XYO Layer One achieved 2-5x throughput increase, Arie Trouw explains

0:00I'm Ashton Addison from the Crypto Coin Show, and today on Blockchain Interviews with Ari Traub, co-founder of XYO, CEO of XYO Labs, building XYO Layer 1 network in the intersection of DePIN, decentralized physical infrastructure networks, data, and AI, and blockchain. Being at the forefront of it since we first started following XYO in 2018, before

0:26before DePIN was even a thing, and they've been going strong with the 2025 release of the XYO Layer 1 network as well, and the intersection of AI as that sector continues to grow. There's a lot going on, and excited to dive into all that today. Ari, thank you so much for taking the time. Thank you, Ashton. Yeah, excited to hear the latest on the growth of not just XYO Labs, but the

0:54Layer 1 and the recent updates to the network that has made it even faster and cheaper than it was. It was it was already good from what I understand since the launch in Q3 2025, but the team continues to improve, and AI is only getting better, and DePIN is only growing more into tokenization, data verification, which I think AI

1:21plays a big role in as well. Maybe we can start with a high level on how exactly XYO ties in AI with data verification and DePIN, and you can sort of break that down in layman's terms a little bit more. XYO the protocol and XYO 1 the Layer 1 that we have it's always been based on data storage and providing provenance for data and verification for data

1:49being on chain so that way basically it's immutable and unchangeable. And with AI coming around there's more and more need for this, and also there's much more data being being used and consumed in there's a lot as far as you know is it actually consuming and using the correct data. Is it using data that it's supposed to or allowed to? Making sure that data

2:09that's supposed to stay private stays private, and you know the data that's supposed to be public is public. So, we really focused on those different aspects all along, and now with AI coming along we've become more of a linchpin as far as the need goes to be able to do those different things, and AI calls for those. So, we've been very excited about the AI growth, and

2:30also the ability we have to put things on chain and to add provenance to it without having to put the entire item on chain, for example. So, our efficiency for storage is extremely high compared to most blockchains out there. Mhm. Yeah, and you mentioned there the private side and public side with data verification, especially with the increased use of AI with everyday people

2:55and small businesses, more people are starting to think about not just the data verification, but the security of their data and the privacy of it because as AI is working so well, we're just giving more and more of our private and confidential data into these AI systems, and there's permissionless blockchains, but there's also permission blockchains or like

3:17public and private blockchains. Where does that fit in where you have maybe data data verification or location data that's public and then private and balancing those on a blockchain network? So, what we do is we really separate the actual storage of the data and then the metadata for the data. And primarily the metadata goes onto the blockchain. So,

3:40for example, the fact that it was stored its hash, its state, and time, for example. You can also have so a reduced version of the data on there. So, say for example, the location can be very precise in the stored data that's in a data lake that's private, but then on the blockchain you can actually store a much less precise location. So, it's like you might know

4:02it might be in California somewhere, but not where specifically. And then the thing which is nice about our network specifically is that once it gets put on chain and that provenance is stored there and the permanence is stored there, you can later on reveal it and show it. So, it's you don't have to go and prove again in the future that it was correct. And so, the actual data

4:25itself, the full data was actually preserved as far as being being persisted and made permanent on the chain earlier on without having to reveal the entire thing. So, it's basically it's optional and privacy, and it's it's reveal on demand privacy basically. Mhm. Can you elaborate on probably the 95 to 99% of businesses that are putting a lot of their data into these AI systems and

4:51the LLMs and ones that are owned by you know, the oligarchy of the FANG of AI companies. Why should they be thinking about data verification on blockchain or the privacy aspect? Well, on the verification side of things, you know, you're obviously when you're procuring models and you're learning learning about bad things it's going to make something

5:14inaccurate. So, you want to have purity of data when it comes in. You know, that's one of the concerns that's always been about AI is like, well, it's really good at reading the internet, and as everyone knows not everything on the internet's true necessarily. So, there's some verification that goes on there because if it just regurgitates what's being reported on or what's being

