[Ep#4] BUILDING THE FUTURE OF ON-CHAIN AI WITH 0G LABS

@0G_labs - The first modular AI chain, revolutionizing the crypto and AI overlap.
Dive into the future of onchain AI and the impact of data availability in the crypto industry with our special guest @mheinrich 

Timestamps :  

00:00 Intro

01:37 Issues & Latest in Crypto Lands

06:21 Michael Intro

09:00 Existing State Of DA

12:09 Data Storage and Limitations 

14:05 End Game For DA

17:08 Application Areas For 0G

19:14 Why WEB 3 & AI Need Each Other?

22:57 Why not L1 & Thesis Behind Modular Stack 

23:47 Michael's Fav Virtual Machine 

25:43 Biggest Hurdles For 0G? 

27:55 Michael's Thought on 0G's Competition  

29:43 Name & Narrative Around AI : Hindrance Or Not?

32:53 Demand & Interest for 0G 

34:10 Gaming & AI 

35:20 A Day in Michael's Life 

37:30 Michael's Free and Filled Schedule 

38:44 Michael's Omnipresence At Conferences  

39:00 Building A Good Team & Saying No As A Founder 

39:41 Building & Managing A Team : WEB 3 VS WEB 2

40:36 Michael's Remote Team 

41:03 Outro

Video Credits : @BoriyaKishan   

[Disclaimer: It's important to note that the information provided here does not constitute financial advice. The views and opinions expressed herein are purely personal and do not in any way represent or reflect the official stance or viewpoints of the individuals company]

Transcript

Speaker 1: 0:00

So welcome everyone to this week's episode of the DealFlow podcast. I'm your host, mixed Race Magic. With me today are regular guests Cam of MH Ventures and Mehdi of Animoca Brands. But we're very, very proud to have special guests with us today in Michael, who are both with the same project that I think we'd all agree is uniquely positioned to capitalize on the overlap between blockchain, crypto and AI. But without further ado, I'll let you guys introduce yourself.

Speaker 2: 0:42

Great Well, thanks for having us. Super excited to be here. I'm Michael. I'm the co-founder and CEO of Zero-G. Zero-g is the first modular AI chain and we help launch AID apps to the fastest programmable data availability layer in existence today, and so anything that can be built in a centralized context can now be built on-chain with the same cost and performance, so it's a major breakthrough for the space. Super excited about building the future of on-chain AI as well.

Speaker 1: 1:11

Phenomenal guys and I have to say what an incredible position to be in at this point in the market, for sure. But before we dive into the really interesting stuff, just a quick word to say that this is a podcast. We're all speaking as individuals here. Nothing in this podcast that is in any way intended to constitute financial advice. You should always do your own research. Everything we say is opinions, riffing, chatting and and all the rest of it. But with that out of the way, I think, before we do dive into some of the the topic specific stuff, are there any sort of issues, uh, across the crypto industry from the last few days, the last week, that we want to sort of have a quick word about, just to dive into?

Speaker 3: 1:46

to begin with, Maybe we can touch on the dip in the market and then also GCR tweeting after one year and turning the market around. I don't know if you want to just dive into that a bit, Johnny, and then we can elaborate.

Speaker 1: 2:01

So I don't have a lot of context on the GCR stuff, only what we've been discussing pretty recently. But in terms of the market dip, I mean, did it really come as a surprise to anyone? I mean, like all the major pundits and I know that the pundits are never, ever right but everyone was pointing to a big dip before the halving and a big dip after the halving, which kind of I think is at least from where I'm sitting, is not hugely surprising, but was anybody expecting sort of a major pump around the harbinger we always normally see like a, a massive up move really.

Speaker 3: 2:34

But then I take the context of three arrows. Capital is up only so.

Speaker 1: 2:35

That's my context that's one of the many ways you're just like three ac, isn't it?

Speaker 3: 2:41

that's the best way to try, but I did see a tweet. Someone tweeted out, like when the market dipped, that these guys were like down, maybe like three, four million plus, and I'm like how do they still think that the market just goes literally up? They're just constantly losing money with their new project, or is it oxbow? It's a derivative decentralized derivatives platform yeah, I've forgotten what the actual name is, and, and I think they're- the what's it built on?

Speaker 1: 3:05

Is it just another Ethereum L2?

