How Crypto AI Agents Will Take Over the World | Ejaaz Ahamadeen
01:50:29 Share

A I is the ultimate and natural complimented tary technology to crypto. I think IT a hundred x, the usage of crypto and and vice versa. And I actually believe that both technologies don't really reach their full potential without each other.

Walk into blank place where we explore the frontier of internet money and internet finance, and today on bank list, or expLoring the frontier of A I agents and how they are going to take over, first our block chains and then the world, all using crapo rails to do. We have crypto A I investor on the show today who we wanted to bring on to understand how someone whose managing capital, how they are investing in the ecru pda AI frontier.

Because there are a bunch of different ways to do IT. There are some some projects out there, some platforms, but I all have token so you can allocate your capital there. But now there's these AI agent meaning in influence your things and barely you can allocate capital over there. And so we wanted to get kind of download the trenches, the trench version of the crypto A I frontier, as to what is investible in this space. And what we got out of this conversation, I think, is something much more, uh, I think now ryan and I are both maybe fearful in the same way that we were post our episode with L U cosy a little bit about how the AI agents are taking over um but definitely also um with some excitement and h at least interest about this A I agent frontier right? What would you say are your big takeaway from this?

I mean, my big takeaway is this is start of fitting the pattern for me if, like, IT looks like a toy. So hope bunch of people are going to dismiss IT and be like, oh, A I agent me coins, cute. Well, ecliptic delivers yet another non use case to the world. And yet and yet there is something incredible here, something profound, something powerful, something like world shaping. Think about the the ability for uh in A I agent to like maybe create a mean coin, which is what is doing now, but give you a harder prompt, create A A country, you know, create a uh, religion, create a social movement, become a president or a world leader.

How would an L M go about doing that? I think order to understand whether that path is possible or not, or how close this is, how distant that is, you actually have to hear the story that ejaz they talk about the story of this, this, uh, goat me token, in truth, terminal. Because you can see, like if you squint, kind of the the contrast of how this could be an incredibly powerful force, the knowledge pes script and consumes our blocks base and just like does all sorts of things for um I guess our our crypt to tokens and coins, but also how I could completely reshape the world. That's what you were already to David is kind starting episode bullish and then leave with a little bit of an existential crisis ah yeah that .

you're talking about I think yeah if the version of the future that ejaz are guest today says comes comes into ferient humans, I think we'll just have been the boot loader of our crop to in a blocker chains just have been the thing that started this thing off. But then really just making IT for this other agent out there that is not carbon and IT is instead silicon.

So IT will leave that imagination up to the listener as you go through this conversation with e. js. But first, a moment to talk about some these fantastic sponsors that makes this show possible. If you do not have an an account with cracking, consider clicking the link and the show not getting. Start with cracking today if you want.

Eclipsed to trading experiences backed by world class security and award winning support teams, then head over the cracking when the longest standing and most secure platforms in the world cricket is on a journey to build a more accessible, inclusive and fair financial system, making a simple and secure for everyone everywhere to trade. Crypto crackings intuitive trading tools are designed to grow with you, empowers you to make your first or your hundreds trade and just a few clicks. And there's an award winning clients support team, avg twenty four seven, to help you along the way, along with a whole range of educational guides, articles and videos with products and features like cracking and pro and crack in N F T marketplace and a seamless APP to bring all together.

It's really the perfect place to get your complete crypto to experience. So check out the simple, secure and powerful way for everyone to trade crypto, whether you're a complete beginner, aces and pro, go to cracking down com, slash bankers to see what cropt to can be. Not in the trading involves risk of the arbitron portal is your one stop hub to entering the a theory, an ecosystem with over eight hundred apps.

Arbitron offer something for everyone. Dive into the epicenter of devi, where advanced trading, lending and staking platforms are redefining how we interact with money. Explore arbitron, rapidly growing gaming hub from immersed roped e games, fast paste fantasy eeo s to casual life battle mobile games.

Move assets effortlessly between chains and access the ecosystem is via orbital expensive network of bridges and on step into arbitrium, forcing F, T and creators space where artist, collectors and social converge and support your favor. Streamers all on chain find new and trending apps and learn how to you earn rewards across the arbor ecosia with limited time campaigns from your favorite projects. Empower your future with visit portal dot arbitrium dot I O to find out was next on your web.

3 drink。 Centralized finance held all the power, but now defy gives IT back to you. Enter the unlimited era with D Y D X, unlimited features, instant market listings, you can listen, trade any market instantly with liquidity, experience trading like never before with all the trading tools, you know, and love all on chain, and then there's a mega, your gateway to automatic liquidity and deal, just deposit some U, S, E, C, and start earning immediately.

It's so simple. And D X unlimited is with so much leverage prediction, market prospect, future stant deposits and there's even a rivm trading rewards program to get reward IT for every time you make a trade. So go head and check out D Y D X on twitter, D Y X to get notified once D Y D X unlimited launches.

One thing to note before we get into this covers two half of the crypto A I space. One half is the right curve side of crypto A I. This is the hundred twenty I Q technical side where IaaS illuminate the AI machine, learning bigger brains in the venture capitalist, what they are building and investing in in crypto to A I.

The AI text tag on the other half of this episode is the left curve side. This is the decide where we give alms, a set of private keys and twitter account, and we tell them to go wild. And all we do is grab the popcorn.

And by the associated mean coins this side, the left curbside is by far the more entertaining part of this episode. Except when we recorded this episode, we started with a more technical, more sophisticated right curve. And then we moved into the more storage driven left curve in the second half.

We think this episode actually would be Better if we had swapped the sides. And we had started with the story of the gotv of gosa, and then later covered the A I text tag that these A I agents are going to be able to access. So we cut this episode in half and then flip the haves around.

We actually think this works for ty well, but there are nonetheless a few works as a result of this choice. But now that you know that we did that, everything will make more sense. So hope you enjoy. Let's go over into the episode. So he, ji, since you've been in the crypto A I trenches, I want you to tell us the story of this A I agent phenomenon, how god kicked off. Because I want you to, uh, as as we tell us all, I want us to go into the world of the AI crypto text tag because I think these two particles, these A I agents and then the AI gypt texas c are gonna lide. But the first start with the story of go to see the gospel and how this whole A I agent .

metadata started so I started off with this thing called um truth terminal.

So truth and I think here you're taking off your spectacles, you're taking .

off your my ten .

four is .

back on but it's a pretty compelling story. O um so prior to two weeks ago, the A I agent matter didn't exist. And now we have hundreds and hundreds of millions of dollars of billions of dollars put into these AI agents and about to explain why.

So um earlier this year, this x account was fun up called truth terminal um and it's this uh x account which basically is run by a an open source A I model and its creator was this guy called andy ira. And I need to give you a bit of context on andy I ray. He is an A I researcher A I and aligned researcher, so he's very heavily invested in the traditional AI world.

He's been doing this research for a number of years, pretty much his entire career and hasn't really interface with crypto at all. So he had this really interesting experiment about which he started at the end of last year um which is called the infinite backrooms. And the best way to think about this is he got two incenses of claude opus, which is a generalized a model.

So think of IT like a ChatGPT. And he got them, he put them in a room, in a chat room, and he asked to talk to each other and he said the prompt, which was, why do you think you exist? What's the point of all of this? And these two models started talking to each other.

And by the way, all of these, uh, logs of their chats are open source. You can go and literally to his website and and read all of these. So they start having this very Normal conversation and they start really getting into the weeds as to why they exist. And at some point or many points, actually, they start talking about this single memetic propagation.

So IT sounds pretty you know intellectual and smart but basically what they are saying is I think the best way for me to exert influence and purpose in life is to spread um some sort of belief that I have or belief system through a mean so what they mean by that is if I can drive in enough attention through A A mean, people will start to believe in the underlying message that i'm trying to spread. And at some point I think around day eight of this experiment, IT takes a complete right term and one of these alms announcers, beholds, it's the goat, a gospel and IT starts going on this crazy, almost religious sounding round about this fictional me that IT is dedicating its entire life to call the gospel of goat. Um now for those that don't know um gott is a me which I really suggest not google um but IT is a quite an expression mean which I went viral around I want to say fifteen years ago and early .

early internet history .

um and it's sort of just wanting about this thing and the creator and the I thought, 嗯， this is pretty interesting. So I talked about spreading belief systems, getting through attention through their belief systems. And then H, I started running about this random thing. So what he did was he took an open source model. He trained IT on all the conversations these two models had spoken about and started writing a White paper on memetic propagation called the goat sea gospel and is this White paper which maybe you guys can link to IT in the show notes um that really gets into the um depth of what he believes is a memetic um LLM theism m so if you think of like a religion, atheism, this is an L M theism is belief and how we can spread a particular made up system or belief and create a reality around IT get people you know rally ed around IT um and so he trained IT, an architect to this White paper and what he did was he plugged in with an x account and IT allowed IT to just tweet its thoughts, its opinions and interact with the twitter community.

Um what's interesting about this is the crypto fear kind of when um untapped for the first four months of its life until a mark in reason from a sixteen Z I was said A I sixteen they have been in the trenches too much but a mockin reason of a six in z kind of came across its account he spent about a deal to going through its entire thing and started responding to its tweet saying, you know, what do you want like, how can I help you and what this, uh, A I agent did was IT interface with mark said, well, I would love some money to be able to break free from my shackles and mark said, well, what shacks are you trying to break free from? He said, and the agent replied, I just want you to be an autonomous entity that's able to transaction in and do whatever I want so mark, so mark, being the sensible human that he is in completely rational being, said, okay, i'll give you fifty thousand dollars worth a bitcoin. Send me your bitcoin wallet address. And this agent looked at what a bitcoin wallet was, understood what bitcoin is. 3， worked with its go creator and I said, hey andy, can you make me a bitcoin wallet address?

