of what is a blockchain, what it does, and then there are the deeper layers right? There are so many very small concepts that people have no idea that they would have to grasp to understand the whole thing. So sometimes the thing we're going to is analogies, and analogies seem like very simple, but they are a really good way to make you actually at least get a grasp on what's going on behind the scenes.

Joeri Billast 2:17

Yeah, for me it works too. Of course, I'm into Web3, I'm into AI, it's a bit that's technical most of the time, but I need to explain what's a blockchain to someone else, sometimes even to my accountant, and then it helps to use metaphors or concepts that they already know. When you

Linux On-Chain Explained

Joeri Billast 2:34

explain Cartes execution environment to a non-technical founder, what is the simplest way to describe why Linux on-chain matters?

Joao Garcia 2:45

I think the most important thing that we can focus when we're talking about this is not necessarily how it works, but why it is important, right? And I think it's that everything that we built so far in history of technology development, not everything, but most of the things were focused on Linux. And every application that runs online runs on two sites, the client and the server. And most servers are Linux servers. Most of the logic of the background is running on Linux servers. It's funny because most of the people at their homes have Windows, right? But Linux is just this whole other bist that it's on our lives all the time without anyone seeing it. And maybe that's the thing about blockchains. Maybe that's how they will be integrated in the world in the future as well. But the idea of us running Linux is that we enable everything that was built and so many things that will still be built to run on-chain as well. So we are basically enabling blockchain functionalities that were blocked to developers that are used to doing things with these tools. So we are allowing people to develop with Python, we are allowing people to use databases on their systems, we're allowing them to use a file system, which are things that people that are not developing don't know the importance of, but it does allow the developer to feel not only feel more comfortable, but to be capable of developing something without having to spend without having to recreate years of development to get to the point where we were.

Joeri Billast 4:28

Yeah.

DeFi Built Around EVM Limits

Joeri Billast 4:29

Now, DeFi has already achieved a lot with Uniswap, AV, Curve, and other protocols. So where do you where exactly do you see the hidden cost of building finance inside EVM's limitations?

Joao Garcia 4:46

I think there's uh human cost in terms of there is so much that has to be done. There is uh sanity cost in the sense that you have to trust the technology that you're using, right? The whole idea of having something that's verifiable by a whole network is that you want people to trust that solution or to trust that you don't have to trust one individual, but the technology itself. And when we take that away, the trust away, it it gets confusing for the users, right? So when you're creating something new, you have to trust that the developer created in the correct way. He's creating so many algorithms for trading, calculations for balances, and all of this that is already defined in mathematical libraries, it's already defined on financial libraries, is now available to be used on-chain. So things that were tried and tested for a long time, right? So I think there is this huge cost that people don't realize that when we are creating things on-chain, there is a whole level of rigor that we that we need that it's hard to prove. And when we use things that have been used before, we already inherit the security of those things, we inherit the safeness of those things. And there is also, of course, the way that the infrastructure works that brings more cost with gas. And if we are optimizing, if we have an optimized environment, there's also financial cost that is saved a lot when we use a proper execution environment that is more focused on computing itself, right? So I think there is this two these two pieces together.

Finance-First Execution Environments

Joeri Billast 6:43

Now, in your piece why DeFi needs a better execution environment, you argue that DeFi is often designed around what the machine can handle. Would change if the machine was designed around finance instead.

Joao Garcia 6:58

Yes. So I think that if the machine is designed towards the solution and not the solution designed towards the machine, we get cheaper cost of execution and we get a myriad of possibilities. You know, I like to doing a small comparison here. It's like when you get a computer that has the hardware and the software made by the same company, it tends to perform better than the equivalent hardware that's running a software from another company, right? It's kind of like that because you are using the software that was developed to run on that kind of technology. So soft mathematical financial software was designed to run on the Linux machine. It was designed to run on that kind of operational system. Now you don't have you don't have software, you you don't have a design that was thought out in DeFi that's not recent, because DeFi is a recent thing, right? So Solidity was not made in the EVM or even other execution environments for blockchain. They were not created with the idea of how do we make the things that we want to work work. It was made with the idea how we guarantee the security, and then all the work is gonna be on the developer that's creating the solo. So we are trying to take that away. We are trying to say we can have the best of both words, we can have something that is secure and is tried, and put it together with something that's creative and is useful and has always proved itself to be useful, right? So it stops being about hey, what can I create on chain that would be useful? And it starts being, hey, I would love to bring this to chain. It was hard before, now it's being made easy. No.