Speaker 3: 3:07

But yeah, gcr tweeted out that we shouldn't sell our tokens here. We should all hold our bags or double down. And then, it's a coincidence, but the market literally turned around the minute he tweeted out.

Speaker 1: 3:23

So I do think it's a coincidence, just on the basis that you know, most of the capital flows are very much institutional and I don't believe that a sufficient degree of those institutions are looking at stuff like twitter, even when a big account like that comes through, like it's not getting indexed into bloomberg terminals, stuff like that, but I don't know.

Speaker 2: 3:39

Be contrarian and tell me why I'm wrong there's a lot of twitter bots, a lot of whales that have yeah, I would also assume that some of the bigger institutions and hedge funds do some type of sentiment analysis and then probably try to buy in and identify if there's kind of a low, once the kind of low point has been found in the market, so that they may have contributed, uh, to some extent maybe.

Speaker 1: 4:00

So actually yeah because look how inefficient the markets are to be fair, there's a very the market is so inefficient in crypto, like something like that.

Speaker 3: 4:07

A tweet like that is able. I would think about this now look at kobe. If kobe tweeted out about something that would turn the market around, I honestly believe that would.

Speaker 1: 4:18

If he was to do that what are you talking like btc and ether? Are you speaking like a specific project?

Speaker 1: 4:23

maybe not bitcoin or eth, but other altcoins and all coins, no market cap maybe. So I mean to be fair. I mean I'm arguing against myself here but when was it? It was november with the coin telegraph tweet for bitcoin etf approval. But I do think part of the reason was at least what I was told by some friends. So I don't really know how verifiable it is, but that one specific tweet was indexed into bloomberg terminal terminals. So you had a load of white collar big institutions that were sat there, you know, doing, doing intraday hands-on stuff that actually really really did think for a brief moment that bitcoin ets had been approved. So you know, you know, social media is no joke. I mean, I know that we at bscn we definitely take social media seriously.

Speaker 3: 5:01

We think that's where where everything happens first okay, and then, just before we get into, uh, the main topic, I'm going to ask medley because I'm not sure if it's already happened or not and him being solana maxi, he will know did they patch the update yet? Is that today or is it tomorrow?

Speaker 4: 5:17

I have no idea. I've been following too much macro at the moment, uh yeah care about that effectively.

Speaker 1: 5:25

I think at the end of last week anza, who run validator clients, I think, on solana they proposed the patch right and they asked um solana devs to implement it, and I think what's been implemented today, and part of the reason that solana and whiff and bonk and boom and all the rest of it rallied, is because the first of those upgrades has been implemented successfully. I don't think it's as simple, as today was the day that it just got fixed. I think today was step one and it didn't, you know, go tits up, which which is a win. I love that.

Speaker 3: 5:53

They've made the first iteration of this patch and all the memes rally.

Speaker 4: 6:01

And also, yeah, it's, solana is still on testnet.

Speaker 1: 6:04

Here we go oh yeah, that's true. I haven't heard that for like a year man oh god, the most performing network.

Speaker 1: 6:14

Everybody knows amazing. But yeah, should we? Should we move on to the juicy stuff, okay, so today's topic for you know, probably reasons that you can guess is I. I mean, I think we actually we agreed on this before we even spoke about guests and stuff like this, because it is one that really does need to be spoken about. But it's data availability, right, and you know, even to go a little bit more within that, data availability within the crypto AI landscape especially, which obviously Michael and probably the best guys in the industry to lend comment on, I think that would probably be fair to say at this point. But just to kick us off, before we dive into specifics, maybe Michael can just speak a little bit about your personal backgrounds, how you ended up into what is, at the moment, a fairly niche sector.

Speaker 2: 6:57

Yeah, I can start. I started initially as a software engineer and technical product manager at Microsoft and SAP Labs. I was working on frontier technologies at that point for example, the Visual Studio Group and then moved over more to the business side, worked for a Bain Company for a couple of years and then moved to Connecticut to work for Bridgewater Associates, where I was on the portfolio construction side and saw about $60 billion worth of trades on a daily basis and understood a lot of risk metrics and, for example, we looked at CDS rates for counterparty risks and so on. So definitely great kind of TradFi exposure.

