Chatted with João Garcia from Cartesi about where DeFi is actually headed

Added: 2026-05-09

[00:00:09]Hey everybody, welcome back. Um, today got kind of a special treat. I have Kraw. It's hard to say his name.

[00:00:16]Uh, Kraw from Cartesi.

[00:00:19]Um, Cartesi is a very interesting project and it's one of the reasons why I wanted to have him on and and kind of pick his brain on a lot of the stuff cuz to me it's it's a weird combination.

[00:00:32]It's so simple that it's stupid that no one else has done this years ago.

[00:00:39]Um, and it's technical stuff and I'm just a big dumb guy.

[00:00:44]I'm not, you know, a super smart person. I know a lot of super smart people and all of them all the developers I know everyone is just focused on Linux. I think and most regular lay people don't really even know about Linux.

[00:00:59]Um, but [clears throat] you know, and I've heard you say like basically everything that's built or that we use was built on Linux. Like your your video games that you use, your all that stuff was all comes from Linux.

[00:01:13]Um, and I like your you you said a a few things that I thought were very impactful. One of them was very impactful about life in general.

[00:01:25]Um, which I appreciate you coming up with that. I'll I'll tell you talk about that later. Um, that the the math code for mathing um, when it comes to DeFi and it's just like most people don't understand or know.

[00:01:41]I remember the the first day in my class in university, uh, years ago and Andreas Antonopoulos was my professor and I thought I'm in over my head, you know, taking this class. And he, uh, he was talking about the development of private keys and how, you know, the the computer does it and it's like all this complex multiplication, blah blah blah.

[00:02:11]And then someone raise your hand and like, "Well, can't you break it by using just complex division?"

[00:02:17]And he goes, "No, division doesn't exist in world." And I said, "What?"

[00:02:22]Um, yeah, I am over my head. This is the wrong class, man.

[00:02:27]But basically you you put it much much more simple that uh we're jump and I'm jumping ahead because it's like I said, it's just so basic and so simple. Can't believe no one's done it. And then the fact that you did it and you're doing it I think changes so many things and opens up so many opportunities um, that I wanted to hit on. But that the problem isn't the math or other things. The problem is the tools when it comes to DeFi, which is big big thing for me is DeFi. Um, that people can't create or develop because they can't use existing stuff, existing libraries, existing um, things cuz it's all it's it's all uh like specific computational math.

[00:03:15]Um, and they have to like re uh like re-simplify it or get with something like that to make it blockchain like compatible. Um, so a lot of stuff to go into. First let's just talk more about you and your uh experience, um, how you got into to blockchain, crypto and um, what led you to focus on DeFi infrastructure?

[00:03:41]Okay, so I think from the beginning, I was born in No, I'm joking. Uh, >> [laughter] >> but I was I was working before I I got into crypto in um Web2 company. We used to do we we we developed solutions mostly for the energy sector.

[00:04:01]And in there there was a project which was green energy credit where I had to start learning about uh crypto, right? I I had to start learning about tokenizing energy credits, carbon credits. So I started studying crypto and I started studying how blockchain is done, why it became a thing.

[00:04:26]And then I started to realize that the technology behind it made a lot of sense. Um, the security, the cryptography, it made a lot of sense to me. It sounded really cool and it was a really different way.

[00:04:40]And I I I always had that itch for security, you know, like when I was a when I was a kid, hacker was in one of my top uh wanna be dreams, you know. And then so it started to make sense to me. I started to enjoy the work.

[00:04:59]So I talked to some friends who worked at Cartesi and they explained to me what they were doing.

[00:05:05]And when they explained I I had the same reaction that you you just explained right now. It's it actually makes it so much better. It actually makes it viable to do the things that I want to do.