5:30said on X or on different social media networks, it's going to correctly reflect what the internet is and says, but not necessarily correctly validate or say what is in the real world. So, I think that's one of the concerns. But there's also the provenance concern where it's like, what did it learn from? And did it learn from something which was actually copyrighted

5:50or patented? Should it be using that data and where it come from? There's a lot of difficulties with that because once it goes into a model, it becomes blurred and it becomes kind of lost that provenance. So, I think for their own concerns as far as defending where they came got stuff from, they're going to have to at some point start having audit trails and

6:10more of a proof of what did you learn on, where did it come from, and how do we know where the answer came from for verification purposes. Definitely. I think with AI, yeah, it's just pulling data from, you know, some of the top Google results, but where did that information come from? How can we use the data verification through through XYO to

6:37prove or double-check that the sources that these AI are are pulling from are verified. You know, I can imagine if you're looking at location data specifically, having a drone or a weather station, having that location verified might be easier than when you expand it out to everyday common knowledge and in the social context, it's hard to

6:59verify every source. And how does the network on the blockchain pull that and verify things in a way that the centralized AI can't? Well, what we do is we do a what we call witness-based system. So, you have these different witnesses that are witnessing something like temperature, for example. Let's say you have, you know, a thermometer. They're just getting the temperature.

7:22What you want to do is you normally want to sample reality, for lack of a better term, with multiple anonymous sources. So, if you have multiple different parties actually checking the same thing, but they don't know about each other so they can't really collaborate on making false data, they can then basically be trusted because if four people or five people

7:42not people, but four different or five different sources get asked for the same piece of data, and four out of the five actually match and the fifth one's an outlier, you can assume that probably the fifth one is problematic. Or say for example, two out of three are correct, you can always just resample and see well, you know, maybe it's not part of the consensus.

8:01So, what it is it's almost like having micro consensus on data collection. Mhm. And you mentioned earlier about about the provenance, knowing where the data came from, it's almost more important than the data itself. And I always take news headlines, you know, newspaper headlines or digital media as a great example. You know, it's there's a

8:22headline, but like who's writing it? Is almost more important to know. Yeah, I've always been kind of fascinated with that where like literally I've written or I've read like different articles that were written by people, and factually they're all accurate, and you know, you can't fault any of them, but you can tell from the tone or you can tell from the way

8:41it's told or what's emphasized and what's not emphasized like there's a different perspective on those things. And people have gotten very good at that at those those techniques to be able to communicate a story in a certain way, which is part of their their added pizzazz to that story. They you know, have to have some personality in there, but at the same

9:02time if it goes so far where it's actually, you know, untrue, it becomes problematic potentially. Or if it omits omission to some degree also can be a case where where it's factually untrue because you omitted a specific thing. So, omission and untruth are not that far away from each other. Also, one of the most important things I'm actually a anti-privacy advocate for

9:25many things. Obviously, privacy is important for security and for all those things, but the idea of having people who are anonymous reporting news, for example, on X and saying, oh yeah, you're you're believing what this account is saying. We don't even know who the account is or where it's coming from. It's, you know, it's titled whatever they want to. The idea of

9:44having some sort of provenance where it's like, you know, this account has been reporting this. You can't take it down. You see this often with people who use APIs for X or screenshots of things like, hey, you know, this person posted this thing, and then 30 seconds later they took it down, but forever they have the screenshot of that." But really what you want is you want to,

10:03have things go on their permanent record where it's you have to stand behind what you say. And if you're not willing to actually put your identity on that, then it should have much less weight because you know you're not going to stand behind it. And also you can't edit the past. You can't say, "Well, I never said that." But if it says, "Well, obviously,

10:20you know, on the blockchain this was the case. There it was witnessed by different people and it was stored. You can't go back from that." So, having transparency into the provenance of something, I think it is a big step towards improving the quality of data and the validity of it because then you can actually use that either as a feedback loop for determining which

10:40ones to trust because you know after a certain number of times that if something is found found false, you just stop listening to it. Or also it could be even used as a way to you know, prosecute somebody if they were intentionally you know, putting false things on there or attacking somebody in a in a certain way. So, I think that transparency is important and we can