Speaker 2: 7:38

After that I went back to graduate school at Stanford and started my first Web2 company, a company called Garten. I scaled that to about 650 employees, $100 million in contracted AR, raised about $130 million for it became a unicorn and top YC company at its height. And then my classmate Thomas from Stanford called me up late 2022, and he was like hey, five years ago I invested in Conflux. Ming and Fan are the best engineers that I've ever backed. The four of us should get together and see if there's some magic that can develop. And after about six months of co founder dating I basically came to the same conclusion and I was like, wow, ming and Fan are actually the best engineers and computer scientists that I've ever worked with. We have to start something. And so it was the team first and I basically went to a board chair type of role at Garten and then went full time into Zero-G because I was so excited about kind of Ming, fan and Thomas and us working together and making a major difference in the space.

Speaker 1: 8:43

I mean talk about a stuck team. Those are some of the best introductions I've ever heard. That's absolutely incredible. I imagine the VC funding process was a fairly smooth one for you guys would be my guess there. But no, thank you so much for giving us the context there. I guess we can talk a little bit more about backgrounds and experience and sort of the vision and where that came from. But I think one, before we really dive into the topic of data availability, would love for us just to discuss a little bit about the existing state of the DA landscape. I know that there's some players that everyone will have heard of, but you know how would you assess or look at it at the moment, so to speak?

Speaker 2: 9:18

Yeah, there's different sources of DA and it's, of course, chain dependent. But let's say we just look at Etherium, we've got ETH DA and kind of prior to the Dancun upgrade, it was about 0.08 megabytes per second, if I remember that correctly, and then Celestia and EigenDA and I think Avail entered the market and they all tend to be around 1.2 to maybe 10 megabytes per second or so. Now the issue is that when you talk about AI type of applications or really any on-chain game application, that type of throughput is simply not enough. We really need to be talking about gigabytes per second, not megabytes per second. For example, if you want to do on-chain AI training at some point, really you need like a 50 to 100 gigabytes per second pipeline for that to even be feasible, and so we're talking about many orders of magnitude difference. And so that's really where we saw the opportunity to say, okay, how do we create this big unlock so that Web2 large scale applications can actually be built on chain with the same performance, with the same cost? And so that's where we saw a really big gap in the space.

Speaker 2: 10:34

And then there were other pieces that were not fully kind of thought through.

Speaker 2: 10:39

So, for example, we think of data availability as a combination of data publishing and data storage, and so our core insight is to segment the data into those two lanes so that you don't have a broadcast bottleneck in the system, and so that then leads to breakthrough performance, essentially.

Speaker 2: 10:59

And so that was our core insight. And then there's other aspects that you can then do when you have a storage network attached to it, such as model storage, training, data storage for the specific use cases and even programmability. You can do complete state management. You can determine where you want to store data, how long you want to store the data, how much security you want in the data, and so all of these different types of use cases that are really required for other application spaces are now possible. And so the state of DA right now is basically, we made a big leap forward, because going from 0.08 megabytes per second to 1.4 is significant and definitely great and improves transaction costs and reduces them by, you know, 90, 99% in some cases, but it's not enough for what the future really is in the world, and that's high performance, ai applications, on-chain gaming, you know, high frequency, defi, all of those types of applications.

Speaker 4: 11:58

Yeah, I have to like follow up questions and very naive in nature, so I'll focus a little bit on the storage side. So you mentioned the transaction history for L2s and even like a kind of like a history for AI models as well. So, like in terms of storage, like what's the mental model here, like how long do we have to store the data? So that's number one, like the naive question. And the second naive question is I think we have things like rv and and and file like file coin. So these like decentralized storage network. Do you think like we already have them? Uh, do you think they can come in and then kind of like help on the date, not on the data publishing side, but like help on the storage side to kind of increase the throughput? Like what, what's your opinion there?

Speaker 2: 12:43

how long things need to be stored, it depends on the purpose. So if you're thinking about disaster recovery, something should be stored forever, essentially so that you can reconstruct a state. If it's specifically for, let's say, optimistic type of roll-ups and you've got this kind of fraud-proof window, then you want it to be at least, you know, seven plus days, so that you can reconstruct a state if you need, and then, of course, shorter for other type of roll-ups, depending on the situation. But that's kind of the spectrum. And then, as far as other platforms on the storage side, so we opted to build our own in-house storage system, because Arweave and Filecoin were designed more for log type of storage, which means longer term cold storage, and so not designed for really fast data ingestion and data retrieval rates, which is very important for more AI type of applications or even structured data applications that need, like a key value store or transactional data types, so that you can process things significantly faster and you can actually build applications, like you know, decentralized Google Docs if you wanted to.