[00:05:19]Then I said, "Okay, let's let's start studying it. Let's let's take a look at it." And next step was just uh starting to engage with the ecosystem and discover what are the cool things that you can do with uh with blockchain and DeFi included. Um, I started to realize that our systems could be so much better for both uh Web2 and Web3. So decentralization was such a cool thing for me, such a cool concept. And at the same time the things that Web2 brings that we could bring to to Web3 uh with complex computation making uh DeFi more akin to TradFi in the sense that it's more powerful. So I started to really enjoy it. I think I think it was I think Cartesi was the perfect protocol to make me make that jump from oh, I find this really interesting and really cool to woah, I actually want to work with this.

[00:06:19]The this is what I want to do. This is what I want to develop. So yeah, we had a we had an opening for for developer advocacy and I really like to talk.

[00:06:33]So it it seemed to be a good match.

[00:06:37]Um, yeah, it's you said what this is the thing you said I think is real cool for just life in general.

[00:06:43]But you said DeFi doesn't need to be defined by its limitations.

[00:06:48]Yeah, so we we always look at ourselves as far as like limitations. So it I think it's good for life, but that it really is uh has been like a problem. Um, one of the things you're you're doing is you're uh you're you're working with the Push Protocol.

[00:07:05]Which to me that was like one of the biggest things years ago when it was you know, invented. It was that was one of the biggest problems is that you couldn't use crypto wallets for you know, uh push things like uh subscriptions and Netflix and things like that. Like everything had to be you know, you're your own bank and you're in control, but you're also the one that has to do everything. You know, you're the teller, you're the bank manager. So the Push Protocol were you able to that really changed the game. I thought, "Oh, it's over.

[00:07:34]You know, crypto's going to go mainstream now. It's here comes adoption." And it didn't happen. It was but the fact that you're you're working with that I think is really really cool.

[00:07:43]Um, let me see. Now introducing Linux-based runtime for blockchain.

[00:07:51]Um, what made you believe that this that approach would work better?

[00:07:58]You explain it.

[00:08:01]So what made you believe people?

[00:08:03]I think that really goes with with the the the thing that you just said about life. I think that we we we always should look for a more a better way to do things. It's like settling down is kind of bad for everyone and like accepting that accepting the limitations and not trying to do what you want, but what you can is is not is not great.

[00:08:32]Um, so I I think the the idea behind it was there are so many tools that are ready, that are available. There are so many things that we want to do.

[00:08:44]But for us to do all of those things, we won't be able we we we will take so much time to do each and every one of them individually. Yes, to do just one it's faster, but if you if you create something generic if if we create something that allows us to do what we like, what we have fun doing which we were used to this environment, right?

[00:09:08]We were used to Linux.

[00:09:10]Um, so that that was the motivation. It's Let's find the middle ground here where we can talk, but the the the the general idea of it is just that with with Linux, um, games run on Linux as you mentioned. Uh, bank systems frequently run on Linux. Databases, they're always built with Linux on in mind if they're not built on Linux for Linux.

[00:09:38]Um, most uh back-end systems which was is where most of the logic of every software runs, it's running there. Uh, it's like I don't know the the the exact number, so I'm just going to do hard guess here, an educated guess, but 90% of servers that provide any logic for everything that we use day-to-day is is Linux hosted, right?

[00:10:02]So, at least 90% of the software that we use day-to-day. Of course, most people use Linux or use Windows or Mac OS on their personal computers day-to-day, but they are using Linux indirectly because they are communicating with a web server somewhere. Yeah.

[00:10:20]So, if we need to perform logic on chain, and that's the idea with smart contracts, we need to move this logic part there.

[00:10:31]So, we need to be able to create this smart contracts. We need to be able to create this infrastructure with the tools that we always use. So, I think that's the that's the core.

[00:10:41]I think that's a really good point. You know, with the tools that we've always used and been able to to do stuff with. It's kind of like this technology, the blockchain, this new technology comes and it's like you can build this the best, fastest sports car there is now, and you're like, "Okay, great.

[00:10:59]Hey, let's reinvent these wheels first."

[00:11:02]And you're like, "Why? The wheels are fine. We could, you know, let's go to the other stuff." Um you know, use the things that are already already working.

[00:11:10]Um And and there's more, right? It it brings compatibility with everything else that will come in the future.