Joeri Billast 9:21

Many people talk about DeFi's next chapter in terms of better UX, more liquidity. Why do you believe that the runtime itself is one of the most overlooked parts of the conversation?

Joao Garcia 9:35

I think there are many parts that are talked a lot. You mentioned UX, right there. UX is something that I'm very critical about of Web3 in general. It's really hard for people to get to learn about, but I think we will be able to get a more abstract solution, a more ubiquitous solution where people don't need to realize that they are living on-chain. But at the same time, I think the way that we followed to get to this life that we live now, that we don't realize we're using technology all the time, we don't realize the data going back and forth, and we don't realize how things are processing was with a great technical advancement in the system architecture. And a big piece of that was the execution environment. So Cortez is not here to solve it all. It's just that we are really good at the execution environment, and we know it's a very big, very important part of the stack. I think it's overlooked because people started feeling what is the most hypeful thing, what is the most exciting thing, and they tend to overfocus on that. And reinventing the will in terms of execution is not that glamorous, it's not that marketable necessarily. But it's one of the most critical parts, right? And we live in a social era, and people kind of chase the hype real hard. And I think we have the vision that someone has to not chase the hype but to make the whole thing work. And I think we're focusing on what is really good from us.

Joeri Billast 11:28

Okay. I think you also use the idea that smart contracts today behave like basic calculators. While Web3 needs a real computer. So where does that analogy become especially important for financial applications?

Joao Garcia 11:46

I mean, there are calculations that you can do. I'm not trying to say you can't do. Of course, you can do things like logarithms on smart contracts. But they're hard to be done, you know, or they cost a lot of operations to be performed, right? The number of cycles a regular processor does in seconds is really high. And the operations are not counted one by one. On Ethereum virtual machine, it's a little more clunky than that. So you have to perform so many operations, and the gas cost starts to pile up when you do it. The way we design the solution, the way we design with specific roll-ups to run this execution environment,

Smart Contracts Need Real Computers

Joao Garcia 12:32

it allows us to not have this very huge processing cost across all the network. And this makes it feasible, and this kind of helps with UX too, which was something that we were just talking about. It kind of helps with UX too in the sense that it's good that you can do an operation without having to pay so much, right? To simulate a result to get to a result that you want to achieve. So it gets cheaper for you to use the solution. Or it gets complex in the side of the developer that they have to create something that is really robust and prove the robustness of it. And that's getting harder to do today as time goes by. So the comparison I like to do is that you don't simply create the math that you you don't simply create the libraries and the calculations that your program uses. Most of the time, you delegate it to someone who's very specialized to do it. So in Python, you use PyTorch for scientific programming, or you use Quantlib, which is a very big library for finance as well. And those things they allow you to simplify the way. And that's why the comparison goes.

Joeri Billast 14:13

Yeah, absolutely. Now, another question I had is can you explain the difference between gas-optimized finance and computation-driven finance and why the distinction matters for founders, builders, users?

Joao Garcia 14:30

Okay, so that is very limb, but when we're doing something like a token launch, for example, right? You follow a price of the token by the amount of assets that have been sold. A very common strategy here is using something that's called a bonding curve. Right? A bonding curve is basically what dictates the price of the token given the number of purchases that happen. When you don't have a full mathematical library backing you, you kind of have to do some weird approximations and force some results to fit this curve, this chart, right? But when you do have these libraries,

