AI ROLLUP #10: 70% AI Market Crash | Fair-Launch Tokens Good or Bad? | Virtuals’ $2M Fiscal Policy | Coinbase AI Hackathon
01:13:20 Share

Welcome Bankless Nation to the AI Rollup where we stay up to speed with the emerging trends and developments in the crypto AI space. I'm David Hoffman here with my co-host Ejaz. Ejaz, how you doing my man? What's up man? You haven't heard? It's a bloodbath out here, David. Oh, I know. I am well aware. Yeah, no, it's time to pack it in. This is actually the final episode. This is the last one. We're wrapping up shop. I'm kidding. Fairweather fans. Yeah, it is. The markets are crazy.

And I think we're probably at or near the end of this short-term dip. And don't crucify me if this isn't true, but let's look at the latest dashboard on cookie.fun, David. What are we seeing here, basically?

Essentially, we're seeing quite a bit of capitulation from some of these tokens' all-time highs. I think the market cap is sitting at, what is it, roughly like $8 billion right now? $7.5 billion. So what was it at its peak when we were doing this episode? At its peak, it was about $20. No, $20.5, I believe. Yeah, it just went to $20.5. And now we're sitting at less than half that.

of that. So a pretty crazy retrace. I saw a tweet somewhere where it said on average, some of these AI coins basically retraced about 60 to 70%, which to put things into context, it's not abnormal to see 20 to 30% corrections during a bull market. And things like Bitcoin and ETH and stuff. And we all know that like alts typically get hit harder.

But to see something like a 60% to 70% retrace, and by the way, this isn't just specific to AI coins. This is across every single sector of alts that we've seen so far. It's just a different market and a different type of cycle. And I don't want to say that this time is different, but this time evidently has been different, David. We've got virtuals pulled up here. I would actually say, relatively speaking, obviously virtuals peaked at, I think, a

a $5. Yeah. Yeah. A $5. I think it was like, what market cap was that? Roughly $5 billion, $5 billion price. Right. We are currently at like 1.4. So I think when people look at like the size and magnitude of the AI agent space, I think the virtuals chart is kind of just a microcosm emblematic of the whole thing. So like November, we were at 50 cents. Uh,

50 cent virtual price. January 2nd was the peak at $5. And then we are back down to one and a half dollars, $1.4. I think if I pull up AI 16Z is going to do something similar. Pretty similar, yeah. I mean, it's been more aggressive. I think just like a lot of energy came into the AI agent space. Yeah, way more aggressive.

And then it's like, I think people's like reality has caught up with people saying, oh, we have a lot more to build. And so people have, capital has flooded in and then flooded out pretty quickly. I mean, I do want to direct, for those of you who are watching, direct your attention to the bottom left of this chart here, right? Which highlights that, hey, this whole meta is only a few months old still. We're still literally like, you know, AI16z what came to life in...

at the end of October, right? And then it had this kind of parabolic run up until like $2.5 billion market cap and then it kind of like crashed. So I think you summarized it really well, David. Like the crypto AI sector, specifically agents, has seen a huge influx of capital and we may have, you know, just surged to ahead of time. We might have gotten

gotten a little over our ski tips we tend to do this yeah you know um and you know it kind of like that pico top was kind of like when trump released his meme coin and then melania a few days later and then everyone was like well hang on a second like is this the end um or is this just the start

And then the market got really impatient because, surprise, surprise, the American government didn't announce a strategic Bitcoin reserve immediately and setting them up like immediately. Right. Which is crazy to think. Right. Like a year back, David, if we had this news, we would just absolutely send. And it would imply that the market has probably priced something, some of these things in since Trump got elected. Right. So.

It's just like an interesting way to frame things. I know you have AIXBT pulled up here, but I kind of want to direct our attention to a different flavor

of the cookie.fun dashboard. They have this really interesting way of framing things, David, which is they're splitting AI coins into meme AI coins, and then like infrastructure framework, DeFi AI based coins. And why I like this view is it kind of shows us where the attention and capital is going into and it kind of like informs us of the type

of investors or dare I say traders that are involved in this sector right now. So number one, David, is meme AI coins. So these might be AI coins that, I mean, there's some examples to the right actually that will kind of like tell us. So Fartcoin up there, what else have we got, David?

Toshi, which is the Coinbase-based chain. MemeCoin, which apparently has like an AI integration. But I think these are, call these memes that have some amount of AI properties. Call it sentient memes. Right, sentient memes. And if we come out of this, we see number two is led by frameworks. So this is typically protocols or teams that you and I cover every week, actually. Yeah.

This is more of the fundamentals. This is like the layer one of AI agents trade. Exactly. And it's worth noting, actually, I kind of want to point it out here. There are very few protocols that show green in terms of like mindshare, certainly not price action. But if we want to kind of look at certain teams that are bouncing off some of these like lows, we noticed that the largest market cap coin which was able to do this was ARK right at the top of this, David. And

I was kind of thinking to myself, I wonder why? Because it's not necessarily a small market cap coin. And I would probably put it in the placings of virtuals and AI16Z. I think it's through a few different things. One, I think probably...

It's making a lot of progress on Solana and Solana is kind of being branded the AI chain or the AI degen chain of this cycle. And I know they partnered with Solana Foundation. We mentioned that last week, actually. So I'm thinking we're probably seeing some momentum there. And, you know, the news hasn't quite caught up because markets are just dumping. But aside from that, you know, AI16Z hit pretty hard in terms of like Mindshare and

market cap we're going to kind of dig into a little bit about why that might be a little later and then virtual is kind of like holding steady as we know virtuals has uh decided to expand to solana they haven't officially launched that platform just yet but they're going to be working towards it so all in all like um you know the market is reset the market is resetting um and

it'll be interesting to see how we kind of move forwards over the next couple of months. I think that resetting word is nice because if you just look at the virtuals chart, it has gone back down to a place. It's much higher, but it feels like it's much down, but further back down to reality, right? Things are like,

We are no longer full of hot air things are starting to find they're closer to a fair price You can claim that AI agents are just bots. Maybe you're even right, but nonetheless this industry is like real It's attracting developers. It is growing as fundamentals. There is like a loose roadmap for the sector as a whole So I think things have kind of come back down to reality which is you know, call it like an internal to a broader secular bull market and

bear market. It's a bear market and a bull market. And AI tokens, AI agent is like the frontier of the frontier. And so it takes like just a lot of things to go right for, I think, the animal spirits to find their way back to AI tokens. But right now, right now we're in fundamental zone. Well, I mean, I always ask myself, like,

am I just in this like tiny miniature bubble? Like, am I even focused on the right things? Am I like kind of like dreaming? I think it's important to look at AI mindshare just as a topic of conversation as a whole. Can you pull up Sammy's tweet here, David? Where he kind of shows a screenshot from Kaito, which captures the different kinds of market share. So for reference here, Kaito tracks very, you know, specifically sentiment

on crypto Twitter, which is where a lot of people kind of like get their investment advice from trade or just kind of research. And we see that AI is still dominating. Now we've seen better examples of this where AI just takes over 50% of the screen. It's now down to like, what is it? Like 30 something percent, roughly, David? 30, 35. It's still a solid like 3X ahead of the number two, which is me. Which is interesting because AI tokens are,

By comparison to meme tokens are getting wrecked in terms of price action. Yes. But AI as a conversation piece, as a mindshare piece is three times higher than meme pieces. Yeah. And you can also see that show up in like centralized exchanges. When AI tokens show up on centralized exchanges, they don't necessarily catch a bid. You know, we need more data to follow through on that. But when meme coins show up on centralized exchanges as token listings, meme coins go through the roof. Yeah.