10:58really provide the transparency and permanence of posts or things that AI could use as well. Definitely. I think that's really important and that's been a narrative that's been growing since around the time of the XYO layer one network launch. There was a shift in the narrative towards more privacy in Ethereum and like midnight network. We just had Charles Hoskinson on, but it's

11:24more of a compliant privacy. It's not like a Monero privacy. It's like privacy that for the things that are important, you can keep private or you can use zero-knowledge proofs to prove something, but you still are tying your identity or your there's KYC there. You're not you know, going all wild west against the government or something, but having the perfect amount of privacy

11:48where it's needed optionally, I feel like can be important especially if institutions are coming in and they're trying to use verified data to make trades or to get a leg up. They don't want all their information public, but they don't they can't necessarily have all their information private. Right. Right. So, privacy for the sake of security where like your passwords or

12:09your key phrases obviously you want those sorts of things. There's also privacy from the standpoint of things people can judge you on or there's social impact from from privacy in many cases. But when it comes to things like accounting for example, you know, transparent accounting is why the SEC has filing requirements for public companies and so on and so forth

12:29to have as much transparency as possible. So, to me a lot of people feel that this is this single bullet of privacy and privacy equals security. It's like in certain lights privacy definitely equals security and in certain lights privacy actually is the antithesis of security because it gives a shroud to somebody who wants to do something that's not

12:48good or it's inappropriate. So, knowing when privacy is actually a security tool and when it's actually a security fault is important. I think people don't often think about the fact that it can have both sides of that edge. Definitely. As blockchain overall integrates AI more deeper and XYO is more getting involved with AI especially on the data integrity side, it seems like

13:16there's a big discussion around there's a bottleneck of AI whether it's in the models or the compute or the data integrity side. Where do you land on that? Well, I think one of the most important things with AI is to learn that the tools around it the tooling is what makes it more efficient. So, for example, if you have a code base and you

13:35say you have a giant repo and you say, "Well, go and find every bug in here for me or every every missing semicolon." This is going to be a very intense AI exercise. Whereas if you say, "Well, just run ESLint and look at the results and go fix those things." Or even say ESLint fix, it'll fix it for you with minimal effort. And so, the way to make AI more efficient

13:55isn't necessarily to throw more H200s at it. You know, that's obviously one way to give it more horsepower, but to give it tools so it can be more efficient. So, it's you know, almost the equivalent of you take the caveman and you say, "Well, just run the caveman you know, stronger and faster and harder." Versus giving him a hammer or a saw as a tool

14:12to be able to make things faster. So, just because AI is really smart, it doesn't mean it can do things as efficiently without tools. And so, I think the tooling around there and XYO definitely XYO one definitely falls in that tooling category where it gives AI the ability to do things in a much more efficient way. Like there's especially in the next couple of years when it gets

14:32to the point where there's nothing AI can't do. It's just can it do it in an efficient way? And it becomes even more of a case where you don't want to run things in the cloud. So, if I want to say, "Well, I want to run my AI model at home on a reasonable PC or reasonable setup that doesn't cost $50,000 for a bunch of GPUs, how do I do that?" Well,

14:52potentially if you have the right tools and you just don't have to use as many tokens you know, as many iterations, you can actually do that. And I think looking at Optimus and some of these new robots are coming out, they have to have onboard computers or onboard GPUs to be able to run their models because otherwise as soon as the internet goes out your robot freezes and

15:10that'll be kind of a problem. So, there's another case there where how do you actually make those you know, use less tokens and you use less horsepower. So, it's kind of similar to computers. It's one side is you make them faster and faster and on the other side you make them more efficient so they don't need as much much battery for example. So, I think

15:30the place where there's a lot of opportunity for the non Fang group as you call it of AI is in that tooling because that tooling can be very specific to different use cases or different domains. And it can really make AI more efficient for people to use and even in some cases easier for them to use because many people who are using AI don't understand how do you make skills or how

15:51do you make make your AI more efficient? They just ask it to do things and expect it to be self-optimizing if that makes sense. Definitely. And it's a really interesting point about the efficiency. Of course, we want AI to be as efficient as possible to not have to use as much compute, processing power, cost. But I feel like a lot of people think that