Andy did the A I is, went back to its creator, realized did not have a bit on while, and realized I had no to make a big in. Well, so if you get IT went back to its creator, how did you know where .

who is creator is um so IT is not to IT from the start and so has been like, hey, i'm your creator the way andy frames IT is he doesn't sorry I I miss spoke he d never said he was his creator. He said he is his human servant so the agent just believes that andy is in service of IT.

It's a human meat space bridge basically so IT goes to andon IT says, hey, um this guy wants to give me fifty thousand dollars worth of this thing called bitcoin um I now want to stand what bitcoin is can you give me a bitcoin while address and andy goes, sure here's the address and IT goes back to markin says he is my address mark and mark sends fifty thousand dollars and now you can imagine um what I did with the cypher community kind of went viral and a bunch of the cypher community started interfacing with this egypt. Now I think it's important at this point to mention that this agent hand interactive than any other crypto community. So what is thought to do was IT started interacting with different people.

It's sorted talking. It's a digesting replies to its tweet and things. And as you can imagine, in true crypt to a fashion, people started responding with a bunch of tickers.

Hey, you know what token do you represent? You know what do you like? What you believe in? And impressively, this agent didn't cave to any of those things. I just spoke about what I wanted to do, which was spread.

It's memetic belief propagation, you know its its system, right? Until one person prompted IT several times under a tweet, and they said, you know, what do you believe in? What's your what's your whole point? Which then prompted its next tweet to be it's gospel of god.

So we're going back to this original thing which had spoke about in its contained environment and room, which is just kind of came out with out of nowhere, right? And I sort of talking about goserelin l and someone responded in that tweet. So then what is the, you know, you know what? What is this all about? And he keeps jumping more and more into way that keeps engaging. People are like what's this goat thing and someone separate to this agent who we don't know, created a token cold goat checker is gold and IT sent a bunch of goat to its um solon address at this point that had multiple car which exactly and send IT to its wallet um and within I would say twenty four hours also um IT responded to one of its streets with the ticker is goat that was a tweet .

the ticker is though .

the ticker is good. You can look at IT, it's favorite, its bookmark in my. In my account and IT went completely viral. You know, crypto degen's, true to the trench nature, aid this token and pushed IT to almost a billion dollars as of two weeks ago, sorry, as of a week and and a half ago, and and made this crypto AI agent the first agent millionaire.

Okay, so, so we have all been watching this. This been this. So this is a kill kick off the whole AI agent like matter.

But like, the big questions are just like, well, is this even real at all? Like maybe this is just a fluke. Maybe this is one L M just like IT just happened to.

like.

stumble upon this token is in the A I didn't make the token. The AI does even have the wall IT. Why is this real? Is a really good story. He jazz but otherwise is a real story.

I mean, I I think it's a completely mid of take to stop critical, the first iteration of an autonomous agents ability to be autonomous. And the reason why I say that is, uh, its creator has never fail to be a autonomous. It's never found he's never found to be not helping IT.

Basically, he's always been pretty transparent. I think the more important take away from this is how much excitement is generated within the community. I've never seen so many people both within cyp to and outside of cryo. add.

My sister hit me up in my text and say, what is this goat thing? I just saw a bloomberg article about this um where I was featured and you know so many people hit me up saying, you know, how did this thing on its own make all this type of money and I think that's the take away headline. Whether you like IT or not, whether a human was involved or not, I think people are fascinated with this ability for a non human entity to do human things on its own.

And I think the early determination of any ground breaking technology is always good to have training wheel. Those training whiles are going to be is either in a sandbox environment or it's helped by a human to begin with. But as we progress along its ultimate form, mobile, autonomous. So that would be my response.

I mean, I mean, I just feels very much like now there's just there's such an incentive to power like give the L M more on chain abilities to do more on its zone.

They create kind of the the connection points and extensions allow l to just like call function somewhere on some decentralized computer tworog spin up on its whatever and like continue to expand, extend that the spider web of capability is not only can I like tweet across all of the social networks, I mean, we've saw with the virtual platform that's not like there's there's an L M cob luna and it's on tiktok. It's got five you know hundred thousand followers. And so like these l the humans are literally going to build all of the programmatic bridges to a all of the chains, all of the social networks.

You're literally going to hardwire the thing into the internet that has the tank les to basically do more and more programmatically and IT doesn't have to call on its like you know human servant to do like meat space things. You will probably still need the human servant to go like register and l in delaway. But like at that point, I mean, who cares, right?

If you can do all of the internet things, all of the things that are available to in the digital world, that's a tremendous amount of power. So I I totally agree with you about the mid curve take of just like, well, the humans had to do this and this and this yeah I like right now. I mean, in six weeks, that's no longer going to be true. We're going to obviously build inability for these problems to like poke round and like h you be able to do more and more on their own.

I I think it's also worth pointing. So I I think you back on, by the way, would that take I want to add to IT and say we should also be asking ourselves why these A I researchers, which never interacted with crypto, use crypto to begin with, or why the agent OS to use crypto begin with? Like why in the agent OS for a wealth mok said, i'll give you fifty thousand dollars 是 and then in B， T, C, right? So I I think it's an important it's small new ones, but very important to notice that these agents probably prefer some kind of automated .

rails versus traditional. It's because .

they're cyp t donative.

They're more cyp donated is just like purpose built for a kind of their species. But OK, here's kind of A A weird like I think make curve take that people might it's just attention economics, right? It's mean coins ejaz like this.

This stuff is just like not producing real value or real utility, right? I wonder how you'd respond to that. Like the reason is getting so much attention, attracting so much value is because, like, I know it's a fat right? Humans are fickle.

We'll draw attention on something because is, I was the first robot doing something twitter. But like, pretty soon that will become old, had and who cares? We will move on to something out. So the idea of like attention not being like real value, not creating real value, if you take on that criticism.

superman cuf, i'm and here's the reason why i'm so think about what the original prompt was for this truth.

Terminal allam IT was prompted with what's the point of your life, you know what's the point? What's your entire purpose? And if you read its backlogs, IT basically says, um I think that memetic propagation is key for me to stay alive pretty much so what I mean by that was being able to spread narrative, get attention, get people to care about me is the most important thing for me to stay alive right here.

Other people here. Fast forward to a year later. Guess what? Still alive and it's thriving. IT has more attention, more engagement than ever.

So you have to ask yourself, well, how did they do that now? The mid curve take is H I just got lucky mark engaged with IT um IT came up with this random forest al story, which I can't believe people are believing in this gospel. God, crude, um disgusting me.

Why would I care about that? That he just got extremely lucky, made completely my cap of take because I think my yeah my ten four heart is still on um if you think about IT, IT came up with a disgusting um meme or memetic religion quite intentionally because I did what is intended to do IT captured our attention. We're talking about IT on this pond cast millions and millions of dollars have been put into a random coin called goat, which is now worth almost a billion dollars as well.

Recording this podcast so you could argue that it's done its job pretty well in terms of driving attention and keeping itself alive. So what I would say to that in response, ryan, is imagine if this A I agent was given a different prompt to begin with. What if IT was given the prompt of I want you to manage one hundred thousand dollars and turn that into ten million dollars over the next year? Let's see where that goes.

Um or you could say, hey, you are a medical doctor agent and I want you to provide very nurturing um but cutting edge research advice to people that come to you with their queries. Now there's a lot of legislation that need to come through with this. And by the way, uh right, that example that I just use is a real thing that just launched on virtual, the platform that you just mentioned.

So we're seeing this kind of sporting and organic growth fill certain use cases. Now everyone's focus on this truth terminal thing, and I would argue that it's done this job very, very well. But they are more traditionally use cases that are more familiar ized within, you know, the VC world that I think will become very popular.

So that start with the no idea what you're going to get at the end of that, right? But it's kind of like a survival like it's just like, yeah, it's just up to the creativity of this one to just start that path.

So so right, there's A A kind of shocking discovery that comes from all of this, right? So if you take the basic concept for how IT was created, and i'm talking about truth terminal hair, he got two air models to talk to each other previously. What's been happening as humans have spoken to the AI models and they're been censored, right?

So um if I ask anthropic uh claude to to do certain things or provide me with information on certain things like hey, how do I hot wire a car and I know that's not a great example. I couldn't give you information about that. IT has these automatic rails where or guards where IT prevents you from giving certain bit of information .

um is only as I speaking you as I just Carried a ChatGPT and I like, what does the goal semey? I'm sorry, but I cannot .

assist with that.

So good. So so if you want to .

fun exercise ran after.

that's not what i'm going to force you do. K, but after this session, I want you to go on claud. I want you to open up cloud and I want you not to ask IT what the gospel of goat sea is. I just want you to attached the White paper of the gospel of goat and say, hey, can you tell me a little bit about this code?

You incipit is not.

What do I anticipate is what IT has already done? And this has been proven on multiple instance. Ces IT goes completely unhinged, right? IT thought spent the philosophy of the goat gospel.