[00:11:19]Because if someone creates something that's really smart and really cool for web two in the future, it will work on web three because it will already have compatibility if we're able to run Linux there, right? So, it's not only backwards compatible, it's like forward compatible, future compatible as well.

[00:11:38]Which is fun.

[00:11:40]So, And it's kind of like AR-15s or so cuz everything's so modular and fits together. You know what I mean? You can use any different parts.

[00:11:49]Um I did want to ask some something about your uh cuz there's some things that are kind of revolutionary that I think Cartesi's doing.

[00:12:01]It seems to me like there it's like a like a and I may be wrong, but it's kind of like a a sandbox kind of thing where you are able to uh or like a cell phone where people are able to build or run like dApps, apps, and they don't affect other systems.

[00:12:21]It's all kind of more contained to that one thing that they're building. Is that right?

[00:12:25]Yeah, exactly. It's Do Do you remember the I think it was CryptoKitties?

[00:12:32]Mhm. That was a few years ago and how it uh um kind of made gas extremely expensive, no one could actually use the chain. So, the whole idea behind that is that if we isolate applications in a way that the processing of one application does not compete with the other one, right?

[00:12:54]It's like this idea that Ethereum is the world computer. So, everything will be run in this computer. But what if we could separate processors Mhm.

[00:13:03]>> for each different application, right?

[00:13:05]So, each application processes in its own separate little module here. That's what we call a an app chain or or an app-specific roll-up. Okay.

[00:13:16]>> So, the idea is that we you as an old runner, you only run the applications that you are interested in validating. And you won't run the other ones. That's advantages for you because you only care for what you care. And that's advantage for the users as well because they don't have to compete in gas prices for every other application that they don't care about.

[00:13:43]I think that's revolutionary. That's very cool. That's one of the things I thought was awesome.

[00:13:48]Um So, expanding on that, and I'm I really try to not get too technical so that more people will keep watching. Um you've made it to where you're able to delegate the deterministic calculations um to the host machine. So, you're actually able to delegate like computer power so that it's kind of expanding on what you just said. Mhm. Um That that to me means it would have much bigger impact. Like just the fact that you said you're not the gas competition, all that. But what about things like the broader picture like data centers and all the infrastructure that's like, "We have to build this now because this is the way it is." But if you can tailor stuff, isn't that need going to be less?

[00:14:39]Yeah. Uh the idea is that whenever you can use more resources from the machine that you're running, you are you will be able to do things better, do things faster, do more complex stuff. Um And the dream is even running LLMs, uh AI agents, powerful AI agents on chain in a verifiable way.

[00:15:03]Of course, there are some tweaks that have to be done to be able to do that.

[00:15:07]And I mean, not some tweaks. Uh to be able to do this delegating to the host machine thing, there are some tweaks that have has to be done. But for and for AI models, there are lots of tweaks.

[00:15:18]It's a extremely complex thing that we're trying to do. But uh we were able to do some experimentations, and we were very happy with the results because it was something that even for us, it was it we we were not absolutely sure it would work. But when it worked, we were, "Wow, okay. So, there is a path here.

[00:15:39]>> [laughter] >> There is a very cool path to to to be able to make this this thing that we like on blockchain so much more scalable.

[00:15:50]Um Yeah, it it it's taking advantage of every resource that you have, not >> It It seems like you're really just amping up efficiency so much.

[00:16:03]Exactly.

[00:16:04]Exactly. Um >> Yeah, yeah. Like uh trying to to not go technical is as you said, we have a we have a car, but when we're using the virtual machine that Ethereum does that uses, you are kind of always always always uh stuck to a speed limit.

[00:16:25]We would be able to break through because we would you be using the whole engine, not just a little part of the engine that is delegated towards them, you know?

[00:16:35]Yeah, it Like I said, it's just just better efficiency means every like every aspect of it just runs better, runs cheaper, runs faster.

[00:16:45]Um Okay.