16:14integrating blockchain might mean that it becomes less efficient. At least when you compare traditional server systems or people are looking for storing things and they feel like they're sacrificing by adding in blockchain network or decentralization. Is that not the case that it actually can make it even more efficient? Well, I would say in general you're

16:37right as far as anything which has cryptographic signing or hashes and those sorts of things, there's overhead to that. There's a cost to that. That's one of the reasons why we you know, store a lot of our data off off-chain and then we link it on-chain and we add the provenance and the metadata on-chain because it would be inefficient for us to store you know, a 20 gigabyte

16:56video for example on-chain. And so, we'd want to minimize the amount of times that we pay that price you know, that overhead for doing that. But I would definitely agree if there's no need for something like provenance or permanence, then I wouldn't use blockchain for it necessarily to solve the problem. So, the back to the I have one hammer and everything looks

17:15like a nail sort of a thing. If you say, "Well, I want everything to run on blockchain." That's not necessarily a solution. The blockchain should be doing what it's good at and you know, use web 2 or traditional technologies for what it's good at. And so, I would not solve every problem that AI has with blockchain necessarily. But there are cases where having a shared ledger for

17:35example makes a lot of sense. You know, when it comes to money for example with Bitcoin, doing that without a shared ledger in a decentralized sort of way is not really very practical. And so, the benefits you get out of it is worth that cost. And provenance and permanence is definitely beneficial to have that. But at the same time, if you were to ask AI to generate a SHA-512 hash or

17:58say a SHA-256 hash of your code, it's going to be way less efficient than just having a SHA library actually run it for you and get that. So, that is an example of a tool where having it just use a CLI tool that does a hash for it is way more efficient than it figuring out the hash itself because it probably can. I've never tried that, but it would

18:17be kind of interesting to say, "Hey, here's a hash algorithm just you know, in your mind without using external tools, see if you can actually generate the hash I want." I bet it can, but it'll probably be a very expensive test. It's very interesting and but of course, we always like it when AI gets more efficient. It seems like the models are are coming out

18:38you know, every month. It's like

[snorts]

getting better and blockchains, some of them haven't really improved their speed and efficiency over many years. But the XYO layer one which you know, launched less than a year ago now, just had an upgrade that improved the efficiency. I read two to five times which is significant. Can you talk about those

19:04upgrades and you know, was that with the help of AI or what does that really enable? Well, I definitely say it was with the help of AI because we're using cloud code for example and codex as as tools. So, the pace at which we can develop and improve our blockchain is a lot faster because we're using AI tools. I think everybody who's developing anything at this point has to use those

19:24to keep up. So, we would have been able to make these improvements over say six months where it takes us maybe one month to do now. So, I'm really excited from the standpoint that our ability to actually iterate and improve our blockchain using AI is really good. And there's a lot of little things like for example, one of the improvements is getting

19:42the current balance for example of an account. And so, when we do the verification and the validation for that, you know, if we can get that balance faster with indexing, then of course the validation of the block that you're you're making goes faster. So, you know, it kind of flows up there. And so, using AI for example also to do things like

20:00constant profiling where, you know, every once in a while besides like running unit tests which it can generate as well, you can say, "Well, you run the profile on this and, you know, make a blockchain, you know, with our protocol with a million items on it. See how fast it runs and compare that with a previous version and, you know, does it run faster or does it run

20:20slower? So, being able to do you know, brute force checks like that A building those things and B having it go and frequently go and run it and check it to see if it's there makes it also a lot easier to both improve performance and improve in efficiency, but also make sure we don't regress because that's always the worst is when you have have improved something

20:40and then, you know, a build comes out that, you know, is 10% slower, but you don't even realize it because it's only 10%. And then, you know, another build comes out and now it's 10% slower again, but you don't really realize it cuz it's going on so forth. So, so making sure we don't slide backwards is something which AI really helps us with as well. And

20:56so, I'm really excited about the fact that our efficiency and our stability and those things have really improved the last you know, three to four months and we've, you know, gotten all those things, you know, the T's crossed and the I's dotted in many ways. Definitely. It's super exciting when you're able to take something that would take 6 months