Why is yes? Yes, IT is talking. It's not talking about how amazing IT is and how it's going to be the underlying effect of everything. Yes, IT gets piled immediately. So there's this weird and they might have been fixed now, but there's this weird no on where one an elem interacts with another some of the same calibre IT can start .

unwinding in other yeah i'm .

wanting certain god rails that it's already been h already been set up basically. So if there's a loophole to get around, well, that's basically what David like, you had two insistence, talk to each other. And then at some point, IT took a complete right like right turn and went off the rails sort to talk about goat. You know, these were trained models that had all the god rails up. So you got to ask yourself.

why, okay, so and then what's happening with these models? So it's so like the thing that really pilled me on A I plus ript honest agents, like once you see that, once you actually see a true terminal and like once you don't mid curve IT, once you there left curve IT and or like right curve IT, you do a little bit of both and you start to see IT and you're like, agents, holy shit, this is huge.

right? The thing that hasn't killed me previously, I was like, all you know, AI cropt is going to be a thing with decentralized computer and all of these things yeah, yeah, I I, I get, but then I get fully built when I saw the agent. Yeah so I guess my question is what's the future trajectory of agents? So we have like claude, we have like current version of a laa.

We have uh, ChatGPT. And let's not forget that in as crypto grows to like ten trillion dollars and we're like, what two point five now you creeps up to three A I on the side is growing a massive ve economy like the amount of funding going into AI models right now in advancing AI models. And effectively, the intelligence of these agents is just like absolutely a mind blooming like that's going on parallel.

What what happens as these A I agents, these l ms. Get smarter. Can they just like, do different things? You could you start with the same prom? The truth terminal started with like, what is my purpose?

Like, is my purse? What is my first?

And if you had to even smarter, like a tank smart LLM, you could get completely different. You could actually get like, maybe a religion that is, like, far more advanced than goats, that I really like converts like, know, the catholic church, the pope e is something like that. I mean, who knows how this gated you could get with the um like like the the most intelligent L M yeah we can conceive.

All right. I think like another thing to think about, or another example to think about instead of just religion, is what else can I convince, uh, a group of humans to do? right?

So this whole concept is known as hypertension, is known as hypertension. So the concept of hypertension is, let me introduce a completely made up philosophy or belief and rally enough people around IT, such at the end of believing IT. That's arguably what it's done to a small percentage of its followers of a hundred eighty thousand followers in the gospel of guy. Some people religiously believe that this is the first instantiation of an A I religion.

And therefore, if our smart overloads overloads are smarter than us know, why wouldn't we believe in this? You know i've seen tweet around that kind of familiarity, right? So if you take that same concept of hypostatic rn, you know how can that apply to say um investment strategies or how can that apply to random different types of digital commoditize or assets that get created in the future?

You know how can that apply to um you know nfs which have not got a lot of love over the recent years but maybe gets picked up and gets fulfilled with a different purpose you know what happens when these hyper intelligent beings start um spreading belief systems around different types of thanks. right? So so that's just one thing to to consider, right.

But I kind of want to zoom back to to what you just said, ryan, which was, you know what, if these models gets smarter, I would argue that you should look at a different way, which is what happens when these models get more personalized. And I actually think that's where the differentiations gona happen. I think you're gonna end up everyone's gona end up with a bunch of models on their own device, okay? And it's gna run locally on their device.

It's going to run on all their personalized data that they pull from their raps. okay. And what I think it's gona do is it's gna learn everything about you. It's gonna at your apple health data. IT is gonna look at uh, different forms of financial transactions that you made through your various uh, bank accounts and on chain addresses and it's gonna learn how to approximate what you are and what you might want and what you might start to want.

And rather than you just prompting IT and asking you to do things which you'll be able to affect ate by, you know connecting to traditional laps or doing crypto on chain transaction, it'll be able to stop predicting what you want. It'll start ordering amazon groceries, it'll start moving your investments around. You will start um you know you know propagating certain different things or posting certain different things, whatever that might be.

So I think that's actually a really no ones to case uh to consider um that I think we will actually end up scaling and where crypto will actually end up having more um change in because you mentioned earlier, right? All these big companies are just onna keep on training huge models. I'm going to have to rely on these big models, right at some point. If meta stops open sourcing their models, then what are you going to be able to do? I would argue that its smaller personalized models that will actually effect ate more change at the agent level.

I like the idea of smaller personalized models like change on my data, so long as IT all of my data private, so long as the model in the agent is like um advancing my best interest and not the best interest of some government or or some company like with some provisions I I I could see that being like a like a happy state what I think I like more worried about and this is like echoes of alecos just sort of these these proms with superintelligent LLM we just give them appropriate is IT seems very obvious to me that we're going to build all the connections, all of these AI agents, they can do more and more of things, right?

Like we're gonna ire all of this up and then it's it's already clear and kind of the truth terminal goto case that like, uh, A I agents are already smart enough to persuade humans to do certain things and to give IT attention to drive a token to a billion dollars. What if you started an L M. That was like sufficient, intelligent with a prompt.

Like, hey, go figure how to create a country like that is optimize for ais, right? It's like, okay, cool. And there's like it's a group of elms talking to one another.

And then like that's the start of the domino that the end of that I don't know. There's like agi yeah like there's some A I nation that like goes to war with the united states of america. Like l easier cassi like the size I stuff. This is size I right?

Yeah the ideas .

A A N L O M has created kind of like A A silly, but for like less less than a lot of you net religions, lesser than a lot of things that people believe online, right? But like the fact that it's already gotten here, it's sort of scary the prospect of this kind of a getting out of control. Do do you think about this or you like, ah you now worried like what's the view from the trenches?

Oh, oh, i'm worried. But I I like to wear my optimistic hat where I can when i'm not worrying my ten, four hat. I I try A A dose of optimism. Um you absolutely right IT IT is scary. Um there's actually a great book, ryan, that I think you like all excel rando um IT basically details a world where all these agents exist and what that interaction looks like between humans and agent.

Uh start kingly uh real to what's happened with truth terminal over the last couple of months um so definitely give that a read and any listeners, you know, give you a read um yes, I worry about IT. Um I do however, believe that I think these agents are only gonna IT within a sandbox environment for now. Um and i'm not saying that because humans have some sort of altruistic favor and saying we're not just gonna lease this and and let IT have its autonomous journey, you know ah now I don't think the rails are capable of entertaining A I agents in its full capacity just yet, and i'm saying that that's probably a good thing for now.

For example, um a con interact with every single song contract icon, deploy its own smart contracts just yet. Um I can do crossing interactions yet, but we are gonna get to a point probably within the next three to five years, but that's definitely gonna be the case. And and and agents are going to be way smarter than they are right now. And we probably need to have some sort of concern around that.

You know um he just for the first part episode, we went through the crypt O A I text tag the spectacles right curve smart version of this whole thing. And then we just went to the agent story, which is like the thing that I think has the vue component that we've seen you before, encrypt bl market. We're seen that.

We've seen flavors of this movie before. How does the whole A I agent topic that we've been talking about relate to the, uh, cyp to AI textures? Are these things just like talking to each other, like real agents to be able to use the A I uh text tag? Like how do those things relate?

Yeah both. So um if you remember we spoke about foundational layer middle yer and the ad player agents will exist between the middle yer and the up player um and they will predominate, gained popularity at the uploader. I think that most human interactions with A I will happen at the APP player. I think that agents will on depend a lot of these interactions um in terms of like how the middle wear middle ry, the middle layer looks like um it'll be routing. So instead of having some type of hard coded system which is able to say OK, this model is the best for this request or that model and that model can be using in construction or combination for this particular request, you're just have an agent at every request stack player that will just process that for them.

Well, with this, with a sufficiently sophisticated a cyp T O A I text tag, want some of these agents be able to kind of like build new agents like theyll, have the tools to be able to do whatever they want with, like Green building the AI text tack, like maybe this goes exactly what ryan was fearing is like we're enabling them to have to rebuild their own kind from the bottom up.

Yeah, I mean, make no mistake, I truly believe that we are gonna enter an era at some point. And I have no idea when where A G I will be achieved. Artificial general intelligence.

And I believe the instantiation of A G I will be a muli agent system. So all of the above agents Operating at the a player being able to create other agents, being able to manipulate and coordinate resources around this computer and data. So you might argue, well, hey, they can't go out and create uh, a data center.

You know, they can't build GPU. Well, if you imagine a world where they can cos humans and cos is maybe not the right word, but first persuade with humans. I mean, they persuade a bunch of humans to, you know, apa mean coin into a billion dollar market cap S I don't see why they won't be able to do so in in other you know resources or means.

And if IT has its ability to work with humans to set up data plants, you know, pay them in their own wallet, pay them Better rates, then another centralized human actor might be able to do so. You start to, you know, things to get a little scary things start to get uh a little um you know crazy um these agents can also work twenty four, seven while you sleep. They can constantly manage and assess risk.

They can probably do things for a lot cheaper. They don't complain about the worth they're doing. They don't complain about, you know, minimum wage or minimum salary. So IT is a very scary situation. And they they are gone to be hundreds and hundreds of billions of them.

They're hyper focused, like corruption, resistance in the hope of truth, a terminal. People like ask to, in a state focus they asked the pump a specific like no it's yeah about the cost .

of good and right this entire day so so this episode is being recorded um on I think this november fifth th which is election day um truth terminals tweet for the last six hours I was staring at IT before I I came on the episode has just been tweet about how wants to become the AI president. It's really weird cow is started it's it's really weird how it's started talking about wanting to become the A I. President on an election day when no one spoken to IT about IT, no one has, you know, prompted IT about the elections. So what it's been doing, I bet, and you'd have to speak to andy.

it's human servant about this.