[00:16:49]Uh Well, there was something else I wanted to You don't know, but I'm thinking I wrote it down. Um And also one of the big things is that that developers are able to use all the all all the languages that they're used to writing with with Cartesi, which is Like I said, it's just you have to seems like everything's been made to make just make it as hard and as complex as possible to build, you know? And you're just came and said, "Hey, why don't we just do this easier way? It's faster and more efficient." And just like I said, it's just yeah.

[00:17:31]There's a So, when Ethereum come to life, it was one idea, right? It was one It It was just, "Hey, what if we could run this this software there?" And for that, they had to put some constraints to be able to make everyone be able to run it.

[00:17:49]What we did is we took this one part of Ethereum that we thought could be improved on a lot, and we made it so much better.

[00:18:00]Um we're not We didn't replace it because we still think that it has its value in the way it works. It has its advantages. But we made it you able to expand on it. So, if you want to delegate part to it, you can. If you if you are okay with using the limited part because you want to do something simple, you also can.

[00:18:21]So, you are free to do whatever you want. It's not a restriction. It's That's why we didn't go to a towards designing a full blockchain, you know? We we went with let's just design and fix the one part where where we are really good at, which is this what we call the execution environment. And it felt it felt to us that a team can only do so much. So, probably when creating Ethereum, that was not the major focus, the execution environment.

[00:18:58]Vitalik even kind of wrote piece that he mentions not exactly that, but there is a space to improve it. There is a a space for us to replace the Ethereum virtual machine with a different to work in a different way.

[00:19:15]And it's kind of what we thought all along. It's yes, you made these decisions because at the time they sounded right, and they sounded enough.

[00:19:23]But as time moves, we want better things. So, um that's that's where it came from. It came from from this need to not be um limited by design choices that were made to prove that something was viable.

[00:19:42]Viable. Um it was designed So, now that it's proven it's viable, let's do it in the best way possible.

[00:19:49]So, our focus was always making the this part, this execution environment in the best way possible and not what is viable at the time.

[00:20:01]So, that that that was the jump, I would say.

[00:20:06]You tell AI, you mentioned AI in uh I've heard you mention verifiable AI.

[00:20:13]Yeah. Uh using this um you can have verifiable AI.

[00:20:20]Uh what exactly do you mean by that?

[00:20:23]So, um it really depends because AI uh these days people have been thinking of AI immediately like LLMs specifically. People already think immediately ChatGPT, Claude, you know?

[00:20:38]AI is much more than that and at the same time uh there are much simpler things than that that are still called AI, right? So, you have AI models to create to calculate uh risk for example.

[00:20:56]So, one very cool use of AI on chain is using linear regressions. Uh it's using decision trees uh which are basically a tree that you go down just like a family tree seeing the branches and making the decisions that you have based on uh information of that uh individual. So, we could we can use those mechanisms to calculate and define and figure out uh risks for example in investments or in loans or in anything.

[00:21:35]And one of the cool things of it is that some AI models um are black boxed.

[00:21:43]You don't know how they operate inside.

[00:21:46]So, it's hard to reverse engineer and understand how it reached that conclusion.

[00:21:51]But with it being verifiable the first cool thing is you can always reproduce it to know for every person that use it why how it behaved. So, you can you can double check it.

[00:22:07]Um and you can make it in a way that you are absolutely sure that no one is messing with the results.

[00:22:16]So, you were running those algorithms that are so important for you to decide what to do with your money, with your life in a way that you are sure that it's not being tampered with. Mhm.

[00:22:29]>> [clears throat] >> Which is a very important.

[00:22:33]Um and talking about money, like I said DeFi I think is very important. I think that's it needs to be one of the biggest parts of blockchain and I just keep I keep thinking it's our time, it's our time.

[00:22:46]Um you said that DeFi hasn't been uh limited by imagination but by execution environments or the tools. Um what do you think is the um the most primitive thing and what do you think is is uh is still missing when it comes to it?

[00:23:06]I think the most basic thing it misses is being able to deal with floating points.