21:14and do it in a month. AI is just amazing at that. And the fact that the chain is more of a I don't know if we're in generation three or four of these chains where you're not sort of bloated from technology from from 2013 like like Ethereum is and having to change things, you know, when you've already spent 10 years is building the old way. I've

21:39done it in traditional SAS. It's no fun because, you know, you're sort of built on old technology. Mhm. Yeah, one of the things that's nice also about the way that we develop our code is that since we try to minimize how much actually goes on chain and what really happens on chain. And a lot of the things that are done are off chain like for

21:57example indexing. And so, Ethereum's kind of that way where like there's certain things you can't get from Ethereum from the normal client or normal provider. Like say for example, if you want to say, "Well, how many NFTs or what are all the NFTs a person has in their wallet?" You can get that from Etherscan cuz Etherscan goes and indexes the entire chain and will know what that

22:12is, but there's no easy way to get that from a provider for example cuz it's not built into the provider architecture. And so, we kind of do the same thing where there's we try to minimize how much the true provider actually provides and have as much as possible that's provided by external sources kind of like an Etherscan type thing where we have our API where it can go index these

22:31things and find them for you. The key though is to make sure that when it does provide that answer, you can still go verify it. And that's kind of how like the NFT thing works for Etherscan where it'll say, "Well, you know, here's the 20 different things that this person has." You can go and check if those are actually true and if they actually have them in their account

22:45pretty easily. The one thing that's hard there is, well, do they maybe have 30? You know, did it was this a complete full complete answer? The it's hard to verify that with a provider. But you can at least verify the deposit. So, our system works very similarly where we try to push as much as we can off of chain. And when it's off chain, you can iterate on the

23:04the indexing or the performance of those things as much as you want to because you know, as long as the truth is the truth and you can actually prove that's correct, getting that answer faster is something which you don't have to worry about altering the primary protocol every single time that you do it. Yeah, it is it's nice to not have to put everything

23:23on chain and there are like on chain maxis where it's like we need everything there. But when you're intersecting with AI and you're especially dealing with real world data, you know, whether it's from the weather station example is good. You know, that it's hard to have the source beyond at some point that information has to come from off chain to on chain.

23:44And maybe not necessarily every single thing needs to come because it bloats it down and it isn't as efficient. Yeah, and also you have to distill things. Like I always use videos an example where if you have like a car for example with a video camera that it uses for auto driving, you can't just stream that entire video over the internet to a server somewhere and say, "Hey AI, you

24:03know, here's a live stream of my 1080p camera in the front of my car. Drive me." what you have to do is you have to have a edge device there of some sort which takes that translation into little boxes and arrows and metadata which is, you know, it finds the different items on there. Finds people. Finds other cars. And then it actually drives using that metadata instead. So,

24:22distilling and reducing data at the edge or outside of the system is something which is very important, but you still want to have a hash potentially of that the frames of that video where it came from. So, if you want to go back and say, "Well, you know, did this metadata get correct correctly generated?" we can go and verify it by trying, you know, potentially pulling that video

24:40off of the hard drive that's in the car for example and comparing it. So, you can still have the provenance of the original video without having the video go all the way upstream. And so, I think data reduction is something which is going to be a big thing and big challenge also on a ongoing basis for AI and XYO and XL1 is very good at that where you keep the provenance of the

25:00original data and the source data, but you also have the ability to reduce it into useful forms that are very efficient. Definitely. And, you know, one of the first types of data back when I first spoke to XYO in 2018 was location data. And now we're talking about the vehicles, you know, the rise in autonomous vehicles and it's a robot vehicles. And soon they're

25:21going to have wallets to pay with crypto as well, I'm sure. Is there a special focus on autonomous vehicles or you know, just physical AI with XYO as AI continues to grow and move more into the real world? Well, we've always been a oracle. So, onboarding data onto our chain and into our data lakes is one of the important things. And so, you can either on