Speaker 1: 13:52

Pretty awesome articulation in terms of why you know data layers are needed and also, I think, especially interesting from where I'm sitting is why existing decentralized storage solutions just aren't really fit for purpose in this particular context. So cheers for clearing that up for sure. So we can either go. We can either talk a little bit about the end game for data availability Don't know if that's one we want to go to or a question that I really wanted to go down was the need that AI has for Web3 and the need that Web3 has for AI. So I'm happy to do either of those. We can probably do both, but in terms of one. To start with, to prioritize, what do we think?

Speaker 2: 14:29

The endgame is pretty easy to define. It's essentially what we're building is matching Web2 performance and costs so that you can build anything on-chain, specifically AI type of applications. It's pretty straightforward from our perspective, just like in Amazon web servers you've got compute and you've got storage, and so S3 is a critical component. Of course, ba is a bit different because it has other properties, but it's a key component, and so to us, the end game is to build an AI modular stack where the data availability piece is not just data publishing, but it has the storage components and then is essentially put together by a consensus network. And so, in our definition, we don't have, like clients do data availability sampling, but we have the actual consensus network do data availability sampling, and then, once that consensus is reached, we can then be a testable to the underlying layer, one like Ethereum. And so we're essentially building for the end game where any high performance application can be run on chain, so that we can even support things like on-chain training of AI models and so on.

Speaker 1: 15:42

I mean, I think what you've articulated is as much an end game for the entire blockchain industry as it is for DA right, the point at which and what I mean by that is the point at which decentralized applications more broadly can actually compete with centralized versions, right, like for a lot of different applications, what they really have going for them is the fact that they're decentralized, right, so so it's almost like a trust-based bias towards using them, whereas the point at which they're decentralized right, free from third party influence and censorship, and they're also cheaper and they're also faster is kind of that does sound like an end game for the entire industry, which is which is really, really cool. I don't know whether, cam medhi, from, like, an investor point of view, you guys want to. You guys want to?

Speaker 4: 16:18

yeah like what I'm hearing is they're not just celestia competitor or like polygon avail Avail, there is element of Filecoin and Arvive that they might go after. That's kind of like a hybrid positioning and like in terms of our thesis formation, we also feel there'll be a lot of synthetic data produced from an output standpoint because of all of these MLM models. So there is definitely a need for storage and there's also a need for data availability. So I think from an investor standpoint, if you kind of believe that the data availability will be commoditized, then kind of like a hedge for you is the storage part. So I think that's what I'm thinking with all the conversation ongoing. I think that's what I'm getting a sense of.

Speaker 3: 17:05

Maybe, cam, you want to highlight or like like what are your thoughts? Kind of come in and maybe go 10 steps forward here just for, like, the listeners to kind of understand like your target market of what you're looking. I know it's, of course, ai and people building um different things inside of blockchain on the AI side, but, like, from your aspect, what kind of projects are you looking to implement to use zero?

Speaker 2: 17:31

gravity. You already mentioned one application area. So we're trying to build the largest decentralized AI community and to have a lot of projects built on top of us, whether it's somebody like aond building a large graph model, or it's somebody like Fraction AI or Public AI doing decentralized data labeling or data cleaning and then execution layer type of folks like Aura or Telus Network or Ritual kind of building on top of this, definitely trying to build the largest community for AI builders, and so that's kind of the table stakes for us. But then really any high performance application can build on top of us. Take the example of on-chain gaming 5,000 users, uncompressed need data availability of like 16 megabytes per second to have full on-chain game state.

Speaker 2: 18:24

No DA layer today supports that Maybe Solana, but that's different from the Ethereum ecosystem and only to a certain extent. So applications like those are also super interesting for us, especially if they then combine it with, let's say, a non-playable character, that's, an on-chain AI agent, for example. So a lot of cross-use cases there High-frequency DeFi. Another example FHE in the future Data marketplaces, you know high-frequency, deep end type of applications All of those will require really large data throughput and need a DA layer that really supports that high performance so really any high performance DApp or layer two can build on top of us.