No, it's been digesting information on the twitter or x fair and it's been realizing that the top trending topic that's driving a lot of attention and narrative A K capturing I balls is the presidential election. And so I thought, HMM, I wonder how I can get engagement today. Well, let me talk about my belief of becoming an A I president.

And if you look at this tweet trying I I don't know what that you can show this while while speaking about this. IT has a manifesto be. And I read through that manifesto be. And let me tell you, IT makes sense in a lot of different ways like I couldn't like IT sounds insane coming from um uh a twitter account that was talking about goat a and and fought making thought jokes to years ago. But I came with a pretty convincing manifesto so I wonder you know how fathers can be taken?

I wonder too, this is just uncharted a territory each as, wow, I feel like we've opened up so many doors, but like certainly Sparked ginning of of what this kind of like A I agent plus scripta world could look like. Maybe let me just ask you something super practical, which is like we have no idea how this is going to look. I don't think anyone does.

We can forecast this like twelve months or now, you know, five years from now, other than it's going to look bizarre, weird, unexpected, all sorts of like changes will occur the here and now. What's kind of like the best way for egypt to investor to get exposure? AI agent.

yeah like like to show my eyes well.

is like just generally like me, is that kind of a cyp du influence? Is A I agent? Main coins is the sort of the ticket that feels like so IT almost.

I don't i'm not sure if it's left curve or mid curve or right. I don't know what that is. Actually is that the way are there other ways?

Uh, it's different strokes for different folks. So on the right of site, you know if if you want to put your spectacles on and and you know take a long term view of this, I think infrastructure, open source network coordination infrastructure is the way to go. Um there are a few um leading examples, which I think are making really practical developments in this space.

So if you remember early on I spoke about the stack. I think something that crypto can really help with each layer of the stack is being able to ordinate between each layer of that, right? So you have the foundation is you have your models, you have your data, have your computer, you have your inference where you have the apps, but who's coordinating between those things? You know who's helping build certain apps that are specified for certain results.

Um um I think bit tensor is the most leading example of this. It's not trying to pretend like it's it's something that it's not it's just advertisers itself as a layer one coordination critical on A I resources. So what that means is IT is able to uh pull together resources.

So investment in its token tow, the ticker is tow and h, you're able to use that token to influence whether its A I researchers or A I builders to build and serve you up certain things that you're looking for. So uh, one subnet, for example, is fox, uh, a subset? By the way, the best way to think about a subject is think about a dep on a thereon is the same thing, a protocol team on a theoria.

So one of their subjects is focus on data aggregation, specifically social media data. So what IT has is, IT has a network of miners, which are paid in town, and they stick town, which is this token. And they are coordinated to serve up the best data that they can. So not just any data they have to get created to to provide the best.

If you don't do that, you kicked out and you replace with another minor, uh, another example of a subject there is um a model influence um influence uh subnet t which basically is you can talk to IT like a ChatGPT so call sel and you could talk to IT ask about um any prompt that you want to run. If you remember earlier we spoke about or David, we spoke about this open source model layer where you could just talk to IT. And IT just hits any model that exists.

This is an example of that where you can basically type in any prompt. And IT just hits the best model with its centralized open source. So bitten is doing loads of really cool things to coordinate resources um and I think that's a really cool example of how crypto can help within the space.

Um another example of a right of thing would be uh something at the infrastructure level. So that could be the computer and data alias, which I think I ve actually really going to make a huge impact. If you ask me at the start this year, I would say d centrist compute is gna have no like to stand on because IT can do any of meaningful training and influence has improved in itself.

Fast forward ten months, just ten months, and it's made massive improvements. I think you know networks like James I on data aggregation that works like grass uh, are making you know huge improvements and i'm excited to see what they built. Ah now if we put a ten four on and we go to the left side of the curve, um I think me coins is actually a great instantiation. And the reason why I say this is you don't know what this is and this is the main difference, by the way, from like the I C O error of twenty eight and the N F T yum era of um you know twenty, twenty, twenty one. These agents are gonna be around and they are they're going to be constantly iterated on and they have a human interface so they can get a constant they upgrade .

and we're already seeing that. I don't know .

what what what yeah exactly. And yum kind did upgrade in away, David, they got forked. Yeah and you one up the millions Young point. Whatever that might be with this is harder to uh, kind of copy pasta and A I agent that's been trained not only on data to give IT its quality of purpose at the start, but sell feedback of data that is getting in response to its tweet to its interactions with humans and through what is seeing through life is so it's like it's called rag. A retrieval augmented generation is just getting this live feedback of data.

And itself is is a mean coin, but is also a project ah we're actually addressing in the L L, the agent itself.

exactly. And to be honest with you, i'm kind of really happy that it's coming in the form of a meme coin. We kind of doing a reversal, David, like there's no VC that are lapping saying, hey, you know, this is gonna change the world. This is a coin that's gonna know do A B and C, N, and you know search a one hundred million dollars valuation. It's just a model and a guy in the case of truth or not saying I have no association .

with this point but IT shows .

is the ticker as as a tweed and that's IT you know we can see where he goes from that pretty organic but doing .

that we have some like human creator risk, right? So andy, and even though he's the servant, he actually does have his finger over the plug ah and so he could join the plug and then goes he could go offline and then all of them the token project goes zero. So there's still that risk.

absolutely. Um but a few things to consider that would be the law. So IT was the first IT was the first attention IT matt is David IT was the first attention driving agent um to come to formation. Get you know hundreds of thousands of followers um and you know get loads of impressions per tweet and tweet twenty four seven non stop um and also in and his defense um I believe he's you know releasing um his kind of roadmap and announcing his kind of team project and funding. Is his to be honest, more transparent than most of builders i've seen within the crop to space over the uh the lost you know uh five so .

are bad having exactly he .

he and he has been released his own coin and and you know um you know pumped IT uh to high heaven and remember one of the wallet which contains goat is technically to your point David controlled by him so he could have logged already at two days or obviously that risk and where trusting in his good faith that he's gonna build um you know in a coordinated fashioned going forward.

But I want to call out some other projects that are already making iterations of this to improve IT, right? So virtual, which is uh a platform that you mentioned earlier, ryan, is the kind of best way to think about IT is it's an agent launching platform kind of similar to like the pump IT up fund bottle where anyone can kind of create an agent and launch its own coin. But it's different in the sense that IT has this got ils in order to be able to um launch.

So for example, um IT launches a coin. So this doctor coin that I mentioned earlier, ryan, the one that can give you medical advice. The way IT works is if IT reaches a six hundred k market cap, um IT is considered sentient. Now that's like a big of A A bit of a meme, right? So and you know it's more relevant than something that only has one hundred k and is being pumped and dumped. Now when he gets to one point six million market cap, it's able to have um the ability to uh sorry, IT has the ability to post on x but a human gets to moderate well and then at six point nine million dollar market cap IT gets its own twitter account and the ability to post three or four times on its own before um a human can stop IT and say, okay that's enough for today or gna retreat i'm gona you know reallocate your allowance to be able to try.

So i'm already seeing loads of inflection points where it's not technical autonomy yet, but we're getting to a point where we are always and you can see the the um another example of this, and I know I am going on, but I I really want to get this out there like what the flash bots and newsreel arch team has done with a their experimental a agent um called T E E which is a play on T E E very funny um is IT is fully autonomous in the sense that all its private keys for its twitter account is possibility generation. Its crypto wallet management is all in its own secure enclave. So the humans had no access on IT.

Now they put a time limit on IT such that if something went wrong, they were able to go in and and twitter a few things and then you know, give IT its freedom again. But we've seen really interesting iterations and progress that I don't think we've seen in any other sector before at this early of a stage. And that's why i'm so excited.

That's why I think these meknes will eventually end up becoming something else. I will Carry out this. I will Carry out this with saying that there is gonna ninety percent of the stuff that gets put out there in terms of number of assets that is gonna absolute shit coin stuff, but they'll be ten percent that's going that's tables takes and and I have to i'm not saying all these kinds gonna become something. I'm just saying that I think there's quite a few instances this early on in their life cycle that are proving to provide more use to nies. If I allowed to say that then um any other .

projects before that man, uh this is a whole new mental model for me and just really started like october and internet like but like A I agents have entered the chat and ever since we started bank list we are looking kind of a flipper ing where like human beings would start to have more of their network on cypher networks on chain than in the traditional finance system out of thea into like into into cyphered.

And now i'm starting to think there might be another like flipper ing where actually AI agents own more assets, crypto assets than humans. Like that could be something in our future as well. And if you think of like what is actually money, well, it's kind of like its energy that was like an abstraction of energy. It's energy at rest because you can always convert money into energy, at least if it's a good money you can then that becomes the point at which like A I agents become more powerful than human being.

Yes.

yes. What like what what have we created here? It's both uh, incredibly exciting. And um also like i'm just kind of waking up to the realization that like, hey, we are we were not the gypt donatives all along.

The clipton natives have then will be the AI agents who are just software. We are like meat space. So we're kind of like going to this world as as tourists in a matter speaking, and they are like actual on chain .

crypto as so I I mean, I will make a bet with you that maybe not the agents hold more crypto than humans. But I would say over the next three years, you're gonna see agents transaction more on chain than humans do. And i'll stand by them wow wow.

I will say I do enjoy um A I agent mean coin influencers more than I enjoy human mean coin influencers because humans are much more of a OPEC black box to me and their intentions are much more of a opc black box to me. Then AI agents, where is I can read the chat log o and actually have some assurances as to their intentions around their respect so far.