25:45onboard it from like a website or something virtual or digital, but you also can onboard it from sensors or from cameras or from IoT devices. And so, there's there's tons and tons of data. I think the IoT ones are more interesting in many cases because that reduction is necessary cuz you can't just, you know, you can't store a snapshot of the entire, you know,

26:05data set of reality on a constant basis. It's just there's just way too much data there. And so, how do you actually take, make senses eyes or ears or whatever it is that these robots or these these cars need. How do how do you take that, reduce it to the important parts that you need to use to be able to steer it and store that somewhere? That's why to me IoT and the

26:26edge devices are are very interesting and I think an important part of XYO and XL1 going forward. And location's important, but time's also very important thing. People often forget that time is actually a component of location. So, because it's XYZT really is what we live in. We live in those four dimensions. The only difference between time and the other three is that

26:44time marches in a certain direction at a certain speed and you can't change it. Your movement in the other XYZ coordinates can change. You know, the velocity can change. You can move at different speeds or you also you can return to the same spot, but you can never return to the same spot in T. But T to me is not a fourth dimension as far as like it's only different. It's

27:03just really the fourth dimension of location. But it's it's a un- uncontrollable dimension of location which is interesting. So, we think time is a very important part of our data storage as well. And we and we include that in the bucket of location. Mhm. That's very cool. I love that. And, you know, with blockchain from the beginning

27:25time has been a major piece of that, you know, understanding not just the chain of transactions from the beginning of time. There's time's right right in there. It's all time stamped on and that can't be changed when when it's on chain. Yeah, blockchains are you know, they themselves are time vectors if you think about it because basically, you

27:46know, you might not know how how long compared to like physical time those were, but they're they're unidirectional. So, you can't go back in time in a blockchain. That's the whole point of it. I can't go and say, "Well, I'm going to go change block four on block on the Bitcoin blockchain." You know, that's not practical at all. In theory you can if you have enough

28:02hash power, but that's kind of like, you know, in theory you can go back in time also if you have a big enough quantum computer perhaps. But it's not practical at all. And so, time is always been a fascinating thing for me because it provides security. It provides permanence. And not being able to manipulate time is what makes time interesting. If time was manipulable

28:19manipulable just like X Y and Z was, then the whole world would be you'd think about it differently. Mhm. Very cool. And with the new upgrade to the XYO layer one network, does that affect the efficiency of the XL1 governance and how the token works within the ecosystem? Well, if anything what ends up happening

28:46is we can make more blocks at a faster pace. So, people who are staking XYO cuz the way our system works is you stake XYO and then you receive XL1 as the reward as being one of the stakers of the network and that sort of thing. So, you'll in theory receive more XL1 for that because more blocks being made, but also more XL1 gets burned because the XL1 that's used

29:09for gas gets burned as part of the components that are there. So, the it does accelerate the time frame basically of our cuz our goal is to get to the point where we because right now we're a slightly inflationary currency because you know, we get block rewards kind of the same way that Bitcoin does or Ethereum does. And the goal of course for all

29:31blockchains is to get to the point where your consumption of tokens equals or is greater than the production of that. So, you kind of switch from being a de facto inflationary currency to a deflationary currency. Even though you kind of still you're always inflationary because you're always producing slightly more of the token, but if you as long

29:51as you burn more burn faster than you produce, you're a net negative as opposed to a net positive growth token. So, you know, our goal and our path to that I think is getting shorter and faster because of the fact that our efficiency and our usage is going up. Definitely. And with having more transactions and blocks, is it more I different IoT devices and physical AI

30:18coming into the network or the same amount of devices being able to compute more things and store more more efficiently or combination of both? I think it's a combination of both, but what if anything we have probably you know, more partners who want to put put things on there at a faster rate. And so, in some cases what it is it's your sampling size changes. So, you

30:41know, if I say for example, I have my private data lake and I'm storing you know, 1,000 items a second, but I only want to go and roll that up and hash it and put it on the main chain you know, once every minute or once every 10 minutes. You know, I can now put it on there once or once every 10 seconds and that sort of thing. So, basically your your