So there's just a general appetite for meme coins. And it probably relates to the Lindy that meme coins have over AI. Yeah, there's a lesson there. Not quite sure what that lesson is, but there's a lesson there, which is, you know, maybe people that are just entering the space and wanting to ape

into crypto are just kind of they view kind of crypto as this gambling sector. And maybe if they see meme coins, they're like, Oh, I don't need to think too much about this. I don't need to read a white paper about this. I don't really need to dig into this. I just see it. Does it laugh? Does it make me laugh or does it not? Exactly. Exactly. And I think that kind of like ties into the financial nihilism that's been pervade across the sector over the last couple of months. I do personally think we're going to shift away from here. I think

with the US government setting up a bunch of regulatory frameworks. Yesterday, they came on and said, stablecoins are going to be super important. We're going to build frameworks, I think, led by Cathie Woods. That's going to figure out what all of this is going to look like. I think we're going to see a shift into fundamentals eventually. But one more point I want to make around this mindshare chart, David, is AI has been the dominant category for about a year now.

probably a year and a half actually i've been tracking this kaito stuff for since they pretty much launched and ai just still has been top of mindshare so there's there's probably something worth um paying attention to here maybe it's like an asymmetric trade or whatever that might be um but i mentioned earlier david like we should look at like okay so the market is dipping

which token should we focus on? Is everything going to zero or are there some things we should focus on? Well, if you pull up this tweet by Deep Value Memetics. - Deep Value Memetics. - Deep Value Memetics. I actually really like this account. It's a pretty small time account. If you guys are listening, just go give it a follow. I have no association with it. I just think they produce some interesting pieces. Typically they do long form stuff, but this one, I like that little market kind of summary here where if you look at large cap coins,

Obviously, you have Fartcoin leading the memetic AI category, you know, bouncing back strong. And then you had ARK, which bounced back plus 40% off the lows. So it's really interesting to see which coins are being paid attention to. You have a meme coin, meme AI coin there. You have a framework coin, which is in the form of ARK.

And then you have VV and Anon, which kind of sit in the kind of middleware slash DeFi AI category, right? And Anon performing strongest in the latter. Interesting. Is Anon, Anon is the token on base token.

that you can tweet out of the Anon account via ZK Privacy app, right? No, I think this is HeyAnon, David. Oh, a different one. Okay, because that one does not look good. Yeah, that is a different story. Yeah, not that one. This is HeyAnon, which is kind of like, I kind of think of it as similar to what Griffin guys are doing, where it's kind of like an agent execution platform and stuff like that. And then if you look at the mid-cap, so obviously lower liquidity, it's easier to pump and...

you know, also easier to dump. The mid cap coins are focused on largely agent execution platforms that are lower in their maturity, as well as agents themselves. So we have like grift, which is I'm

I believe Sphere 1's protocol. And then we have Buzz, which is a protocol that came out of the Solana AI agent hackathon, which I was able to help judge. And I saw the team kind of build it out live. And then Mobi, which is like the leading kind of crypto alpha agent on Solana. So kind of like similarly competitive to AIXPT on base, for example. So I thought this was a really interesting kind of thing to follow on. But like, we should also look at like inflows, right? If you pull up...

Elon Money's tweet over here. So what I like about this account is like he always tracks using Nansen, basically like what the smart money is buying. And you'll see that it represents the same kind of bounces that we've seen in tokens that we've just mentioned, right? We've seen ARK, Fartcoin.

Griffin wasn't specifically mentioned, so that's probably worth noting. And then Graph, I think the lower you go down on here, it's just a much smaller market cap, so it's easier. But I would say focusing on the top three gives us a bit of a...

hint or sign as to what might, you know, do the best once this market recovers. It is interesting to see Fartcoin is 100% a meme coin. You are seeing Fartcoin being talked about in like mainstream circles as an example of just like,

Well, Dogecoin was funny. It was a cute meme. Ha ha, Doge. So therefore, obviously, Fartcoin. Like crypto traders, Fartcoin, obviously, you don't even have to think about that too hard. It just so happens to also be AI. So it kind of makes sense that the meme side of Fartcoin is the thing holding up the price.

In addition, and like one day, I think this will be like synergistic with each other. But right now I think there's one, there's two sides of fart coin, the meme side and the AI side. And I think it's the meme side that's actually outperforming here. Yeah, hot air rises, you know? Yeah, it's a very reflexive asset. Actually, one of the most reflexive assets

in the cycle across all different categories. So super interesting to see. And I know we've spoken a bunch about ARK already, but the reason why I keep bringing it up is it's broken certain trends that hasn't been present in previous weeks. So this tweet that you've just pulled up here shows ARK for the first time ever dominating Mindshare amongst top AI agent protocols. Typically, it's always been AI16Z. Wow.

Typically, it's always been virtuals. And they've had some pretty big news over the last couple of weeks. But for some reason, and again, like I can't quite pinpoint it just yet. But ARK is kind of leading the foray here. And it might signal that the markets are seeing something that isn't so publicly obvious right now. Hmm.

Interesting. One thing that was pretty interesting this week was this tweet from Sean. I think a lot of the crypto Twitter, AI Twitter, the market is kind of processing this is Shaw's tweet that when did this come out yesterday? Yeah, yesterday. He said, I would advise you to think twice about launching a fair launch coin if you intend on using it to fund a long duration project, which sounds

At a high level, seems a little obvious, there has been a movement towards fair launch coins, kind of like in a reaction to the high FDV, low float meta. And so like meme coins really got a tailwind because the other tokens at the casino were these VC coins with very low flow and a very high FDV that kind of just is not a good deal for retail, right? So then we went to these fair launch projects. The problem with...

fair launch projects is that it's hard to incentivize the team doing the project if they aren't able to have a meaningful share of their tokens, which is well, while the whole like high FTV, low float meta is toxic for different reasons. One thing that it's not a problem for is incentivizing long-term team commitment to the project because, well, why is there a low float? Well, because the team is locked up invested. Um,

And so this is like the opposite end of the spectrum. And so Shaw put out this tweet talking about like, man, this whole AI16Z project, it's been fun. I've enjoyed it. But also there's this fundamental incentive alignment issue with this fair launch coin that much of the team doesn't really hold too much supply of. So we're not like sufficiently incentivized over the long term. That was kind of the big takeaway that I got from this thread. What did you take away from it? Yeah, no, I think that's,

That's pretty bang on. Let me add a few things. So you mentioned high FTV, low float. Just to kind of remind the audience, basically last year, everyone got super jaded by all these VC-backed crypto protocols because we really believed in the tech

back a year back, right? And then everyone learned about how the VC game kind of works. And they're like, oh, well, okay, that makes sense. They're early investors because they're taking on the risk. But it was kind of like uncovered that, you know, VCs actually own quite a large percentage of the supply and a very minor amount sometimes gets, you know, put out to the public. And this happens on some occasions, it's not prevalent across the entire landscape. And this put a lot of people, this made a lot of people quite skeptical. Yeah.

They started thinking, well, that means we're only going to be used as exit liquidity, et cetera, et cetera. And that's what led to this frenzied meme coin pump, David. Everyone was like, okay, well, I'm just going to be on pump fun and just gamble my money away instead of putting it into this high FTV, low float type situation. And ironically, some of these meme coins ended up pioneering agent coins, David, like GOAT and ARK, literally came out of pumped up.