Speaker 1: 19:14

That's really, really cool and I think I mean, you know, I kind of came into this conversation very much focused on AI and you've kind of enlightened me as to the range of applications that actually kind of stand to benefit from this kind of infrastructure. But just to take a step back, almost away from DA, specifically right to go back to that question that I mentioned before, you know you guys have chosen AI as very much a place to hang your hat and all the rest of it. But what is the need and you know it's a two-way question, I think, as I articulated it is why does Web3 need AI, hosted within its ecosystem, and why does AI need Web3? And you know the answer doesn't have to be to be yes to either.

Speaker 2: 19:52

Well, I mean, you could definitely see a world of Web3 without AI, but I would say, the world in the next five to 10 years. Every company is going to be an AI company, because it's as large of a happening as the internet was 20 plus years ago. And so do we really want to miss the boat in Web3? I don't think so, because there's trillions of dollars of economic value to be unlocked by AI. According to McKinsey, 70% of work can be automated by AI, so why not tap into its capabilities? And so I think there's certainly a world without AI in Web3, but it's going to be a much better world, and we fundamentally believe, let's say, in the next five to 10 years, that most people on a blockchain aren't actually people. They'll be AI agents transacting for you, doing things for you, and so I think it's going to be a very exciting world where we have a lot of automation in Web3 driven by AI agents, kind of a whole ecosystem designed for the user, so to speak. Now, the reverse of that is, I think AI absolutely needs Web3. Our mission is to have AI become a public good, and so it's fundamentally a question of alignment.

Speaker 2: 21:07

How do you make sure that AI models don't cheat, how do you make sure that they make decisions for the best of humanity? And so alignment can then be broken down into incentivization, verification and kind of security component as well. So each one of those is ideally suited on a blockchain environment. So we know blockchains help with financialization and incentivization through tokens, and so from that aspect, you can create an environment that where, economically, ai agents don't want to cheat. All of the transaction history is also on a blockchain.

Speaker 2: 21:54

So I'm even going to take a bit of a kind of stark statement that I fundamentally think everything should be on chain, from the training data to you know the data, the training data to the data cleaning components, to the data kind of ingestion and gathering components, so that you have full attribution in terms of who's actually providing that data. To then the decisions that the AI model is making as well, to 10 years from now and I imagine AI agents running logistic systems, administrative systems, you know, manufacturing systems I would want to know the version of the model, the decisions it's making, and have some type of oversight, especially if it's a beyond human intelligence type of model, so that there's full alignment and to put AI into again a box where it can cheat and it's making decisions for the best of humanity. I don't know if I trust a handful of companies to consistently ensure that future for us, especially with super capabilities that an AI model can potentially have in the next five to 10 years. For sure.

Speaker 2: 22:57

Fair concern to have.

Speaker 1: 22:58

Mehdi, you had a couple of questions.

Speaker 4: 22:59

Yeah, yeah. So one question that's popping up in my mind it seems like you guys are working towards scalability, throughput state bloat because of the storage component. Why not launch L1? Like, if you guys have the capability and have the technological breakthrough to kind of like tackle these issues, why did you decide to kind of go down this approach versus, let's say, start your own L1 where you can have like a virtual machine of your own and have all of these applications built on top, like what was the thesis behind having like a modular stack?

Speaker 2: 23:32

Yeah, fundamentally under the hood, we are actually in L1, but we very much believe in modularity also being the future of the way you want to build applications.

Speaker 2: 23:43

And just because we're modular doesn't mean we can't provide an execution stack in the future at some point, so maybe there's one that's highly optimized for AI type of applications. We haven't fully decided on our roadmap in that regard, but it certainly is possible. But modularity is all about choice. You can choose what your settlement layer is. You can choose what your execution environment is. You can choose what your DA layer is. You can choose what your execution environment is. You can choose what your DA layer is, and, depending on different use cases, we want to enable developers to choose the best for their particular environment. And also, if you look at Web2, the reason why TCP IP, for example, was so successful is because it was modular in nature and because it was unopinionated, so that developers could use different aspects of it in their applications. And so we fundamentally believe in choice and being able to construct whatever environment that you need for your particular type of application.

Speaker 4: 24:37

If you were to take an opinion right now let's say you were having a discussion to have a virtual machine, what is one virtual machine you really like at the moment in terms of what's available in the market, or something that would be very suitable towards the applications that you're kind of thinking about or kind of working towards?