30:59your your resolution or your your frame rate to large degree is getting getting faster which is good for them because they can have a better frame rate as far as being able to have permanence to this on there. And they can have more data that's stored on there because the amount of hashes they can put inside of one transaction gets bigger. So, I don't think if I ever had a shortage

31:20of data that we want to have on there. It's just how much how much of reality you want to capture and put on a blockchain. And if we can capture more and more, you can analyze more and you can do more and you can and learn more. You know, one of the things we're working on also is like for people to capture their own queries and responses for AI. And if you

31:38can do that, then you can actually use those to build context and train your own models. And that sort of thing, but that becomes very voluminous as well. So, you want to have these data sets of your your own data so you can propagate those into a learning feedback loop basically for your AI. That's very cool. Are there developers that are working on

32:04the layer one right now or XYO you know, in the combination with AI to be able to do things like that to build their own models or to capture more data? Like can anybody that's a developer go to XYO right now and start tinkering with that? Well, you can definitely the simplest thing to do is if you just install our wallet for example in browser then kind

32:24of like Ethereum you can go and grab the gateway that injects into the page and you can you know, connect to that right there. We also have an SDK where you can do it like from node for example or from other things. Or if a person is really industrious if they want to do it from say go or those things, they can just call our the RBC endpoints for submitting

32:44transactions, but we're we're working on SDKs to be able to make it easier for people to be able to roll the transactions and submit transactions from other languages besides JavaScript at this point. But again with AI really helps us with that because to say, "Hey, we have a really perfectly working JavaScript version of this. Make me a Kotlin version of this

33:05or make me a Go version of this." is pretty easy with AI where with a human going through and porting to a different language is a lot harder. And so, I think the ability for us to actually make unlimited you know, SDKs that all work as well as the primary ones we've developed is going to be much more rapid than it would have been even a year ago. Mhm. Definitely. Well, it's

33:29great to see the updates happening quickly with XYO and you guys are still building strong whether it's bull market or bear in the crypto industry which is just a small piece you know, AI and physical physical AI and IoT is going to encompass the entire world outside of blockchain. Looking forward, how do you see the pace of these different devices moving on chain

33:57and into XYO compared to the rate that's that's happened so far? Well, I think everything in the world is going to be moving faster. I think it's a extremely exciting time to be alive as far as developing software and developing technology and products. It's also a scary time because you know, like sometimes I'll be I'll wake up in the morning and be like, "Hey, I just this

34:14really cool idea." I'll write it down and I'm like, "Maybe maybe we should do this." And then a week later somebody else ships it. So, it's like the race to be the first first to develop the or deliverable of that product is really hard now because everyone has this ability to develop things fast. So, knowing what's novel and knowing what's interesting and

34:34building tools that are are slightly better perhaps than somebody else's because of usability and those sorts of things are are becoming more and more important. But I think everything's going to move a lot faster once we have I think Optimus is supposed to ship later this year. They're their first robots right now. There's like a few Chinese companies you can actually order

34:50robots from. The point where people have robots in their homes and they want to store the data from that. They want to have an audit log also where if something bad happens, they can go back and look at what did the robot do? You know, you want to know what they did because you never know. You know, like like why did they do this or why did they do

35:07that? So, I think that the need for storing provenance data on online or in a blockchain is going to just explode over the next two to three years with things like any sort of self self-motivated items like robots or cars or any of those things where people need something to help them as a human have some sort of I wouldn't say defense against that, but some sort

35:31of layer of understanding and comfort of using those because I for a lot of people it's very uncomfortable to think that this thing can do things that you don't even you know, you don't even understand how it does it. Yeah, it's going to be a crazy world. I would love to see my car being driven in sync with XYO or the robots in my house when when they get better

35:55which you know, we've seen some of the early videos and it's not just remote controlled by by a person, but quickly it's going to be getting really good. And I'm sure they'll be the top surgeons in the world and top janitors and everything. Have that all on chain is I think every every blockchain fanboy's dream. So, I'm excited to see that and wishing you and the team the best on

36:18playing a big part in making that happen. Yeah, thank you very much. Thanks for having me on today and as I say, I'm very excited about this space and it continues to be a great journey. Yeah, thank you so much, Ari.