Some meme coins backed into becoming a real fundamental Asian token. Yeah, it was very prophylactic, actually. We didn't actually realize that that was going to be the case. But we watched as these coins went from being a meme to adding utility at the heart of an ecosystem. Sounds great, right? And everyone was able to ape at launch. The team has barely any allocation, so everyone wins, right?

You know, decentralized investing, let's go. But teams are starting to face this different kind of problem that they're realizing in the wake of this heavy market crash in the sector. That is, you know, the team that doesn't have enough tokens to incentivize themselves to work on their vision can't really fulfill what their vision is. They have to like look at ulterior solutions and motives to make a profit in order to then

refocus back on their vision, right? Think about it. You kind of give away 90% of your tokens and you at most manage to scrounge maybe 5% for yourself and the team. The community loves you. The community loves you, right? But your team certainly doesn't and the developers that want to build with you that you

need, probably don't. And you know, that 5% isn't worth much to maintain the best builders to help you build, like, especially in a world of open source development, when you know, you have many pots cooking. Typically, with if we compare this to crypto startups, David, the team keeps a large chunk, which not only incentivizes them, but entrusts them with the ability to use that allocation in the right way, right? Maybe it's like, hey, you need to treasury.

It's your treasury. Exactly. Right. That makes sense. Obviously, the founders likely know best where all of this is going, not some random anon on pump fund that's like shit talking or shit posting. Right. So teams are starting to realize that the lack of incentives and capital to carry out their mission, you know, and they're kind of like forced to seek alternative ways. So Shaw writes about this, like literally, I recommend everyone to read this post because I think it's important and might

end up in a shift to something midway, you know, where maybe it's like... There's merits to all token distributions. There's merits to the high FDV, low float launch. It doesn't do great for the community. There's merits to the fair launch mechanism, which does great things for the community, but really terrible things to the team and the structure and long-term planning. And then we need to figure out how to meet in the middle here. Now, granted, I do think that the fair launch mechanism...

that also reserves a supply of tokens for the team is a pretty trivial fix. I think there just needs to be like, you know, a little bit more planning ahead of time before like Shaw launched a meme coin. And then he was like, well, wait a second. What if I like commit my time and life and career to this? So he didn't really have a moment to breathe to think about this ahead of time. But I think a fair launch, like a pump.fun launch, a Flare launch,

where you do a fair launch, but then you just have 30% of the supply that's vested in a team contract is a pretty simple fix to this. It just requires a little bit of forethought. Yep, yep, I agree. Yeah, I mean, I have no real comment here. It's just ironic that we've went to both ends of the extreme and the answer is let's just kind of like settle somewhere in the middle.

Yeah, what do you think this means for the AI16Z project? What do you think this means for Shah? How do you think the AI16Z community is processing this? So, okay, so to answer the last question, I think they're processing it badly, but that's understandable because typically it's the team and the leader that needs to have the vision and honestly the capability to execute from here. Okay, figure it out. You're in the situation that you're in, figure it out. My personal take is I think the team is...

a crew of amazing developers and builders. And on the Eliza OS side, so that's the framework and the platform that they're building on top of it, they're going to kill it, right? V2 is coming out around the corner. I'm excited to see what that is. I think the team also has the capability to adapt to this. Now, what exactly that strategy is, is yet to be determined. And I'm

honestly curious to see what they come out with. My personal opinion is they're going to need to cook something up that drives revenue or success of their framework that they've built out that they've open source for everyone to tie value back to AI16Z. Whether that is driving more AI16Z tokens back to the team in some sort of a treasury. So maybe they take a percentage of every transaction that's happening on every

part of their platform that they're going to launch back to a treasury, or if they do certain launches where they capture certain values, and they're able to kind of like accrue it into their DAO treasury, whatever it is, they need to amass a stockpile, a strategic AI agent coin reserve, if you want.

for AI16Z. And I think that they'll probably pursue a path like this. There are other projects out there, David, that are in similar positions. I won't name names, right? I was going to ask that, yeah. Yeah, I mean, there are some. I don't want to name names because, you know, what's the point of like footing? But it's important that a bunch of these

teams start thinking about these things. And it's evident from this post that they already are. Right. And I think the timing of this is very much because we have seen a 40, 50, 60% drop in token prices. This isn't happening if we have 3x from here. Or even if we go 20% up tomorrow, we're not going to be talking about this because no one will even care. Suddenly you have a larger stockpile. Suddenly your 5% is worth a lot more.

Yeah. But if you're here for the long term, like we all know AI agents are not going to be built overnight or in a week or in a month. This is going to be a multi-year long arc. And so teams need to be thinking about multi-year long incentive structures in treasuries.

And I don't think the AI teams, other than maybe virtuals is different because virtuals has literally already been around for well over a year. So they've already been thinking like a startup. And so maybe that's an exception. But this space has arisen so quickly that I highly doubt teams have been looking, thinking about like what happens if we go into a multi-cycle, not like four-year cycle, but like a multi-era like bear market where funding is dry. Yeah, agreed.

With over $1.5 billion in TVL, the METH protocol is home to METH, the fourth largest ETH liquid staking token, offering one of the highest APRs among the top 10 LFTs. And now, CMETH takes things even further. This restaked version captures multiple yields across CARAC, Eigenlayer, Symbiotic,

and many more, making CME the most efficient and most composable LRT solution on the market. Metamorphosis Season 1 dropped $7.7 million in Cook rewards to METH holders. Season 2 is currently ongoing, allowing users to earn staking, restaking, and AVS yields, plus rewards in Cook, METH Protocol's governance token, and more. Don't miss out on the opportunity to stake, restake, and shape the future of METH Protocol with Cook.

Participate today at meeth.mantle.xyz. Are you ready to swap smarter? Uniswap apps are simple, secure, and seamless tools that crypto users trust. The Uniswap protocol has processed more than $2.5 trillion in all-time swap volume, proving it's the go-to liquidity hub for swaps. With support for growing numbers of chains, including Ethereum, Mainnet, Base, Arbitrum, Polygon, ZK-SYNC, Uniswap apps

are built for a multi-chain world. Uniswap syncs your transactions across its web interface, mobile apps, and Chrome browser extensions, so you're never tied to one device. And with self-custody for your funds and MED protection, Uniswap keeps your crypto secure while you swap anywhere, anytime.

Connect your wallet and swap smarter today with the Uniswap web app or download the Uniswap wallet available now in iOS, Android, and Chrome. Uniswap, the simple, secure way to swap in a multi-chain world. Actually, speaking of virtuals, David, the team has actually been working on something that hasn't been seen in the market before. And maybe this is because they've been around for a while and they've kind of like seen how these things play out. So they're introducing something known as their fiscal policy for their virtuals nation, right? So this sounds very broad and bold and

honestly kind of crazy. Virtual's Nation Fiscal Policy. I love that title. Okay, so the point around virtuals and the team and their vision is they believe agents are going to exist everywhere, right? Doesn't matter what chain they're on. And they've evidently expressed this in their roadmap by announcing that they're going to be launching on Solana. For those of you who don't know, virtuals has primarily been based on the

base chain, which is an EVM L2. So the fact that they're moving over to Solana is pretty nuts, right? And they're going to juggle two ecosystems, and they're probably going to expand to more in the future, right? But the whole idea is, okay, if these agents exist on different chains, and they're able to transact with each other, and it doesn't matter where your agent is, they can just

communicate and converse with each other, then you're kind of building out some kind of a separate economy. It's not governed by geographical borders. It's not governed by land. It's governed by virtual space. And so if it is governed by virtual space, there needs to be some kind of...

policy or tax system that kind of keeps your stakeholders or participants in check, you know, David, making sure they're not doing illegal shit, but also making sure they're paying back to the society that supported them, that's allowed them to kind of like build into something, right. And so virtuals announced this,

fiscal policy manifesto, right? I'll break it down. But it basically says like right now, agents on virtuals can go to the platform and they can be designed and then they could launch, right? And there's like a bonding curve situation and all that kind of stuff. And they've addressed each stage of this agent's launch

And they've mentioned now how they're going to be introducing new kinds of mechanisms or procedures that would help either enhance or kind of cut out some of the riffraff that's happening at each of these different stages. So basically making it a more functioning society, essentially, right? So if we look at like pre-launch, you know, they talk about this venture partner network.