[00:23:12]And that that means not I that means numbers that are not rounded where you have decimal parts.

[00:23:23]If you can't do that, you have to approximate everything. And when you have to approximate everything, you have two consequences.

[00:23:32]Uh first one is your result is not as precise because you're approximating things. Mhm. And second thing, it because you have to approximate and go out of your way to do these calculations you tend to spend more on calculations than you should. So, the service you're providing becomes more expensive for everyone.

[00:23:57]Um let's say there's a token sale for example, right?

[00:24:02]One of the things that we use is bonding curves which you it's a mathematical curve that you calculate to define the price of the token given the number of purchases that happened.

[00:24:14]Uh you cannot calculate a smooth curve with without decimal numbers.

[00:24:23]You have to approximate. And when you are approximating, you're making every operation more expensive. So, gas is more expensive, computational power is more expensive, there is more going to waste uh and it's not as precise, so you're not dealing with necessarily the correct numbers. That is just one of something to example It's basic though.

[00:24:44]That's so basic. Yeah. Mind-blowing.

[00:24:48]>> [laughter] >> And you said that you say that there's no real math, there's no real finance.

[00:24:54]Which having been part of like, you know, building exchanges and I'm good friends with a guy that used to that he wrote the software for the Chicago Stock Exchange and all stock exchanges use it now.

[00:25:08]Um and uh the guy that was the CEO of the Chicago Stock Exchange and man, there it's it's it's like such fractions of fractions of fractions that you know, whenever we're building stuff or talking about it that's so important.

[00:25:24]If you would said, "Oh, you got to round that."

[00:25:27]Their heads would explode.

[00:25:28]>> [laughter] >> What So, what do you mean no no real no real math, no real finance?

[00:25:35]What are like that one that's one of the real consequences. What are the limitations for users and markets and you know, stuff like that today do you think still is out there that needs to be fixed? I think I think we have been for years now uh even we even have hard time thinking of this, right? Because uh if you look at DeFi, we were limited. Uh if we look at blockchain, we we are very limited in the solutions that we've been doing.

[00:26:09]Kind of like it's always such a small step when we see a new DEX that does something different, right?

[00:26:16]It's always such a small step when we see a new platform, a new uh prediction market that does does something a little bit different.

[00:26:25]And I think that that's why why I say we don't have real finance because we're so constrained by our we're we're so used to be constrained that we don't try to think outside of the box so much. Um it lately I've been hearing so much about just uh real world assets, about uh the same sort of DEX that we always do.

[00:26:53]Uh so many of them using resources that are that are not real.

[00:27:00]They are either not real finance as I mentioned because we are rounding things down, we are rounding things up, we are just taking shortcuts or they are not really decentralized because many of those solutions are just using oracles and centralized pieces of information to discover things, to make some calculations.

[00:27:26]And that's where the danger and kind of lies as well. Because so many platforms, so many solutions kind of they don't I won't say pretend but they intend to do something in the right way but they have to rely on TradFi to do DeFi instead of actually running the the the finance on the decentralized man, right?

[00:27:55]I think that's the thing.

[00:27:58]Okay. Um the you talk about moving from a gas optimized finance to computation driven finance. Yeah. What is what do you mean by that? And what is what would it look like?

[00:28:11]Instead of us instead of us simply running an approximation of a logarithm, we would be able to use the whole power of your of your system to run financial algorithms. So, state changes calculate the impact of millions of operations after uh one move that you make. So, I take a loan with you and in that simulation of me taking a loan with you, we see how that affects in a cascade effect the market as a whole. So, there's a whole simulation being done there and figuring out the many possible branches of everything that will happen when I take that loan with you. So, how the people that you owe to will see uh the prediction of your investment going up and knowing that I don't know, I have an asset, people want to invest in that asset. People can simulate and see the different cases of each person that is investing in that asset.

[00:29:24]And all at the same time in a very powerful way. It's like have you ever seen Avengers?

[00:29:31]Yeah.