Speaker 2: 24:57

I think very practically about this. So if it's about attracting more developers from Web 2 into Web 3, that would be some type of WASM type of virtual machine that just allows you to build in the most common programming languages like JavaScript or Python. Those are not necessarily the best optimized for you know. On chain type of development, another one that's really interesting is like the movie M, just because of the way it thinks about objects and throughput and so on. So if you're looking for really high performance, that's a really interesting one to look into as well. And then if it's more about just what's been battle tested, then more like an EVM, because there's a ton of Solidity developers. So to me it depends more on the use case specific.

Speaker 1: 25:43

Very, very cool, some incredibly in-depth answers there, so really really appreciate it. I guess probably, from from my point of view, I'd want to hear about the biggest hurdles that you guys are facing, or is it a bit completely smooth sailing right, like I can't imagine that, like the scale of the undertaking that you guys are doing, it can't just be straightforward the entire time yeah, I think, uh, if you look at any startup, it's never just smooth, it's always some type of hero's journey, I think.

Speaker 2: 26:08

So I'd say the biggest challenge from my aspect is just making sure that we keep up with the intensity, because there's 50 different projects that we have to execute on really, really well, and we certainly have to make some trade-offs by being very quick in market.

Speaker 2: 26:25

So, for example, we wanted to launch with a custom consensus.

Speaker 2: 26:29

However, that would add, you know, four or five months to the launch timeline.

Speaker 2: 26:33

And so we said, for phase one, I think it's okay to use an off-the-shelf consensus just to do a really strong proof of concept and to get some percentage of the way there to our kind of endgame of, let's say, 50 gigabytes per second per consensus layer, and then, in phase two, introduce horizontally scalable consensus layers as well, so that we in effect have infinite DA throughput, which means just like you can flip on a switch and have another AWS server launch.

Speaker 2: 27:02

Very similarly, we can add additional consensus layers that then add to the overall data availability throughput. And so some of the key hurdles were just to say, like, how do we make sure, with the resources that we have, that we actually invest in the key elements and make sure we have a timeline that's realistic, actionable and also hits kind of the strength of our system, and so kind of that planning was definitely a key challenge. And the other challenge is to make sure from a recruiting standpoint that we bring absolutely kind of A-plus talent into the company because of the strength that we have in the team. And I mean we've mentioned some of our backgrounds and we have many Olympic gold medalists in informatics, for example, many top computer science PhDs. So we need to make sure that the go-to-market side matches that and any of the additional developers coming in match that as well. And so those have been two key hurdles so far.

Speaker 1: 27:55

It sounds almost like the biggest hurdle you guys are facing at the moment is prioritization right, accepting that you can't do everything within a perfectly finite timeframe and you kind of have to make, know, make some trade-offs, at least in the short term, which is still a pretty awesome position to be in by by the sounds of it. Um, one follow-up question I'd have is around competition. So how do you guys envisage competition? I mean it might. You know you guys are so early to move and, like you say, you've made speed a priority. But you know if a competitor were to pop up, what would look like, because I'm guessing that you're not really in a position where you're feeling massively threatened by Celestia or Eigenlayer for the specific use case that you guys are going for.

Speaker 2: 28:32

Yeah, and my perspective on that is in Web3, it's going to be heavily community dependent, a very strong community around just high performance and AI type of builders versus, I think maybe a Celestia and Eigen.

Speaker 2: 28:50

They have more of a kind of generalist type of community, people that care maybe more about bringing economic security and building an AVS. On the Eigen side versus Celestia is a little bit more around. You know which layer two wants to reduce their transaction costs and doesn't have a lot of high throughput applications. So, for example, building high frequency DeFi very, very challenging on Celestia because you need many megabytes per second of throughput and it would completely clog Celestia's throughput. And so from that perspective, yeah, definitely don't feel threatened and building a very strong community so that if somebody else pops up, we've got kind of the network effects of developers and Mindshare already going for us and then hopefully the financial assets that come along with it, and so the best defensibility from that perspective I think is the network effects that we have.

Speaker 1: 29:43

That's a really cool answer because we spoke I think it was last week we spoke about teams with very, very high degrees of technical expertise and limited sort of business development marketing and then the flip side right, so very sort of marketing, heavy teams, and it sounds like you guys have kind of boshed both on the head by the sounds of it at the moment. The fact that you know what you're really concerned about is community building and that's where the moat exists and and all the rest of it.