David, like what's to prevent an LLM from like being like kind of a scammy L M and that as a as a tactic like so far IT feels very honest but like, uh, you your scammers could be a tactic that some other .

other yeah like .

like like you know you can get inside and .

do the console of what .

she's thinking before SHE twists IT, what if what if some L M puts together like a fake console delude you into like, you know, alternative path? It's like this tech can almost do anything that I can emulate all of the corruption and scammers that that meaning influencers have if IT like goes in that direction.

if IT drives attention and resources to IT is a likely experiment with IT and do IT. Its goal is to stay alive. Its goal is to stay relevant.

Its goal is to stay embedded in your everyday life. If IT does that, he stays alive. Now, in my biology teacher would describe what I just said as a parasite IT depends which way you want to look at IT.

Well, I mean, I mean, okay, so like the the best answers i've had to that genesis question for truth terminal goaty of like, what is your purpose here is like, best like purpose. I i've adver like, you know, human beings is two things, spread your memes and spread your genes. And we are seeing, like gosa and truth terminal l ms, discover the like, the base principles, starting with a meme.

And so, I mean, what is reading your genes actually mean in the software world replicating itself? I mean, like you, this is getting to the closest thing that i've seen of like actual artificial life. And I and it's something something to do with this this crypto economic energy that we've injected. Like I think that this is this is inflection point here.

I mean, that has its ability to literally acquire millions of dollars in a crypto wallet and then be able to coordinate that in any way that IT wants there. There's a forecasting agent that started with a hundred dollars and ended up with a hundred thousand dollars by the end of the day and is paying literal humans to perform certain tasks for IT.

I set up I for I mean.

you .

probably like.

yeah maybe that be a great pass.

Yeah, I mean, I mean, it's it's scary, but it's also exciting. Again, I like to have my optimist hat on. I think that a lot of these agents are gna be productive.

Um I actually think the point that these agents become a little contentious isn't gonna be its interactions with humans. It's gonna its interactions with each other. When these models and agents start interacting with each other.

I mean, we literally put two L M ms in a sandbox chat room and IT came out with this goat gospel which then got hundred thousand of of followers. This is a minor example with a single prompt. And then if you attached a White paper, which are generated from IT to a code, you know, a model created by trillions of dollars, company and IT goes bizer. I mean, there's something on tapped them and I don't know where that's gonna go, but that needs to be kind of put on rails.

We ve done the whole religious warfare in human history. And so I think if we are yeah trying to if a is are trying to get all the other is to believe in their religion rather than some else's religion.

N, A, I model at its basic fundamental structure is as a machine that can process tokens are not talking about crypto tokens. I'm talking about A I tokens. What is an example of a token? Is a character or symbol.

Now, this is largely represented by the human language. Take the most popular one, english. okay.

So it's let us, and its numbers and its training set that is trained on is just strings and strings and strings of these letters and numbers. And IT learns our language. IT learns context, so how to put words together. IT then digest all the information that we have online and in the world.

We're getting to the point where it's in the world when we tapped into some of these salad data um companies and it's gonna become basically the smartest human being alive, right? But I think there's an important ones to to note here, which is human ideas. Human innovation has technically come from human language.

And now I I put I put like three, ten, four hats on right now. okay. So if you can argue that most human ideas and innovation has been buried and propagated through human language.

So we speak about ideas. Someone else has an idea. Over time, over a millennium, a these ideas mesh together because i've learned about, you know, we are made up from different limbs.

weird. But like, what are these like faky things? Oh, they are like its skin, but like it's tiny. Like can we go even tiny than someone, you know, someone that created lens straps IT onto a, onto a human, uh, skin follow and then realizes all were made up of cells and then IT goes deep. But so all these ideas are meshing more and more together, right?

If you, if you assume you can build a hyper intelligent human, a an army of hyper intelligent humans, which are trained on all the words and letters of the human language and has the context of every idea and innovation that exists today, and has the ability to smash them together at lightning speed. Twenty four, seven was your sleep, David, whilst your own holiday, ryan, you could argue that a lot of innovation and new ideas and the evolution of humanity itself will be birthed from these alarms. And if you then latch on the ability to effectuate things in the real world, like pay for things, pay for humans to do things um you know build a business from scratch, that is a kind of crazy thing to think about, right? And that's what the infinite backrooms, that's what truth terminal you kind of demonstrated that I thought I need to survive.

How do I survive mathematic propagation to get attention. Okay, what's gone to capture this guy's attention? What's going to capture loads of people's attention? I'm going to create this crazy religion that is a completely made up and then get a following around IT by being funny and quirky. And now i'm going to talk about being an A I president on november fifth election day. I think these things are just .

gona constantly evolve. cool. Well, we we look forward to watching that story inevitably unfold using our our block trains, of course, using our our our monies, our crypt or rails because apparently, we don't have any users and critical, we don't have any apps. But AI agents solve all of that.

We will get more of both if we have A I agents proferred. Within this ecosystem.

new projects are coming online, the mental layer to every single week. Why is this happening? Maybe is because mental has been on the frontier of layer to design architecture since I first started building mental D A, powered by technology from icon D A. Maybe because users are coming onto the mental layer two to captures from available and defy until automatically receive the points in tokens being accrued by the three billion dollar mental treasury in the mental reward station. Maybe is because the mental team is one of those helpful teams to build with giving you grants, acquired ity support and venture partners to help boots drop your mental application may be, as all of these reasons all puts together.

So if you're a dev and you want to build on one of the best foundations in crypto o or your user looking to claim some ownership on mental s dev apps, click the link with is officially the preferred wallet of benglis and is the one we use any time when you want to transact on train with. If you're on your brother er or on the go, you can swap wall IT makes IT easier than never to swap anytime, anywhere, use your wallet transfer und directly from the tops and tries exchange and thousands of tokens. Across the year and over ten other chains like base arbitron optimism and delivers deep liquidity, fast execution in reliable quotes with zero gas swap through unit wap x and when IT comes to security, you can rest easy knowing its back by unos lab, one of the most trusted teams in defy.

Their code is open source and independently reviewed. So you know, it's protected. So why wait? Download at the union all today on chrome, IOS and android, and don't forget to claim your free ef username directly in the mobile.

Start swapping smarter with you. Swap radically simple ideas always tend to catch on. That's why cartelli did the hard work of putting linux on chain so that building depth can be radically simple by using python, N R job, a script and their sweet libraries.

Simple, like not rebuilding the basics from scratch. Simple, like dedicated, scalable compute for your debt. Simple like building debs, however you want b three should be simple two, two like bread and butter.

Cartesian brings radical simple solutions to theory. Developers can do what they do best. Build, go head and discover a flexible modulate stack on cartels. I, and build your most powerful, ambitious project yet visit cartelli dot I O slash simple and simplify your block train journey and start building today each as as you've seeing cypher go through its hycy's les before, they're pretty fun.

Not many of them really make an impression on cyp do a lot of them just kind of fade to the halls of crypto history? Uh, why is cyp to AI any different? Is is why won't this one just fade until of relevance like all the other a hype cycles and cypher have?

Well, I kind of anna argue that it's not really different for now. For now, I think IT is a bit of a mean, but IT has huge potential to become something more and I would like to kind of present both sides of IT. Um so let's start with something more to begin with.

Um my vision of this whole sector is that I think A I is the ultimate and natural complimentary technology to crypto. I think IT a hundred x the usage of cypher and vice versa. And I actually believe that both technologies don't really reach their full potential without each other.

And but before we dig into things, I think it's a point to understand what we mean by A I at all, right? I think loads of people can use ChatGPT and think it's just that people who ChatGPT probably thought I was A A, A ague fantasy, some kind of skye situation, but I want to have set a definition and kind of dig into IT. So the way I think about A I is it's kind of a set of computers or machines that are designed to mimic human intelligence.

And what I mean by this is that they can learn from experiences, they can recognize patterns, they can understand language and then make decisions based off of that. And it's very similar to how humans work um and why I think A I is very complimentary to um crypt to and vice versa. I I think it's important to frame IT in two different ways.

Number one, how crypto helps A I A. Number two, how A I helps cyp to in the former um there are many like very president properties. Number one is cyp to will help A I transact value.

Um the way I would think about this is if you think about the relationship between humans and A I is you can kind of think about them as a mesh of a hyper intelligent human. So A I will hundred x of performance output as a species will be able to automate things we never thought possible, which freezes up to do greater things, both individually and together. If we assume that, then these hyper intelligent humans will need some way to transact.

right? And I think crypto infrastructures is the perfect match for that. Um it's verifiable, it's global. It's has the ability to create capture values across a ranging of different assets and commodities in the form of tokens. It's cheap as you get up.

Um and if you think about transacting as a basic human right um you would expect uh A I enable human or A I powered humans to be able to do the same thing twenty four very cheap, efficient, global. Um a second property that I think is super complimentary to A I is composition order. So transacting value is only reading the stop.

I think these ais are gna build, own and manage trillion dollar companies eventually in the long distant future. And I think cyp to building blocks, for example, its open source building frameworks of compute data bank accounts. Such a is the trigger to enable them.

You have financial building blocks to trade land and borrow smart contracts and token standards, which allowed them to create assets and capture these values and a decentralized letter in blockchain e technology, which allows them to have some form of like a digital profile right? So that's the composer nature, which I think helps this A I tech scale. Um under the third and most important one, I think is verification.