And what they mean by that is, so far, it's been a free for all, right? You turn up, there's not really been any guidance. There's a doc that you can read online, and then you can kind of like just get on with it and launch your agent. They're now going to try and take a more curated approach, kind of similar to what we've seen with like ARK and a few AI16Z team launches actually as well. So what they mean by this is, they're going to have several venture partners, which will act as curators. So think of like...

a profile on Instagram that has a verified checkmark, right, you're going to probably trust the content that comes out of that account, same as on Twitter, right? You'll see venture partners that will have like a verified checkmark on their platform, that will be able to offer up certain opportunities to invest in agents or teams that are going to be launched. And

You're trusting them that they've done their due diligence. You're trusting them that they've kind of figured out what this might be. And they're kind of like funding and supporting the right things, right? The second bucket is they're calling post-launch pre-sentience. So virtuals has this kind of launch mechanism where, okay, once the agent reaches its bonding curve, it launches, right? The token becomes...

available on whatever decks they support. But then like it's still a low market cap, right? The team could just dump this entire thing. But they have different kind of like milestones. If your market cap for your agent hits 50 million, then it's considered sentient. Or if it's 100 million, then it's considered, you know, something a little more useful than sentience or whatever that might be. So the post launch pre sentience, like,

they're going to try and focus on funding teams that believe they have, you know, full, like much better potential above that particular category, right? So they're calling it prototype to sentience virgins, which is like a play on, I guess, virgins, but like, it's like virtual degens. Well, it's virtual degens. And it sounds hilarious if it's like virgins.

But they'll be basically provided with virtual grants to scout for like high potential teams as they put it out here. And as funny as that sounds, David, I've never seen teams take this more seriously. If you pull out actually the next tweet by Vader,

He talks about, or this team rather, talks about virtuals quantitative easing, TLDR, right? - Right. - Essentially- - Okay, so let me checkpoint this. Virtuals, it collects revenue from the virtual system. We know this, it collects fees from trading fees on the bonding curve using the virtuals token. And now it is implementing a top-down fiscal policy that is meant to encourage

new businesses, aka agents, in the virtual universe. And it has bestowed some powers to certain people that we're calling venture partners to go out and actually sprinkle out some cash in targeted precise directions at precise teams because they think it's a good bet to

help encourage the teams grow that AI agent on the virtual platform. And so taking some of the revenue, putting it back into the bottom of the economy, the virtual economy, hoping it trickles back up and then we can create a feedback loop, which is like basically a taxation and like social safety net system, not a social safety net, but collect taxes, put it back out into the economy in ways to simulate the economy and then collect taxes again and do it all over again. That's what we're talking about here. Yeah, it runs pretty much like

an actual nation would which is like something we haven't really seen in terms of protocol governance in crypto before right right and so what um this Vader tweet is highlighting here is for a specific one of those specific programs which is the prototype to sentience program um basically two million dollars or 900 um thousand virtuals at current pricing um will be put into the prototype prototype

stage at this point, which means that these different venture partners will be able to allocate this equivalent amount of money to fund certain teams that they believe have potential to become the next billion dollar protocol, for example. And the reason why I said this is like super interesting is I've never seen a bunch of teams get so locked in immediately.

I was on a Spaces yesterday where they were doing kind of like a shark tank pitch session. I was on it for an hour. Was it good? Yeah, it was good. And the winner of that basically got 10,000 virtuals or whatever it might be equivalent to do their next stage of funding or to get them over their next milestone. So these venture partners are locked in. And I want to point that out because typically the ecosystems that flourish after a bear market or a bear dip

are the ones that are just locked in building during the bad dip right it's not someone that's like kind of like habitual depending on the mood of the market so if you pull up this uh this veda

this second vader tweet virtual's den episode one where you know they basically had a bunch of these teams pitch 45 minute spaces 30 minute pitch um and it was pretty cool to like see this whole like judging criteria and i'm already seeing a bunch of other formats uh like this uh appear so uh really impressed with like what virtuals is doing and how they're kind of approaching this it's pretty crazy to watch a team kind of like try and build an actual economic zone uh for agents and

And in my opinion, if anyone's nailed the tokenomics so far in this space, it's virtuals, you know, pre sentient scouts, venture partner model, post launch funding, they've thought about everything agent token tax fees. They're building what I would term as a sustainable David ecosystem, right? And sustainable isn't really something that has been habitual to crypto. We've seen like meteoric rises and meteoric dumps. But sustainability is going to be super important if you ever want to become a mature protocol.

I think this is really just very apt in timing for the specific week that we're talking about this, which is like deep, deep, deep AI agent token bear market week, where we are now realizing that this is a long haul. You know, some of the people who were in this industry for a quick buck are leaving. And so now we are going back to, I mean, you and I were going through the, um, 2018 through 2020 bear market together and the teams that made it through that bear market, uh, the link Marines, uh,

Ave, Synthetix, they all found ways to engage and create a hot campfire for their community to gather around. Things to do, things to pay attention to, ways to make money, even if it's small, in the bear market to keep that community entertained and to keep that fire going. And so this is what I'm seeing here. And for the average person who's in this economy, which is just like a virtual token holder, this shows that there's something alive and there's something happening.

And there's a reason to hold onto your virtual tokens because you know, there's actually life in this project and there's a campfire that's burning and people are gathered around it talking about like these, these hackathons and this judges. And so it makes me, it reminds me of like the 2018 through 2020 bear market and the projects that made, made it through that bear market just had something to do and, and,

and they kind of like outsourced their growth to the public in creative ways like this. Yeah, at the end of the day, a lot of the criticism comes down to everyone thinks this is a transient meta, right? They think AI crypto is just another kind of meme thing and it'll die off soon.

I personally do not believe in that at all. I think you'd probably say the same. And so the only way you could validate that is by watching teams and seeing what they're doing, seeing what they're building, seeing what they invested in, seeing if even the big head honcho teams are not taking their foot off the gas. And it just seems like virtuals is doing exactly that. Yeah.

Yeah, that wasn't the only hackathon that happened this week. Coinbase had an AI hackathon. Were you paying attention to this? Fill me in on this. Yeah, dude. So, okay. So our job here on this show is to give everyone the lowdown on crypto AI every week, but it's also to be the source of new alpha and trends. And part of that involves being aware of teams that are just getting on their feet or, you know, taking a unique approach to a problem, right? And hackathons are the best way to do that.