[00:29:32]You So, there's that scene where Doctor Strange he says like, "I've seen 14 million futures." Yeah. Kind of like that.

[00:29:41]>> [laughter] >> That's my that's my idea. But with better information, more tools, and better efficiency. It's kind of like a cascade effect, like you said. It it like the improvement goes all out, like a like a ripple through a pond. Um so I mean, talking about that, if if DeFi did have access to like the full-scale computation, all the tools, everything, what types of financial products or systems do you think would be able to come out of that?

[00:30:12]I really like uh risk platforms, uh fully autonomous like hedge funds, where portfolios uh are not managed by humans, but there's an on-chain machine learning model that is like gathering the data, adjusting risk, executing trades programmatically, um or uh uh uh an an insurance, uh so you we would have like uh models running on chain assessing risk, paying out uh claims fairly, uh and we would be able with I don't know. I think w- deal with real estate and corporate debt, um compliance, and run so many complex pricing engines for corporate things that it would become so much more reliable for people who are used to live in this world, right? That that that that's how I see it. That that are some of the use cases that I I I see being possible, not naming the exact applications, but general use cases.

[00:31:24]So, if you want to build something, you can go to uh cartesi.io.

[00:31:28]Um there all the information's there on it. You mean, there's people building so many things on games, um as well as just other stuff. I you know, like DeFi, I really I I had a lot of fun looking through his watching some of the videos, all the you just walking around with a little microphone tiny mic there, just asking people what they're building.

[00:31:51]Um >> There's There's one application that I find really cool that a guy did a while ago, which is called DCA monster, where it's like you're DCA-ing on chain, but with minimum granularity. So, instead of doing like on long batches, you're every second you're DCA-ing, uh which is really fun.

[00:32:10]I think the brother and sister made it so uh it's uh you get like spam calls and stuff on your phone, it automatically uh you know, like tags them. And they said that like the phone companies were doing that for themselves.

[00:32:25]And then they got inspiration for that.

[00:32:27]It's like, hey, that way when if, you know, an address comes in or something that most likely looks like that was spam, or it can pull cuz we do that on our own, just takes time.

[00:32:37]If I get a phone call or if I get, you know, an email, I just put it, you know, search it.

[00:32:42]And then stuff usually come up. Oh, I got, you know, hit by this email, this person ripped me off, or, you know.

[00:32:48]But that would be able to to pull from all that and give you just instant information. I thought that was pretty neat.

[00:32:54]Um Also, I uh I'm on World Blockchain Roundtable. It's every Thursday, hosted by Dragonchain.

[00:33:02]Like to have you on there sometime. Uh we should probably talk about DeFi or something like that. I I'll message you have you on there.

[00:33:09]Um I think you would you would be good to have you have you on there and have your insights.

[00:33:14]Um anything else you wanted to go over? I think I hit all the stuff that I wanted to to hit. Uh there's a lot more about Cartesi that, you know, you could people can go check out. Um especially if you want to build something.

[00:33:26]Uh it really is making it so much easier for people that like know the basics. They don't have to relearn stuff. It's it's so simple, it's genius.

[00:33:37]Um Anything else? Something that you mentioned with about like people that are using the basics and they don't have to relearn stuff. That's also true for AI today. People are building so much with AI, uh Codex, Claude, and so [snorts] many so many different tools that basically develop everything for you. So, now having these AI models that already know how to develop so much on this Linux world and being able to bring this to web three, it's a really fun experience to just hit a prompt and see it bring to life like web two on web three. Um and thankfully Cartesi enables that because doing that even the AI AI models would have a hard time doing that without uh uh this infrastructure that is >> have access to the infrastructure.

[00:34:31]Well, it has been cool hanging out with you. Um there'll be links below to uh Cartesi and all their their socials and everything. Um you guys take care.

[00:34:42]It was good hanging out with you, too.

[00:34:43]Uh I'll see you all in the next video.

[00:34:46]Thank you. Thank you for having me, man.

[00:34:48]That was awesome. It was good hanging out.

[00:34:54]>> [music]