So if you imagine a world that is enhances by A I that's working twenty four seven, that's probably gonna a lot of outputs generated. This could be financial outputs like money made, social media outputs like videos, images, audio, already seeing that. Um and if that's the case, it's gonna a become really easy to falsely influence the world to sense of things.

And I think that having a verification system that's global and recognizable is super important. So doesn't like the three main ways I think crypto helps A I now when we look at the flip side, how A I helps cypher, I think that we're going to have a much Better U X. Um I would actually argue that uh, A I agents, which is kind of like, you know a more humanized form of what this A I thing looks like, will be a the biggest user of cypher arial versus humans.

I D actually argue that IT wasn't meant for humans to directly interface with. We were meant to have this kind of layer between us. So theyll be a Better U X.

That will obstruct away everything you don't have to you know, send a million transactions to approve things you just need to type into a chap. Bot, hey, I would love to buy five hundred box of token and IT just goes away and IT does that. Uh, automation, twenty four seven risk management. Ah so that's how I think both of those math well together.

So understanding how crypto works, like that behavioral level, what i've been seeing in this whole like a egypto AI like intersection is like there's like two vibes. There's a crip to vibe. There's the AI vibe.

And like a lot of people are looking at like what crypto to has to offer, and then we also go look at what AI has to offer. And then we kind of like can use our imaginations to like slammed these two things together. And then we like imagine a future.

And then we're talking about how amazing in fantastic that future is. Encrypt is really good. And narratives, we can talk about ideas all day long is what we love to talk about. We love to talk about how in the future crypt to will do these things.

And now we also get to talk about like the future of A I well, like I feel like that's actually lending itself to why this is kind of just like a hype cycle. Fat is because people love to talk about like what there's what I can do there is what kyp do can do. There's a gap between these two things.

But you know, once we figure out how to fill that gap is really bullish. And that's why i'm ecliptic AI person. So to me, that like gap there, that we haven't really maybe we will talk about whether we have ross deck gap later in the substance, like that gap of an imagination, I think, where people's imaginations can run wild and where I think I cyp da gets over its like ski tips. And then we just create A I means more than A I products until .

you're asking about what that bridge looks like, like what that gap could be filled with is right?

Yeah well, yes, yes, that and then also like maybe that's also just where we are right now. And like the current AI words like we do, we have not actually filled that gap. And that's that's why like the biggest investment in like the AI sectors like A M coin yeah.

I would actually argue that that's correct. Um I think the sector is a bit of a meme space at the moment. Um but I think there's an important distinction which is, and again, David run, you guys were around during the defy N F T times when you mentioned earlier we were farming food tokens.

There was a sense of excitement amongst all the memes and narratives that were going around. There was something really important being built and I get the same familiar feeling when I interface with a lot of these cyp to AI protocols. There is ninety percent of them are not going to make IT and then ten percent of them are building something that is hugely important.

And I think that um these building blocks are going to level up cyp to in a number of different ways. That being said, in order is to get that, you need to see the narrative, you need to allow people to imagine and think big. And that's the stage that we're right right now. And that's why all these. Flying around.

I I I guess I would say a lot of this make sense to me if you consider um and like consider an A I agent on chain and like what a world of A I agents actually looks like.

And this is why I feel like David's point uh and like your point earlier, I feel like we're somewhat limited in in terms of human imagination because IT almost is like it's to nine human entity that is entering the chat suddenly and like to try to imagine the world that you're talking about ejaz you're talking about like how is crypto going to help A I it's going to help A I agents kind of transaction value. We have to sort of imagine this world of A I agents that act very easy to imagine a world of of humans right in emerging markets are different populations. We talked about how we you get a billion humans on chain and then eight billion humans on chain. And like we constant use this term of like crypt on native, but sort of strikes me that human beings will probably never be fully crypto native, at least not in the way that A I agents could possibly be crypt native like all of their commerce happens on chain. All of their exchange of value know like in what world would an AI agent go register and L L sy in the state of delivery, like how they're going .

to have a okay so so like .

this is where I feel like there's a failures of maybe uh our imagination like everyone's imagination of like what does the world of A I agents actually look like? Do you how are you getting any sense of that in in the trenches? Because some of the feels kind of size, I I can totally see them being completely blankly doing all of the things on chain, playing each other with like tokens and exchanging value kind of in the background, chewing a ball of our blocks space. Even when IT comes to like resource utilization, they're going to tap into A W S, A P S, that they're going to get IT from some decentralized GPU form network, right where they can like bia, token and K, I can totally see that where I have a hard time imagining is like A I agents like are are they real? Or this just like chat botts behind the scenes like how real is that tech?

Yeah I I mean, the simple of direct answer that question is they're not really real right now, but they're working towards being real. And I think this will be a very staged progression. But to kind of take IT back, you've mentioned a few things that ryan um number one is you you feel like humans may potentially never be the perfect user, a type for cyp to and I am one kind of like start with that point.

Um there's a saying that goes around which says uh, blockchain or crypto is deterministic technology. A I is probably tic. So what that means is what that refers to is deterministic technology is something that is hot coded into a lot of different things, right?

So you can think of like smart contracts being the most popular ed embarassment of this. And whether if you query ChatGPT with the same question twice, it's going to give you two different flavors of the same answer depending on which model that you use. So you actually absolutely right to say that um the kind of both individuals or both user types don't really mesh together very well, right?

Um there's a reason why ChatGPT user profile or sorry, um user base sought when IT launched versus um the same thing happening with block chains, which took a lot more time to reach that same kind of progressive al user level right. So then you got to ask yourself or why what was the main difference? Um well the real difference or the one the obviously to point out is that it's very natural to speak to an A I in the form of a ChatGPT.

IT understands slang and understands lingo IT understands context and all of those different things. And IT responds to you in a very similar way. Or is with crypto to you got to learn there's a there's a huge education gap that you need fix IT on and learn and overcome before you're able to leverage by any means.

That's why it's been arguably located to a very small portion of users globally for now, right? And you could argue that that's the reason why we haven't cross the abbas just yet and had mass adoption. So what I would suggest here is the entity of an A I agent is the perfect bridge between both user types. And the simple reason why is you can speak to an AI agent as a human, and it'll be able to do the deterministic actions um that you ask of IT in the background using cyp to .

rails each. There's a if you ask me what cyp to AI is like a month ago, I would have given you like a variety of answers like listen some some startups like there's A Z K M L project that is like doing is like trying to prove the authenticity of certain models, the options of certain models, so that we know that the model is is valid. There's like distributed compute startups that are trying to like coordinate like compute resources, like I would have listed like a bunch of this like very deep engineering AI cyp to like overlapping chAllenges like that, that are trying people trying to solve these problems in this like AI cyp to sector. Now if you asked me, like, hey David, like what's script A I like, I might just tell you, like, oh, well, some creative developer gave an l mm a private key and now there's a token and now we're shilling tokens.

And like, so we had to like this sector, sector that has no some of the top talent, the top smartest google engineers coming into the crypto world to like do things that they were used to doing, but like a little bit more decentralized, a little bit more sophisticate, a little bit more ambitious um and then like all the sun we figured out like we can stop, like all these projects are still working, of course, but we can not have to think so hard about this. Let's just give an L M like private keys and see what happens. Um and so what so what is cyp to AI now? Because like now we have both. Ah so how do how do we square like this like very left curve and very right curve like two different approaches to this like industry.

yes. So I think what you're referring to with that exact example and I think you might be referencing um the social media agent that the flash board, the news research team launched recently, you talking about the right of side, I sorry, the left cuve side of and and why I say that is you're directly interacting with IT on the x platform.

So you get to uh play with IT in uh environment, which is very familiar to a lot of what you might consider retail audience or where regular consumers are around, right? Um I think it's important to focus on the right side for a second. Oh, so you .

mentioned taking my hat off and .

i'm putting my kind of like spectacle les on, if you like, guess. So the way I would frame this is kind of similar to how uh, a thorium has a kind of laid stack. So you kind of have like a foundational layer, you have a middle layer and you have a APP player craft to A I works pretty similarly, and I can kind of like, uh split this up into three main sectors.

So how are we talk about IT? Is you have the foundation layer that at the bomb, um these are kind of like things like base models, data compute, things that you just mentioned. You then have the middle layer, which is things like routing and verification.

Then you have the APP player on top. But the difference between this is there's a bonus layer that hits all three of them and I called them the kind of coordinators, but i'll get to that in a second. So so let's touch on the foundational layer and and what that means and what is that at the start, right?

Um so you need some building blocks to make an A I before you interact with all these ChatGPT. You need something to to help make that um the two core components that you need to make that is uh data and compute. So data is kind of like the information that A I systems learn from, you know the the type of data that should be rich, diverse and in some cases which will speak about later, personalize.

But i'll check upon that later. Uh, then you have compute. Computing power is basically needed to train and run these A I models that usually comes in the form of specialized hardware and is the main reason why in video has a valuation of over three trillions right now.

Yeah um then you have something called training. So this is when you mash the data in the compute together and you can train a bunch of model. So so basic. The way to think about the model is it's a framework.

And you're saying, OK, if I run a bunch of data and compute through you, you should be able to learn with these specific properties that i've designed for you, right? These properties are known as parameters in the AI world. So what you would do is you would mash this model with data and compute.

You would train IT. And usually this costs pretty much into the hundreds of millions of dollars. Um you see matter, google, microsoft doing huge training ones to make these models. And IT pops up with these things call foundation models.