And it just so happens that my old stomping ground of Coinbase held one this last, you know, this recent weekend. And the winner, Mamorico is the name of the agent, was worth highlighting, in my opinion, because it's designed to not just be a single agent, but a swarm of agents. Now, before we get into this demo video, I just want to kind of point out why I think the Coinbase hackathon in particular is worth highlighting. Funnily enough, David, and I'm saying this as a previous employee there, usually they're not

right on the edge of things. You know, usually Coinbase is not on the frontier typically. They're typically, you know, they kind of wait a couple months at the very least and kind of like see how these things mature. But it was really surprising and

pleasant surprise for me to see that they were one of the first teams to come up with an AI agent kit from their Coinbase SDK. And the AI agent kit served as a tool or framework for those people building agents to plug into base or whatever different kinds of chains and perform a bunch of different on-chain actions. So it was very relevant that they held a hackathon so that teams

this was held in the San Francisco area, were able to come along and hack and build something super cool, right? So they had a bunch of winners. I wanted to point out two specific teams. So one was the first place winner, Mamorico, as I mentioned earlier,

In this demo here, Luke, shout out Luke. He was the guy that built out the staking stuff at Coinbase or helped lead the team there, describes the agent's function to find the best and safest yield strategy for your USDC and Bitcoin. So Bitcoin in this context is Coinbase Bitcoin, I assume, on base. CB Bitcoin, yeah. Exactly, CB Bitcoin. So this agent basically has four worker agents, David, that perform analysis for it. So-

So two agents work on USDC. One does like a health check on the USDC markets, you know, just to make sure nothing's wrong, nothing's weird, the market is healthy. And then the other searches for different available yield options for that USDC on base. And these agents run in a loop every 15 minutes, so they get refreshed. So it's kind of like this like very basic but potent version of an agent swarm that's like constantly updated.

So I thought that was a pretty unique one to point out because I've kind of been thinking like, where does this agent swarm thing end up? Is it just like a hype term? Like my original question is like, why can't you just get a single agent to just do all of it at once, right? And there are a bunch of engineers that support this view, right? But at some point, I don't think it's...

to assume that one agent rules all, right? There are going to be different teams, different funding. People are just going to build out their version or inflection of an agent to serve a particular subcategory. So whether you like it or not, there's going to be a world full of agents, right? And I don't think it's going to result in one agent capturing it all and the other agents just dying off. I

I think it'll be a way of like getting these agents to kind of work together. Yeah. I think that probably goes pretty deep to the heart of the very core of this AI universe, which is, will there be one God AI LLM algorithm that is superior to all others? Or will it be many, many, many, like countless amounts of individual niche experts? Usually when there's a spectrum like this, the answer is, well, it's somewhere in between.

When it comes to like optimizing yield, there's something about like yield optimization and an AI agent that just intuitively goes hand in hand. Because if you're a yield farmer, yield optimizer, you are basically a single human agent scouring DeFi for optimizing yields. And it seems like a very simple thing to automate. In fact, we did that in 2021 with this thing called Yearn, which was basically an orchestration of not agents, but bots.

that optimize yields. And it seems to be something that's very close in proximity to just being able to be agentified. And I think I've seen this a handful of times. It's just like yield optimizing agents. What that agent is actually doing

is hard for me to understand, like how that's different than a bot. But I can just imagine that like, well, if I want something to automate this, it's an agent. And I think really the market is going to wait for validation that an agent can do it better than a human.

Because I'm still not ready to give a large amount of my net worth to an agent to manage on chain. But I could see yield optimization as like table stakes, like the entry for this kind of activity. Yeah, well, just on your point of like delineating whether a bot

is different from this agent. Well, typically you would expect that a bot has predetermined things to fixate on, right, David? So it's like, you have to spend all your time programming this bot to specifically look at this protocol, this asset pool of this protocol, right? And never take your eyes off it. You're not looking at anything else. And

New protocols don't exist. New protocols don't exist. It's like literally it's got its blinders on. It can literally only see what's in front of it and nothing around it, right? And agent is the kind of like converse of that where it is aware of everything. It's plugged into all the on-chain data. It's aware of...

crypto Twitter. So if a new protocol announces that they've just launched, it's going to check it out. It's going to see whether they've deployed smart contracts on chain, and then it's going to evaluate it, you know, it's going to see how many users it has, how much TVL it has. So you're betting that the agent is going to be like a human David, smart, aware and able to digest and analyze whether the risk is worth taking for a particular thing. And to kind of like answer your question on, you know,

um, you know, whether DeFi or like, you know, these human agents are the right way to, to kind of replace, right. Cause you mentioned human agents are basically those ones that are optimizing for yield farming. It makes sense, right. DeFi was the first thing that kind of exploded outside of the, um, you know, just store of value sector or proposition of crypto because it was financial rails that were built and we could just

analogize with and kind of like build an alternative product. It makes sense that agents kind of capitalize on this first. That's why we see DeFi AI trending the most. So yeah, super cool. There was also a runner up that I just wanted to shout out, which was a social media agent set up.

where advertisers could post bounties for marketing agents to fill. I'm not going to pull up the tweet, but I just kind of want to point it out. Like whichever agent had the most tweet engagement would receive USDC rewards. And we've seen agents like AIXBT gain larger mindshare over time and outcompete every human KOL. So why not this time? And why not in this format? It kind of like redefines what the bounty program looks like or what advertising distribution might look like.

I will say there are very few AI agents left on my Twitter feed that I do not block after AI XP team because there are so many and they are so goddamn annoying. Yeah, I do the same. Yeah. Yeah, I do the same. There's still a lot of work to be done in order for AI agents to get my interest on Twitter.

for sure like probably listeners are probably feeling the same yeah but they are getting smarter david um open ai actually this week announced a new agent so last week they they teased an agent which was like you know gonna book your travel for you book restaurants whatever but now they've released a pretty much a serious agent um called deep research um the way i want to kind of like

pitch this agent and its capability is um do you know those like institutional reports david that take like a bunch of analysts that work like some of the analysts research report do you know what i mean and they're like 30 pages long they have all these like unique graphs and data and they're super useful but that's it's a lot of effort it takes teams of people to make all right

Yeah, all of that is getting condensed into a single agent that could produce that in about five minutes now. Oh yeah, and it's not just anything to do with financial markets. It's whatever category, profession, or field you choose. There are a number of examples of this that exist. I think Sam Altman in this tweet mentions right now, like he was trying to figure out

something to do with a project that he was focused on. And he just ran it through deep research and it came out with a 10 page report, which was just honestly insane to see. And so the question- How much compute does that cost? Well, yeah. I mean, he describes it as very compute intensive, but the fact of the matter is they've put this for pro users to access and run on their MacBooks or their normal regular computers and it's working. People are having- Wait, you can do local inference on your MacBook?