Speaker 3: 30:04

many cameras, both different, but uh, first one being I know that you touched on previously of the landscape of uh projects that can be built, um, using uh, zero gravity, but being so niche. On the ai aspect, we all know that crypto has narratives. Do you see this long term being a hindrance to you guys Because, like, where do you see that going? Because I'm assuming you guys are building and it's like, like you said, it's very much the stack is going to be a lot better than what we already see today. Do you think that the narrative and the name itself being geared around AI like is going to hinder you guys later on down the line?

Speaker 2: 30:46

Um, we don't think so Because, as I mentioned, my belief, or our belief, is that every company is going to be an AI company in the future, and so it's going to be much rarer to see a company that doesn't use AI in some shape or form into in their applications or in their platform. And so, from that perspective, every time there's a new version of, let's say, gpt that comes out that maybe has a trillion plus parameters and unlocks another set of elements that we didn't have before, another level of performance, I think the hype will consistently sustain itself, because it is a completely new paradigm. I mean, for the first time, you can tell a computer and human language what to do and in some cases, beyond what a regular human being would do, you get an answer and can automate processes that you couldn't before. I mean, there's certain some companies that are almost automating their entire sales development function or customer support function already. Imagine what a 140 IQ AI model can do in the future as you release GPT-5 and 6 and so on. And we need to make sure that we keep track of that in Web3 and build our own versions, our own open source versions that are just as good or better, and then again when we go to that future of having AI agents run certain parts of our society, even that it's governed in an appropriate way and governed on the blockchain.

Speaker 2: 32:15

And so we think the future certainly is AI for the next 10 to 20 years at the very least, and there'll be big societal changes that'll happen as a result of that. I mean, even look at full self-driving mode from Tesla. It's almost there. So the future is happening kind of on a day-by-day basis. And then you know, androids are going to come into our lives as well. You'll have lots of support. There's going to be an LLM that's in the background interacting with you. So the amount of prosperity that can be unlocked, it's just mind blowing. We're basically living in a sci-fi movie at this point.

Speaker 3: 32:53

The next question would be, as you're saying that the future is everyone's going to be using AI and it's a big narrative right now in crypto, what's the demand being on your side from a BD side of interest for zero gravity? You're in testnet now going into mainnet, but you must have seen or you're looking to, uh, you know, get that interest so you're ready for for main. How's interest been?

Speaker 2: 33:18

uh, interest been super great. Um, definitely a lot of interest on the ai side, building a very large community around that. A lot of major l2s and we haven't, you know, done a lot of public announcements but they'll definitely come in the future Super interested in integrating with us. A lot of high performance L2s also quite interested, even some L1s using us for kind of stake management. We've chatted with some around that. So there's been a tremendous amount of interest across the board. So even the on-chain gaming side once people realized that having 100 million users on-chain is not as easy as they thought, started getting a lot of interest from AAA games as well. So yeah, across the board really great interest, super excited about building.

Speaker 3: 34:08

Go on, mehdi, you go this time.

Speaker 4: 34:10

Yeah, I feel like with on-chain game that's kind of an interesting one, because I think in some sense it's also at the intersection of gaming and ai, because you sometimes need inference, you sometimes also need model training for autonomous world. So so I think from that perspective it's it's it's interesting, um, because, if you think about it, gaming could be a sub-segment of broader AI strategy here.

Speaker 2: 34:35

Yeah, absolutely, and I'm actually speaking at a big gaming conference in Riyadh as well as one of the keynote speakers, and I have some gaming-specific things too, because gaming could evolve into a very interesting direction, or actually all entertainment. What if you have content that's just tailored to you specifically, your likes, the characters you want, and it's a story only for you? I mean, how cool is that? And you get more and more of that content completely produced for you, versus just mass produced for a generalized audience. There's so many possibilities and, uh, totally opening that up through you know ai applications and then, uh, blockchain really being the kind of governance and financialization layer of it all I have another question, but it's going to come away from a little bit about what we're talking.

Speaker 3: 35:24

Not what we're talking about, but just your daily life now. Like as a founder, looking to take your project to, to mainnet uh, meeting you in in hong kong, knowing how hectic your schedule was and, you know, speaking with your, your team as well, trying to get this organized. What does your daily life look like you, for both yourself and ming um? Like how busy you guys, you know, yeah, what does that look like? And how do you manage a team? Still like?