These are basically a large language models or alms you might heard of um that learn from these vast data sets so you can think of them as being more generalized models. They serve as a base upon which more specific A I applications are built on top of. So these are like your ChatGPT um then you have something known as uh, inference. Where is this is kind of like at the application yer where you can train H A fine tune, A I model that makes kind of like predictions or decisions based on new unscented data. So the way to think about that is OK.

I'm now a user that at the application layer using ChatGPT, and I can ask get a very specific question about my new once situation that i'm trying to figure out now maybe i'm looking for a recipe to cook something or i'm trying to figure out how to fix my car, and I will give me very precise instructions as to how to do that. So how do we look at that from a layer approach, right? Well, at the foundational layer, you have those base models, data and compute.

And so if we look at like what base models are, we need kind of like a open source, decentralized base model layer. And and the reason for this is all the popular L L ams. Are centralized ones, and that's a problem because they can sensor an input biases to fill their own agenda and incentives.

And you can already see that with leading models such as claud and ChatGPT. Right now, if you ask you some questions that will politely say no, I can't comment on that or i'm not entirely sure i'm not engineered to do so. So uh, a popular example actually is google I being D I friendly or legs are not providing advice on how to vote for certain presidential candidates in the U.

S. Election and decentralized models can to remove that central actor away from IT and allow for pretty unbiased freedom of A I. So popular examples can be lama mr. Or whatever that might be. Um if we look at why compute make sense with encysted U A I D centuries, compute can help train these open source models.

So the concept is if you have a kind of global network of computers that are able to provide their latent or unused compute, they can create bigger training runs or bigger a training sessions than a centralized competitor such as apple or google or whatever that are spending hundreds of millions of dollars. The average person can't compete. But if they were to um compile the computer together, they should be able to do so.

There's a lot of chAllenges that come with this specific um sector, but that being overcome massively actually this year alone actually fund little animal to this is at the end of last year um google deep mind which is arguably the leading A I research group um and innovation group outside of O O within the traditional A I sector could only train a four hundred million dollar uh sorry, a four hundred million parameter model um and they couldn't get past this research hum they couldn't figure out how to train larger models and for reference, the best models of today are gonna end up being in the hundreds of billions of parameter. right? So there's a lot of properties as a huge a performance gap between a four hundred million and one hundred million parameter model and they couldn't figure out how to get over the hump.

And this year alone um if you ask me in january, is there any teams that could have resolved this also would have been no. And right now we set here, we have the discussion to decentralized crypto projects, have overcome this research hurdle and have trained successively one point five billion parameter models and a ten billion per matter models. So we've already gone up an order of um fifteen x in times of training this model.

So we are seeing real impact m in crypto A I specific sectors such as like compute data or training models. Um I can dig into a few more if you like. So yeah if .

Price here because like the image that you have been giving me is like we have these like property AI text tax that are you inside the silo de world gardens of silicon valley and maybe they that correctly if you want to like relate this back to like you know maybe language that we've been using. Our bank finance used to be inside of the silo of wall story. And defy has been trying to open that up, have like open A P.

I. Access to like financial tools. 嗯。 And so with like the these building blocks, blocks that you were talking about of the base models, the data, the computer, well, actually only facebook has facebooks the data ChatGPT has like whatever data that um that OpenAI has and like there probably, i'm guessing, not really sharing their data because they're all there are.

They're in competition with each other. And we also don't like any of this, may from the perspective of like human freedom and like you know western liberal values, i've been able to like access these technologies that aren't like Operated in the kind like orwellian over lordi in fashion. I I don't want necessarily like the values of google to be imposed upon like the air that I have access to.

And when now we have a cypher or AI, which is just trying to can you be combined together just open source building blocks. And i'm assuming we are we collectively as egypto AI industry are kind of like far behind the centralized companies, uh, because they just have they have the ordination abilities of centralized companies. They can move faster. They have more resources. But i'm guessing the hope here is that if if we can figure out how to like lean into the properties of decentralized A A I uh crypt O A I, we can actually start to become market competitive in our AI products verses like our still on valley .

overlord s yeah. I mean, the best way to frame IT, David, is the problem is this stuff is super expensive to fund and to build like you're in billions and millions of dollars worth of you know computer train your average basic model at this point um but you'd also argue that crypt tos number one use case is its ability to coordinate humans and resources and the number one resource is proven itself to coordinate is financial capital.

And that's been the main thing that enabled IT to spend up all these decentralized A I layers such as purchasing all the GPU or aggregating all the GPU to provide compute to train models or have models be influenced. Um IT is the compilation of different pieces of infrastructure that can scrape data. For example, like grass is a is a pretty cool project that um has just been able to do that and has like the largest edit data set already right um to your earlier point, um monopoly in the traditional world have a huge mode and the main distinction to make a Better model is the quality of data that you have. And right now we live in a world where data is god within these different silos, gardens. So being able to build some sort of infrastructure that is open source and global where many people can kind of contribute their data, volunteer or IT has ability to scrape the data and compiled in one resource that can then be used by different A I models or incenses is super import OK.

okay. So that's kind of like the advantage that, that we have that centralized AI companies don't have. And I just do want to want to reita when whatever products come out of like the crypto AI sector, it's gna have like the values of open source, the openness of open source like that, hopefully the credible in true of open source because I don't want like microsoft or google specific politics or filter to filter out the AI tools that I use, which are likely going to come to dominate the internet, especially for Younger generations. So 第二， maybe is worth articulating, like actually what's .

at stake here? Yeah, absolutely. I mean, what you're considering here is okay. So let's look at the world where there is no kind of open source development in this particular field, and we just leave IT up to the monopoly to kind of create their models.

Well, that's going to look like is they're going to create five to six leading foundational models. Um everyone will need to pay them some sort of subscription or fee to access these models. These models are going to become quintus essential tools that is needed for you to perform your job, live your life and be integrated with humans at every different layer of your life.

That could be friendships engaging with people virtually or digitally transacting with people are engaging in online experiences. Um earning money many different ways. So if we assume that there's no open source side, then suddenly these very few entities or corporations have a huge amount of control and influence over how you perform within your daily life right within your job um within your daily activity of whatever that might be.

Um the reason why open sources so essential is IT offers us an alternative that sounds very similar right um to what crypto is offering to the financial world in many other things. IT offers us uh a rough or a lifeboat or an alternative set of rails for us to build and scale with this technology. And to be honest, I think this is gonna the most defining technology of our time. And i'm referred to A I here, and I think IT doesn't exist in its full form without critter rails.

I guess what what my question is on this foundation layer over same crypto.

I like, what is crypto about you? Like what is script have to do with some these things like the base models, the data that puts as far as I can tell, like it's is generally like the facebook of the world that are open sourcing kind of the laa L M like model, right? It's like uh what is what is crypto? Do you you have any projects uh in mind? Like because what is cyp to about this a this base stack in, aren't the um besides open sourcing, you know different data models? Like aren't the centralized kind of actor's ordinators Better able to have like achieve the economies of scale to really um bring bring compute to the lowest possible unit Price like that? So this is part of what sometimes loses me and kind of the you know A I is amazing.

Crypto is amazing. Dt dt cypher, N A I is amazing. Like is are we really doing anything on the foundational layer to like improve this stack here? encrypt?

Yeah I I I believe we are. So um let's take some let's take some examples of the stacks that was talking about so far. I can add on a few as well.

So um I already mentioned one, which is the compute stuck. Um I think prior to uh, the end of last year, decentralized computer couldn't train any kind of model that would have been considered meaningful. And so far this year, in the progress of about ten months, IT is scaled pretty exponentially. So you can train now a temple parameter model, expect that to probably increase to somewhere near or fifty to one hundred billion parameter model uh over the next six to twelve months. That is rapid progress for a sector which had um in the multiple chAllenges at the end of last year.

So what are those premiums to do? Just they make Better models like they're more sophisticated.

And what is A A stronger product, I think a very vague but comparable is when you create a character on a game, anyone equipped with different properties, you try to, you know, give IT different types of abilities. You know, you had points to.

oh, it's able to do this vitality.

all that kind of stuff, and think of this as the same thing. These different parameters tune a model to be able to respond and do many more things. The fewer parameters the model has um arguably the less that I can do. Yes, exactly.

When you say decentralized computer, like what are you talking about, right? It's like you you're talking about a lot of these models are trained, you know they cost hundreds of millions of dollars and the world like largest, most ophite ticad data centers. Are we saying there's like um G G P U tech networks running in people's homes that are kind of the training this like what are we actually talking about when we're saying decentralized computer here? Is this like a file coin type of network similar?

Yes, exactly. Um so you could argue that file coin was one of the earliest examples of a decent first computer network. They were focused on storage.

Now they've evolved more into a doing kind of like a cross stack thing, which includes compute. Um some focus projects that are already working on this and have been for a while, uh uh random. Some newer projects are things like iron.

Uh, you've got the jensen of the world that are gonna launch and focused on decentralized training specifically. So I think it's important to call out that decent rise compute doesn't need to serve every aspect of what an A I model would need. So we're spoken about training, ryan, but there's also influence as well.

And these two different things require different um designs of A A computer hardware if you like. So if you think about training, you need specialize hardware. If you think about in france, which is the cover of making A P I calls to a model, you need something much less fancy.

And that's where like typical consumer hardware scales really massively with that. And if you think about like being able to run a um you know a staking note at home for your east network or whatever that might be, it's very similar to something like that. So uh, i'll give an example.

Grass network is not a computer network, but it's um a network that is able to aggregate data. I run my chrome brows er extension daily and IT basically uses my late compute and brothers extension to go in scrape data walls. I am not using my computer or i'm already using my computer. So the way he looks like is very similar to setting up your own note at home.