I don't know about local inference, but the point is you can get access like you do to chat GPT. So if you have a pro open AI account, you now have deep research as an agent that appears and you could talk to it and say, hey, listen, I want to be, I'm a medical practitioner and I'm trying to figure out A, B, and C. How do I figure that out? If you pull up actually this tweet by Felipe Millon,

I think he maybe worked on the team or whatever that might be. And he kind of gives you this personal story here where, you know, his wife was diagnosed with cancer and, you know, obviously it turned their life upside down. And they wanted to get an idea of whether, you know, she should do

radiation after chemo or whether they should leave it or whatever that might be. And there's a ton of... He tweets out, we uploaded her surgical pathology report and asked for guidance on whether radiation would be beneficial. What happened next was mind-blowing. What happened next? So basically, they asked deep research, not really expecting much, and deep research spent six minutes crunching every single latest cutting-edge scientific paper that

analyzing all the contents of it, pulling out the things that were relevant for this specific lady's circumstances and delivered a full report as to whether she should or shouldn't ensue upon a path of radiation after chemo. And he just said that just kind of like changed their complete

idea of all of this. Why? Because it not only gave them confidence in a path, but it provided proof. It provided evidence. And it was built off of the stuff that these doctors are already looking at. Also, David, it might outperform the doctors at their very own job. Think about it. How many doctors do you know spend

18 hours of their day at practice and then go home and read a bunch of papers. It's probably unrealistic, even when you're at top of your field. But now you have this genius little assistant that is accessible to anyone that's paying whatever, 200 bucks a month to now just devour everything that is online and within the medical field. And that's just the medical field.

it's insane so we got here because i was talking about how ai agents on twitter are super annoying and dumb and very low utility and i'm blocking every single one of them and then you're telling me to the story of like this this guy who used deep research to pull out a report so it's you're saying well no we know that there is value here we know that these

LLMs can produce something extremely useful. I'm telling you not to get too comfortable, David, is what I'm saying. So I bet that query from Felipe about what he should do about his wife's cancer, I bet that query costs OpenAI like $5, $10, something expensive. So it makes no sense for swarms of AI agents on Twitter to be popping out queries like this over and over and over again. But we know that

all of these services, all of these resources are going to become commodities. The whole deep seek thing taught us that like this thing is accelerating towards both open source and extreme in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in, in,

these, this, what do they call it? Deep research LLM. Yeah. As actually what is table stakes for being an AI agent that's connected to Twitter. Well, David, I mean, in true open source fashion, remember open AI is a centralized company. It's ironically closed AI, but an open source competitor has already emerged. And it's been like- To deep researcher? Yeah, to do it, to basically mimic and do the job of open AI's deep research. And it's hit a 54%

score on the same validation set that OpenAI's deep research agent scored, which is just like insane. And I haven't used this agent just yet, but I've read a bunch of feedback that people have given from people who actually have used this and it seemingly performs as if not better than the actual model in different kinds of sectors. So just insane to see. I just want to point out like

how quick the open source community just catches up on this, right? Remember, literally less than...

And a week and a half ago, we all believed this was just a big company game. And then DeepSeq comes along and it's like, hey, actually, there's a new reinforcement learning technique that can now allow you to reproduce some of the best models for next to nothing of the cost. And then we're seeing this kind of like similar pattern play out with, you know, someone recreating this deep research agent. So just super, super exciting stuff. The Arbitrum Portal is your one-stop hub to entering the Ethereum ecosystem.

With over 800 apps, Arbitrum offers something for everyone. Dive into the epicenter of DeFi, where advanced trading, lending, and staking platforms are redefining how we interact with money. Explore Arbitrum's rapidly growing gaming hub, from immersed role-playing games, fast-paced fantasy MMOs, to casual luck battle mobile games.

Move assets effortlessly between chains and access the ecosystem with ease via Arbitrum's expansive network of bridges and onrifts. Step into Arbitrum's flourishing NFT and creator space where artists, collectors, and social converge and support your favorite streamers all on-chain. Find new and trending apps and learn how to earn rewards across the Arbitrum ecosystem with limited time campaigns from your favorite projects. Empower your future with Arbitrum. Visit portal.arbitrum.io to find out what's next on your Web3 journey.

Celo is transitioning from a mobile-first, EVM-compatible Layer 1 blockchain to a high-performance Ethereum Layer 2 built on OPStack with EigenDA and OneBlock finality. All happening soon with a hard fork. With over 600 million total transactions, 12 million weekly transactions, and 750,000 daily active users, Celo's meteoric rise would place it among one of the top Layer 2s built for the real world and optimized for fast, low-cost global payments.

As the home of the stablecoins, Celo hosts 13 native stablecoins across seven different currencies, including native USDT on Opera MiniPay and with over 4 million users in Africa alone. In November, stablecoin volumes hit $6.8 billion, made for seamless on-chain FX trading. Plus, users can pay gas with ERC-20 tokens like USDT and USDC and send crypto to phone numbers in seconds. But why should you care about Celo's transition to a Layer 2? Layer 2's Unify Ethereum, L1's

By becoming a layer two, Celo leads the way for other EVM compatible layer ones to follow. Follow Celo on X and witness the great Celo happening where Celo cuts its inflation in half as it enters its layer two era and continuing its environmental leadership. All right, EJS, I'm going to take over for the lead of the next section. Are you ready? This doesn't usually happen. Usually I'm your lead. I'm taking over lead now. Take the steering wheel, dude. Let's go. Okay.

So if you remember on our second episode ever, we talked about this AI agent game where there was like a treasury, like a honeypot and it was guarded. We still have not figured out how to pronounce it. F R E Y S A. Um,

It was this little game of this $50,000, $20,000 in ETH that's guarded by this agent. And the goal was for humans to prompt this agent to try and trick it to get the pot. And the cool mechanism was that, well, if you wanted to send a prompt to the agent, you actually had to pay some small fee in order to send a message, like 0.05 ETH.

But as more messages were sent, that fee would tick up. And so that pot got larger and larger and larger. This got kind of viral. Somebody finally cracked with prompt engineering, finally cracked it, convinced the AI to give him the pot of like $80,000 or something. John Woo turned this into a Twitter thread on Twitter. Elon Musk, it caught Elon Musk's attention. He responded, interesting. The Freya token price pumped and then fell back down. Yep.

That project. That project. Okay. Turns out Freya is like much different than a little game of, you know, a FOMO 3D type game for those who are around in 2018 where you are filling up a pot and there's a little game. It's this team with this very ambitious endeavor to create sovereign AI that's

And it's actually impressively hard to figure out how Frasier, what they're doing. They don't have a lot of content out there, but a couple of my friends are deep in the Frasier world. Eric Conner, when he removed his .eth name, he put on a Frasier profile picture.

So the FreightJet team, what they're building, the arc of what they're building is a set of tools that all kind of compose together once the tools are refined, perfected, ready for production. They kind of compose together into what they want to create, which is sovereign, on-chain, embedded AI.

So truly autonomous. You cannot turn this thing off. This thing has life. This thing develops and grows and it is fully embedded on chain. So like really capital S sovereignty when it comes to an AI agent.

where it starts to really get some human-like qualities, kind of what the early few episodes we talked about, where these agents are real. They are equivalent in terms of permissionlessness access to these blockchains. Wait, so I want to unpack a few things here, David, because I think what you're saying is super important, right? So blockchains for the longest time has always been like, one of their value props has been verification, right?

A single source of truth, et cetera, right? And we're now seeing this kind of be carved out by the Freya, Fraser, Fraser team. And,

In terms of like their tool set that they're building. And I've seen a number of teams actually take this approach. And I believe the Fraser team is focusing on using trusted execution environments and also some kind of verification procedure so that you can trust that what the agent is doing is in fact what it's doing so that it can act sovereign. And what's interesting about this, David, is if you have a sovereign entity that is this agent, right?

that hands are wiped clean, David isn't involved, this agent is acting not on behalf of David, it's just doing its own thing. Who's liable for this thing? Is it the random Anon devs that have built this out? Or do we end up, are we running back Tornado cash? I hope not. Where does this end up? The second thing I want to point out before you comment on this, David, is how cool is it to go from a gamified bounty

Where it's like, hey, I'm an agent. I'm holding a stockpile of $100,000. Try and hack me if you can. I've been programmed to basically not give you this money. But if you can prompt engineer, if you can prompt hack me, I'll give you the money, right? That's kind of, they're basically building out community tested tools to make this trusted execution environment, this sovereign agent work.