Compute-Driven DeFi In Practice

Joao Garcia 15:18

you know that you're calculating and approximating much better with much more precision. You have levels, depth of levels of precision to get to have the best precision with the least cost. Because it's a very small marginal gain that makes a difference that you have to compute so many more steps when you're doing it on Solidity. But when you're doing it on Python, for example, you don't have to stress about this marginal gain because you're delegating it towards a full CPU that is able to actually compute that very fast. And there is very little difference in the number of steps that the CPU takes. So it's not a very small difference, but what translates to the cost of processing in the end of the day, the gas cost that the users pay, and the cost that the validator who's running the node is paying, it gets much smaller when we use this technology, right? So the computing-driven environment, one of the key pieces of it is that it is taught out to not have competing applications wasting resources off each other as well. So when you're running something on Ethereum, everything is running there. And then you launch an application, crypto GitIs, the NFT creates, and you completely saturate of the chain, and then you have an issue because other applications get more expensive. And when you're saying okay, it has this application-specific rollup that runs this application that is very important for my financial, for my personal account, my personal investments, then you need this specialized computer working for it. You know, you cannot necessarily be sharing this space with everyone. Because a surging price may really make it difficult for you to get your assets out of it, or operate, or liquidate, or do whatever you have to do on that moment. And no one wants this kind of surprises. Then you can iterate much better, you know, because you have a full processor for your application, it's one per application, then you don't need to optimize in a way that it won't be expensive for you to just make a withdrawal, or you don't need to operate in a way that you cannot update everything in the ecosystem or everyone's wallets, or make a huge state update on what everyone has because market is rising, market is going down. You have a much smaller cost embedded to this big stateful operation, as we call. Because you have to alter the values of everything that everyone has, right? So you have like a prediction market and you have to calculate results and gains of everyone, and then the gains they can be more abstract, they don't have to be simple, it's not one person that won and one person that lost. It can be divided and shared across so many accounts because you can operate on such a large database of people that have money invested there. That's what I mean when it gets different. You can make more complex operations, you can make lending protocols that you take out not from one person and give to the other, but from 100 of people, and then you give back to another hundred of people, and you're iterating so much and so much faster. And that's what this finance-specific computation allows you to do.

Joeri Billast 19:08

Okay. Now let's talk about trust. It's a topic that is in every podcast episode these days. I've written a book. Uh, I also speak about trust and for the marketing side, the future CMO. I talk about a shift from attention to trust. Now, for technical infrastructure projects like Cartesi, how do you build trust when the most important value is actually invisible to most users?

Joao Garcia 19:36

That's actually very hard. The most important thing that we always done is that we always try to not reinvent the wheel, right? Even when I speak about Kartesi in the sense of we want you to be able to use the old libraries, we want to you to be able to use the old, not the new libraries as well that are being created for things that already exist. We aim for this discourse of not reinventing the wheel because we want to stand on the shoulder of giants. We know that Ethereum, for example, is a very safe place and we trust their security. So we're gonna

Trust, Proofs, And Settlement

Joao Garcia 20:12

instead of recreating it, forking it, and creating our own version or creating our own L1, we want people to feel safe using what we have. So they need to use Ethereum as a settlement layer. So what we did with a lot of research was figure out a way to settle there instead of creating the new thing. So we created a very robust fraud-proof algorithm, right? And our way of building trust was always with science. A few of Cartazi founders are very research-oriented. Some of them are university teachers. So all that we do is very strongly research-backed. And I think a big part of trust that people have on us comes from knowing that we're doing very serious mathematics in the background. So we use this math and we use this giants that already exist that created something big, so we can trust each piece individually as well. Not only sell what we do. It's much more about proving what we do. That's the important thing about.

Joeri Billast 21:19

Yeah. Now when I speak about trust, it's also often in AI context, because part of what I'm doing in my work, also with my retreat that I'm having here in CINTR, I am exploring how AI is changing strategy, execution, visibility. But how do you see AI assisted development tools like cursor, clot, and Cartesis execution layer, changing how Web3 applications get built?

Joao Garcia 21:50

I think it's exponential growth in there because AIs are very good to use what is well documented and what is well used. Right. And when we created an environment that makes everything that was built so far available for you, then AIs can drive so much. I personally am a big fan of spec-driven development. Have you heard of it? No, but I can imagine, yeah. Yeah, it's basically there there are some steps for you before you start. Go for assisted coding.

AI Tools And Spec-Driven Building

Joao Garcia 22:23

And it's very oriented towards what you prompt before, right? It's how to structure your prompts, how to structure your project even before you start doing things. And personally, I think some methodologies that are proven they work really well. Spec-driven development is something that has been working really well for me. I know a few other people on Cartes use it as well. And it's something that really makes us more confident in what is being brought to us. But it's I I do think that it helps to have a sense of safety in all the speed that people are developing nowadays.