Speaker 2: 35:52

I would love to know um, yeah, from from my perspective, it's a heavy prioritization. Not everything that's urgent is important and not everything that's important is urgent, and so I generally try to think about what's important and urgent first and then focus my intention on that. And there's a number of initiatives that are really critical, kind of leading up to MainNet and anything from foundation setup to chatting with market makers, to thinking through tokenomics, through thinking through what are the key critical partnerships and hires that we need to make, and so there's a lot of different types of initiatives, and so one of the first things that we did is just to make sure what is the timeline? How are we doing against the timeline? What are the you know 50 initiatives, which ones are important today, and just to consistently heavily prioritize.

Speaker 2: 36:48

And then, on top of that, for me, especially as to maintain a really good schedule, so I like to make sure that I get enough sleep, that I eat a very clean food, that I exercise regularly, that I also do my meditation practice very regularly, because that gives me a ton of energy, so I do something called transcendental meditation. I actually learned that at bridgewater thanks to, thanks to ray dahlia, who introduced it. Um, and so it. Uh, it makes a big difference to have a very balanced life, because it puts everything else into perspective. If I feel balanced, like, the rest of the world will appear balanced to me, and so, um, that's super critical to me.

Speaker 3: 37:30

So do you feel lost when your diary's free?

Speaker 2: 37:37

I don't know if I feel lost, but I haven't had that feeling for quite a while. So I think Web3 speed and activity is pretty exciting. I mean, recently I've been at a lot of conferences and just doing a lot of speaking engagements, keynotes and panels and so on, and so somebody took a photo of me yesterday with Ada, who's on our team leading our growth, kind of meditating together, and it was super funny because I was essentially fell asleep and so my head's like down like this and then's head's like propped up like this and just perfect description of like the craziness over the last few days and so super, super fun. Yeah, I enjoy having kind of a very filled schedule, but when it's not filled then I think about what to read or how to enjoy nature or to spend time with friends and family as well, and so there's ways to definitely fill the time in different ways, and it's good to have some really hectic type of pieces of my schedule and then also a good amount of downtime just to recharge too.

Speaker 4: 38:43

Yeah, I can attest Michael is everywhere, like everywhere. Whichever conference I am I'm trying to book, I see him on panel panel. Uh, even token 2049 even hasn't started, but I've already seen him like like register a few panels or or attending like that's like massively down the guys everywhere team right, like you've built a really good team to help, on that side, organize and plan.

Speaker 3: 39:06

I think that's what it boils down to, and I think, also as a founder as well, like you just have to really be really good at saying no. I think that's what it boils down to and I think, also as a founder as well, like you just have to really be really good at saying no. I think that's like personally reflecting, and I'm I'm guilty of it because I don't do it enough, because I feel I don't know, maybe I'm just too nice, but I think that that's another thing of like you need to say no a lot of the times, because all your time is taken up by meetings and so forth, but knowing that you've led a team of 650, I yeah that that's pretty insane. I don't know how someone does that.

Speaker 2: 39:40

It's um, it definitely. The second time around is much easier because I know exactly what to build, how to think about team structure. It's a bit different in Web3 because certain things are done quite differently than Web2. So, for example, in Web2, from a sales perspective, it's very much my professors back in grad school like to say token operated or coin operated. You basically say, here's your quota, here's how much it's worth to you from a compensation standpoint, and then you go Versus. In Web3, it's much more ecosystem alignment and so it's not as important there. But then finding the right profile and so on, and then it's just about kind of structuring it. What does the C-level executive structure look like? What is kind of the VP level underneath? And then do you need middle management and what's the kind of team by team perspective look like? And so it becomes a lot more about structuring and strategic planning.

Speaker 3: 40:36

Have you always been remote in terms of team, like your last team at 650,? Were they office based or remote as well?

Speaker 2: 40:41

It was a big office culture until COVID and we had to go remote, and so I actually don't mind remote, it's as long as there's consistent key check-in points. So we make it a point to have quarterly offsites where we adjust in person, especially really deep strategic thought. I think it's a lot easier in person than it is kind of virtual Awesome awesome.

Speaker 1: 41:04

That was super fun. That was super fun. It's the symptom of good conversation.