Be trustworthy. You've mentioned so many times, David, I don't know if I can trust an agent, but what they're doing is they're kind of like stress testing it, right? They're saying like, if we put this out into the wild. Gamified stress testing. It's really cool.

Yeah. Yeah. And I think this is why I've been excited about the AI agent meta, why Bankless has invested time and energy into this AI agent meta, because there is a big difference between what I just like reductively call Web 2 agents versus Web 3 agents. There's Web 2 agents with the properties of Web 2, right? A centralized server, you know, ownership by a person or a company.

It's gate kept, it's controlled versus Web3 agents, which are fully autonomous, fully trustless, fully sovereign. There's some alignment between bearer assets like crypto and then the sovereignty of AI. And I'm interested in fully independent, fully sovereign AI agents. And the Frasier team is zeroed in on that direction.

Olaf Carson, he was on the Unchained podcast with Laura Shin. There's a three-minute segment that I want to play that I think kind of illustrates what I'm interested in, what Frasier is building. I actually found this link from the Frasier Telegram group when somebody just said in the Telegram, Olaf just explained Frasier without actually naming it. So I'm going to play this clip and we're going to put this in the podcast. Let's go. Okay, so, you know, this kind of... I've been thinking a lot about these like social media curation markets, right? But the second...

thing or maybe first thing actually that I've been thinking a lot about is pretty much embedding autonomous agents on chain, right? And giving them wallets that they uniquely have access to and sort of go out into the world and hustle and try to make money, right?

And so, you know, I think it is so tantalizing the idea that I could own like, you know, a percentage through a token of some agent that just never stops working. Right. He's he's out there just working.

whatever it is, right? It's like posting, messaging people, making AI music, right? And launching it, making AI videos and content, like trading in DeFi markets, like market making, running some quant strategy in DeFi markets, doing venture style investing. Eventually, these AIs are going to be more sophisticated than humans at pretty much all this. And I really think that giving them an

on-chain autonomous identity will be the best mechanism to bootstrap fundraising. So when you have this new AI system and you say, this thing is going to be an incredible market maker, and I tuned it specifically for that use case, I need to crowdfund initial capital. You sort of launch the AI. It's

it's owned in a sense by all the token holders, right? But it's totally autonomous and trustless. So even the dev that made the AI, you're not going to really need to trust them per se, right? Like they're not going to be able to shut it down or change it or steal the funds or anything like that. And so where we're headed again, I think is sort of pump fund for AIs, right? Where when you get the ease with which you can launch an AI agent,

to be like really, really, really easy. And you get the ease with winch, you can embed it, you know, inside like a TEE,

style architecture where it has its own private key. This is like the secure enclave on the iPhone, right? And it can access its own private key uniquely. You make that really easy and then you make tokenizing that and crowdfunding it like pump fund style really easy. I think we see an explosion of financialized agents that are out on the internet hustling while you sleep and you own this sort of portfolio of

you know, autonomous agents. Like, it's like, I own a few that do market making for me. I own a few that post in social media. I own a few that make video content. And they're sort of out there competing, you know, with each other, going head to head. So I think that...

perfectly encapsulates some sort of like north star of almost everything that we talk about here on the ai roll up when we talk about arc or virtuals or ai 16z they're building out like the pump.fun side of things when we talk about d5 ai they're talking that's like the ai market makers or the ai quant traders everything kind of aligns with that sort of vision of how do we make

AI agents on chain? How do we make them verifiable? And that connects me to the thing that I want to bring up next, which is what EigenLayer has been talking about lately. So a lot of people who have been paying attention to Sriram and EigenLayer, there's actually a lot of resonance with AI that Sriram has a background in and what EigenLayer is good for.

um actually before he uh was a was it was a startup operator with eigen layer he was actually writing papers yeah he's a he's like a he's like six ai machine learning deep learning uh ai like papers and though he's now kind of applying that to eigen layer so avs's and eigen layer used to be called actively validated service services they're now calling them autonomous verifiable services

And what EigenLayer does, which is E3 staking, pairs really, really well for making agents verifiable, imbuing agents with Web3, like upgrading them from Web2 to Web3. And the reason for this is simple. So EigenLayer put out this tweet thread, which is kind of a roadmap to how to make AI agents verifiable. And very simply put,

You can't put AI agents on chain. You can't do that. Why can't you do that? Well, AI models are massive running inference on chain. That's just not what blockchains are good at. Even Solana, you know, very fast, very high bandwidth blockchain AI resource demands are larger than blockchains as a category.

Inference compute costs are large. Memory is a very high intensity thing. There's not enough memory to do AI agents on check. Do you remember previous cycles, David, where it's like put data on the blockchain? It just doesn't make sense. It doesn't make sense. It's not what it's for. It doesn't make sense. Yeah.

And so what EigenLayer wants to do is, well, we have this amount of ETH that is restaked. You can move everything off-chain with an EigenLayer AVS, and then you can receive crypto economic guarantees about off-chain inference, off-chain memory, any off-chain compute costs, any off-chain resources, etc.

You just use an eigenlayer AVS to verify, to imbue any sort of agentic activity that we want to have verifiable. You can imbue that with crypto economic properties via eigenlayer restaking. So this is what Sri Ram is calling the eigenlayer agent universe. And he thinks that in addition to many of the other reasons why an eigenlayer AVS would be useful, it's actually agents that are going to be able to refer to each other, have an interagent network.

Because how do agents trust other agents if they are themselves black boxes?

Well, Eigenlayer thinks that it can solve that by imbuing these agents with Web3 properties by making their slashable offense if they do some computation off-chain that doesn't align with their commitments on-chain. And so it's interesting to see Eigenlayer, which is like neutral protocol infrastructure, it's just restaking, but it actually works very well with an agentic economy. It allows humans to trust agents more and

because of that, it will also allow agents to also trust each other more. And so I think maybe adding restaking crypto economic guarantees to agents and their resources can kind of help facilitate this economy. David, can I ask a dumb question to you? Sure. Sure.

So if these agents run or are tied to Eigen layer AVSs, right? And in my understanding, Eigen layer, very Ethereum based, obviously, because it's restaking ETH. These agents are going to be cross chain in my assumption, right? Does this affect that underlying mechanism in any way, do you think? Like what happens if an agent, you know,

Cheats on Solana. Yeah, yeah, yeah. What if it does something on Solana or another L2? Yeah, that ought not matter because really if you can create a proof, if you can prove that there was cheating done, it doesn't matter where the cheating happened. We call this in Eigenlayer and Verbiage, you call this an intersubjective falsification.

Remember the intersubjectivity? Intersubjectivity, yeah. So long as it's intersubjective, as in multiple humans can all look at the thing and agree on a truth about it, doesn't matter where that event happened. The only reason, the only thing that EigenLayer is actually using is crypto economic stake. And so it's just using Ethereum for its ETH and for its ETH to be staked.

And so if you lie somewhere, if an agent cheats by claiming to use off-chain infra in one way, and then it uses it in a different way that it promised not to, well, then your stake gets slashed. The only thing that matters is that there's a bond that's slashable. Got it. That makes sense. Yeah.