Joeri Billast 23:04

Indeed, because you know, I come from before I was into marketing AI, I was a project manager business analyst, also preparing specs, of course. And when you've done the work instead of going speed, speed, speed, and you did some work on the specs, then you have a layer to build. The next phase upon it doesn't mean you cannot be agile, but yeah, I totally understand what you're saying. Exactly. So you're in uh defral, it's uh partly education, but also partly community, I would say, partly narrative. How do you balance technical accuracy with making the story simple enough for people to act on?

Joao Garcia 23:46

Yeah, so technical accuracy is always one of our pillars. We always say that whatever we do, whatever we say, we have to speak the truth, right? So everything starts from that. And then there is a lot of logic behind how can we present something that is so complex, so hard to say, so hard to define to people keeping the technical accuracy there, you know, and still making people wanting to engage, making them interested. So this is a huge challenge, but one of the things that we came up with, this three-levels challenge that you mentioned, is actually a solution for that because there

Community Layers And Where To Start

Joao Garcia 24:31

we can make all the analogies that we want, explain in the very simple way, but with progress the content towards this more robust and much more complex way of thinking, right? So I start speaking about how you navigate in a road driving your car, and then I finish saying, Okay, because this bite and this bite, you compare them, you know. So I think it has always to be gradative, but it's always important to speak to all of the community, those who are very technical oriented and those who are not. But in general, the first step is the window where people go through is the more simplified way. It's a more abstract thing. And then we have layers inside our documentations, we have layers inside our resources, and we have deeper discussions, even in our Discord. If you go to our community, you can see we have the community-oriented channels where people engage with ideas, discuss what they would like to see, what they would like to build. And then we have the research and development section, a little bit more on the bottom, because as people navigate, they go towards it. So there is some study and uh UX of it as well, how you navigate through. So when you go through these research-oriented channels, you start seeing discussions of core contributors there. I think this is how we should build, this is the correct way. And then there are even deeper levels of threads within threads of how something is implemented, and then linked towards GitHub repositories where we have even more in-depth solutions of not how things should be on a theoretical level, but how they are implemented, and then you can follow everything there. So I think for more specialized people, they get closer to the environment where they want to be: a small chat, a specific forum, a GitHub repository, and they can navigate through it. And we put a very big effort on this way we organize things because it's very important for people to be able to navigate through different levels within our community. Even more abstract in our Discord, we have our Telegram chat where people are explaining things. We have their AI tools that are constantly replying to people and clearing their doubts. We have me and Cynthia there, always there to talk. So it's a very open for every level, it's a very open forum for every level of discussion that anyone wants to have.

Joeri Billast 27:08

Yeah, Joao, the time went really fast. We're coming towards the end of the podcast. You mentioned Discord, you mentioned you and Cynthia. So if now my listeners they want to learn more about Cartesi, about everything that you are building, where would you like me to send them?

Joao Garcia 27:24

Yeah, I would say send them to our website and our Discord. I think that's the best place to get started. Follow us on socials. Because we're creating some good content on X lately, you know. Uh see people white coding, it's being nice.

Joeri Billast 27:39

Yeah, actually, I've been doing some research. I've seen you quite active. There are quite some videos and you uh doing your thing. Um, I'll ask my listeners know, Joao. There are always show notes, and everything you mentioned, the links will be found in there, as always. Well, Joao, it was really a pleasure to have you on the show.

Joao Garcia 28:00

Great being here, thank you for having me actually.

Joeri Billast 28:03

Great, weather amazing episodes. I'm sure you know people around you that could benefit from this episode, so be sure to share this episode with them. That are in the same field and that could really learn and would be interested in what Cartesi has to offer. If you're not yet following the show, this is a really good moment to hit the subscribe button. If you haven't given me a review yet, give me 5 stars because it really goes a long way. And of course, I would love to see you back next time. Take care. A special thanks as well to Leo, the main sponsor of Feb3CMO Stories. Support the conversations at the intersection of Feb3 AI and the future of digital leadership.