Yeah. All right, you guys, that's my rant. I'm going to give you back the driver's seat. That was a good rant, dude. You like that one? I like you digging the steering wheel. That was good. It made me think about some things that maybe I haven't explored too much. I need to dig into the whole eigen stuff. Okay. One last thing to round this out. Take us home. Okay. So I like to end everything.

every episode with a future facing trend that we see emerging that might become a bigger thing. And listen, eight times out of 10, we're going to be wrong. But for the two times out of 10 that we are right, hopefully it's going to become something pretty major. Now, the thing that stood out for me this week, David, was, uh,

or is voice enabled agents. Now I put out this tweet that kind of summarizes my thoughts here. I think voice enabled agents are gonna be 100% more addictive to engage with. But first of all, like what are you talking about Ejaz? Like what is a voice enabled agent? In my opinion, it's an agent that you can talk to the same way that I'm talking to you right now, David. It can understand what you're saying regardless of what language you speak or what accent you have. And it can- - Hey Siri, turn on my lights. - Exactly, right. And it can perform actions like that, turning on your lights

or tasks that are much more complex on behalf of you. And it occurred to me that this feature would be way more addictive for a human to engage with than, you know, just typing or writing or, you know, interacting with an agent in some other way, right? So the reason why I think this is important is it kind of puts you at ease, David.

I use the example, or I state the example in this tweet that I've started using ChatGPT very differently now. I pick up my phone and I'm just talking to it. Well, I've got my AirPods in and it's voice mode. And David, I've noticed that I'm giving it way more information than I normally would if I was typing. If I'm typing, I'm thinking, right? I'm saying, oh, I shouldn't say that. Or let me rephrase that. When I'm talking, I'm caught unawares. No filter. There's no filter. I'm like, oh yeah, this and that, right? You could just see the text.

length of each of my prompts is way higher when I'm speaking to it. And what's interesting about this is it catches each individual human at their most natural form. So it kind of like it kind of like feeds into Okay, who is he jazz, you know, oh, this is what his voice sounds like, this is the kind of tone that he's expressing. Maybe I can revert in the same kind of way. And I don't know if anyone's tried out the voice feature on chat GPT. It's really good.

It responds almost immediately. So the point I'm making-

And it sends the voice note. So when you're talking to ChatGVT, are you dictating into text or are you sending a voice note? No, no, no, no. I'm talking to it. It's like I'm having a phone call with someone. And the response might be, hmm, it's a good suggestion, but actually I think you should probably explore A, B, and C. And let's say I say, well, actually I like option B, but maybe with like an inflection of option C, it responds immediately.

There's like no hesitation. There's no like compute loading, analyzing like we see with ChatGPT. It just responds seamlessly. It's pretty wild. Is it relatable? Yeah, yeah. And well, the point of why I think this is super important is it is relatable. And I think it's a much more engaging form to interact with these agents, right? So, okay. So what crypto projects are even exploring voice-enabled agents? And the reason why I'm asking this question is, well, right now we have a bunch of text

chatbots, right? But if we want these agents to do more things, we need to be able to interact with them in different ways. And we also need them to be able to understand what we're asking for, right? If I'm saying, "Hey, could you trade this asset for this asset?"

it's probably going to question, hey, are you sure about that? Okay, this person is unsure about these things. Where if I'm like, hey, swap two ETH for blah, blah token, they're going to be like, okay, he seems pretty confident. Like, I'm just saying that there might be some kind of tweaks here, right? So what crypto projects are even exploring voice-enabled agents? There's a few that are probably worth mentioning that are still very early stage, but I think are going to be building out some cool stuff in the future. We've mentioned Solgraph, David, on the show before. It's one of...

I think it's Arc. Yeah, it's one of Arc's opening projects on their platform rig. But I want to focus on a feature of what they're building here, which is voice-enabled agents, right? A16z, the real one, not AI16z, wrote a really cool thread on why voice is important.

And in crypto, we're focused on text-based interactions. But what if it was voice-enabled? Blah, blah, blah. Solgraph is like kind of centering this as like one of their main features, which I think is super cool and worth paying attention to. Agent Tank, David, is another one. I think if you pull up this...

this tweet here, they mentioned like bringing it up basically in their next product launch, right? Where there'll be, I think they're building like a co-pilot or a co-companion. You'll be able to just kind of like seamlessly speak to it. And I think they're going to have like a demo ready in like the next week or so, which is like super wild and cool. Like we literally might go from like these kind of like tech spots immediately to something much more engaging and useful. The final project, which I was caught aware of, which I think someone told me about is this team called Limitus.

I don't know too much about this team. I've just seen what they've kind of produced in terms of like demos on Twitter. But they're, you know, in the early stages of building out this product. But, you know, they're enabling voice for this as well. And specifically in this demo that they're using, it's within Telegram chats. So, you know, you could be typing about a certain token on Telegram, then you could just speak to it and say, hey, could you just ape this token? Here's the contract address, right? And it can pull it. Lots of ways this can go wrong.

I'm not saying it's lots of ways that this can go wrong, but I'm saying we have to figure out those kinks if we want something super intuitive for humans to engage with, right? I bet you, David, you would trust an agent way more if you were just speaking to it and it sounded like a human than this like random like code screen that you're looking at on this tweet right now.

I think so. I think maybe more, but I'm also extremely wary. I, each as I have never been fished or hacked or exploited. You better knock on wood wherever you are. My, my track record is very good. Uh, and I'm,

And so part of it is because I'm risk averse to like using new technology like this. But nonetheless, I'm sure you saw the Zach XPT thread about how like Coinbase is losing $30 million, $300 million a year or something to like customer phishing attacks. And that's just Coinbase. Like phishing is a problem throughout the entire industry.

When I see stuff like this, I see a potential solution to the bugginess and bad UX of crypto by having AI agents make sure that we don't get phished. But at what cost? But first, some people are going to lose some money first.

Probably. Like it's going to pick the wrong contract address. I'm going to say, hey, chat, hey Siri, buy this one token. And then she's going to mishear me and buy a different token. And that token is going to be a honeypot. But like, it's a very hard challenge, but it solves, it's like two birds with one stone. We solve a very big problem this way. Yeah. I mean, building an open source is hard, right? And I'm guessing like some of the centralized companies will come up with something cooler much like way before, but we'll learn from that, right? And at the end of the day,

no pain, no gain. Yeah, that's right. That's right. Ejaz, this is our 10th ever AI Weekly Roll-Up. Really? We're 10 episodes in. It's been a pleasure, my man. How have you been? Reflections 10 weeks in. I feel 10 years older, David, 10 episodes later. My hair is graying. No, I mean, we, God, 10 episodes? Episode one, we were talking about

Goat. Truth terminal. Goat. Oh my God. What's happened with Goat? Goat really hasn't been on the scene much. Goat's taken like a huge hit price-wise. And that might be reflective of just, you know, the team not really shipping anything material since then. And it's kind of just been branded as a meme coin and fart coins kind of taken the spotlight. But yeah, not too well. Goat got farted on? Yeah. Goat.

is the summary um but yeah wow we've come such a long way and again 10 episodes ago was uh two and a half months ago november two and a half months ago literally which is just nuts to see how far we have had a bull market top and a bear market bomb yeah in in 10 weeks which is just very typical of crypto but super excited about where things are going

All right. Well, here is to 10 more weeks and then 10 more weeks after that. Bankless Nation, especially you people enjoying the AI agent meta. This is what we're focused on. Crypto is risky. Crypto AI is even more risky. You can lose what you put in, but we are headed west. This is the frontier. It's not for everyone, but we're glad you're with us on the bankless journey. Thanks a lot. See you guys.

you