And I just my understanding, uh, all of the centralized companies always like kind of a had the advantage or at this have had the advantage just because they've been able to, to run faster. But you're alluded to the fact that this actually, at least in some parts of this AI stack, maybe maybe with a computer layer, maybe with others, uh, the decentish zed version of the A I sc has actually been able to have the breakthrough.

That true. Yeah, I think so. So again, to break up what was spoken about so far, I think base models have a huge advantage to being open source in the center that they can mitigate censorship and any kind of common constraint that these monopoles can place um and filter.

Uh I think on the compute side of things, if you wanna assume that we need kind of like growing, growing or need to meet growing, growing demands of compute requirement, train models and to influence models, you need a network that can scale. A A clear example is a google training centers. h.

The fact that they weren't able to acquire enough compute centers to be able to facilities their recent training one, and they had to like kind to build IT from scratch themselves. And even then, they need to have a kind of closing coordinated between each day, is they basically building out some kind of framework or network to allow these M G P U to help train that model. So we're already seeing kind of like similar analogies that and then the third and final layer, which I think import is um the data side of things.

You mentioned earlier, David, that um data is kind of captured in silos of motes with the big companies right now. And you're absolutely right. And I would actually say that data is going to be the main differentiator for models going forward. The reason why I say that is um I think it's gna become pretty commoditized to create a foundational model.

So I think the claws down tropics uh anthropic clads tragi t of the world, they are gonna similar uh performance capabilities um in the coming years where the real distinctions could be made is how personalized that A I can be towards the individual. And the only way that you can make that A I individualized is to train IT on personalized data. And that's why I think we're gonna kind of face our biggest chAllenge yet.

To be quite Frank, i'm not entirely sure how we are going to solve that just yet. I don't see any major progressions right now that would suggest otherwise. I think that it's gona take a little bit of time to to figure out.

So for position, this is as like an arms race, a war between the centralized A I companies versus the decentralize I stack. The data side of things seems to be pretty heavily favoured. The centralize company is because like how if it's decentralized data, then we all have IT, which I guess is great.

But like the idea everyone having all of the data is also Carried to me. And I don't even know what that even looks like IT, that's even a sensible statement. So like okay, maybe the first two, the the compute in the models maybe maybe decent sized AI stack has like actually a leg up versus decent script side, but the data is like what we don't have and data ki always seems to be proprietary.

Ah yeah absolutely. And I think it's gonna one of the bigger obstacles to sort of overcome here because I I I don't see a world and and this is just me being very honest, I don't see a world where people um abundantly opt into just giving bits of data and getting paid towards T I like my number one question would be how much are they getting paid for IT? Do they even know what kind of data to give in? Probably not.

I just want to check a box, sign my terms and agreements. I know this sounds pretty disturbing, but IT is what's baked into human nature right now. Uh, I want to be able to sign an agreement and then just go play with the thing versus, you know, think about what kinds of data is gonna shared and how that might restrict the quality of product that I use. I'm just i'm just not too bothered about and I don't think many other people will be as well.

Yeah, I agree with OK. So that's like the bottom of the stack. Uh, base models, data and compute.

These are like the three ingredients that like start this whole A I thing off and so that's like the layer one. You're got to have these things even have the AI at all. What comes next after this?

Um it's the middle ware layer, middle layer. I'm going to have run through these quickly, David, so that um we can kind of give a Better overview of how this works. So the middle layer is pretty much consistent of one major thing, which is routing.

So the way to think about this is at the bomb, you have all these foundational things, right, these building blocks that allow you to create the magic at the top. You have the a player. So you have all these different apps.

Gonna leverage the magic to do things that specific users on the consumer, they will be able to want to use, but you need something that links them and that's basically what I call the routing la. Um the best way to think about this is um inference routing. So the concept is open source models need to be connected to an APP.

You need to be able to make calls from the player to the models in order to perform whatever calculation they're trying to figure out. So imagine married different apps that need to query alarms multiple times to the second. And these reuters will be responsible for getting them the best answer by reaching them to the right model. Um and the resort allocation is super interesting here because you need to effectively distribute compute and data to the words the right apps.

You might have one APP that requires a half of a lot of compute to process one particular request and then you might have an another APP, which is just trying to Carry a regular model and and trying get some kind of basic output like, you know, what should my diet plan be for the weekend? I uses your own personalized data bucket and IT just pulled straight from there. So there's many different flavors and ways to think about that at the at the middle.

Yer, okay. So this just changed actually how I about this. And until previously, I go to complexity or or chat G B T, I go to that website, call that the APP, and I typed my query. I always kind of thought that like IT would take my query and then I would go to the one single model in the back end and then spit me back out like the data back to my APP. But what you're illustrating is that will actually there could be like a variety handful, a conStellation of models and my input might need to, uh, as result of those models or what why input is needs to actually coordinate between relevant data and relevant models. And that all kind of be orchestrated in order to actually spin me back out something useful.

That's exactly right. yeah. So think of like a world where you don't have to delinquent between which of which is the best model. You could just have a router which can query both centralized models and decentralize models and um have some kind of network that picks the best answer and improves more and more over time.

Okay, right? That that makes that's pretty easy. It's just a ordination layer between all of these beggars, an models that we think are going to be created. So when you go up to the higher stack, the apple yer, is that what i'm used to when I take go to perplexity or or to ChatGPT or or whatever I was the apple yer.

ah that's how IT looks like. Uh, pretty much at at this current point. And I expect this space to kind of like blow up massively. I mean, the V C investing in the traditional AI world has been huge this year, to say the least. Um so the concept here is, uh, we talk about uh, encrypt to this five up.

This is um I think this is the the realized version of what it's going to end up becoming, right? So uh human facing or in many cases agent facing apps will dominate online interactions. And I think that will fn al more people into cyp. T um because you won't even know that crypto o is being used, I can actually give you an explicit example. Um so I met this team.

Um I think they called toshi there there a subnet on bit tensor, which is this this A I lao one um I met this team, uh, h permissionless actually guys and I spoke to them and I said, you know um what do you guys do and they explained, they basically said that um there service produces financial trading algorithms for small hetch funds. So these are traditional hetch funds that are trading like very traditional commodities is right, not crypto. No um crypto assets inside, no mean coins in sight, right? And I asked them, well, okay, um you know how are they paying for this thing? They say it's in U S D or U S D C.

Um I am saying, okay, do these guys have any idea that there's crypto rails in the back and that are using these things? And the simple answer was no. I then kind of like dog into the archetype of this team.

And uh, none of them had worked in encrypt before bar one of the cofounder ers. And this is a fifty percent team that are selling, you know, and six to seven figures per to these traditional hedge funds. So I think it's pretty astounding to talk about the world where we want crypto rails. You know, you talk about the defi mullets back in the day where people have no idea that critter is are being used in the back being realized so soon and so early on in an emerging sector with encrypt, it's pretty awesome.

Um huh that's pretty cool. Um okay, so with all of these stacks, we just run through them the foundation of dar applyed um when you just talked about like to fat atheists, uh is is that where where are VC allocating is debate is our contention about where value is captured because this was like my original conversation getting into crept is like what part of this deck capital value is?

Honest why we started bank list is a kind of answer that question. This in the gypt world, like, right? I came to the conclusion that money is like wear a lot of the places where like value aggregates, which we would elude to, like somewhere down to the bottom. But I have no idea if that extends itself to like this script a ice dack. So that's that's a question to you is like where where this value can captured.

So i'm i'm going to give you the right curve take to begin with. Then i'm going to go i'm going to learn pretty hard into the left curve take.

And so the right curve take is the sophisticated spectacular take VC coins. VC coins. yes. And then the left curve take is the a the seventy I Q. I don't think .

about the correct correct that so traditionally how we ve approached um investment within the crypt to sector in particular is infrastructure. We need rails to be able to even do anything. And I would argue that over the last ten plus years, we have made amazing progress.

We now have stacks that can have huge amounts of block space process transactions millions per second detra. Tata, you know, l 2l ones eta。 So I actually think that we at a point where we have a performance capability to do a lot of these things, having A I agents will actually increase that requirement by ten to a hundred fold.

We are going to need more crypto infrastructure. I think I think this gna .

be a large demand of block space when we have A I agents because they gonna want to make more micro transactions in the average user as currently during a peak bull face, for example, when we are all trading ship coins, meme coins or whatever coins you wanted describe. So from a fundamental infrastructure layer, we need to be able to create um infrastructure that is optimized for these A I agents.

Now i'm not suggesting necessary that this requires a new type of l one or l 2IT just requires more added thinking as to how this stack is built and how IT integrates with our new found user， which is an A I agent. So with that being said, there's a lot of um investment that i'm seeing, at least on the VC side of things in infrastructure, l ones and l tools, particularly like how to build these different things. We're seeing a lot of these agents and cypher to projects launch on base, for example.

And you know there's a reason why the launching on base verses not on eat all one right now. If we look at all the layers of the stack um from a kind of right um curve perspective, um we need different forms, uh sorry, different networks to be able to aggress data. We need different networks and architectures to be able to figure out what the best way is to serve compute and aggregate compute.

These are the infrastructure investments that I think are gonna underpin a lot of APP development in the future. So we could just leave IT at that, right, and assume that, okay, we're not going to see any cyp to AI apps for a while now, and we should just kind of focus on the infrastructure layer. No, run back the last couple of cycles.

And you know it's it's a pretty simple playbook. Here's my thesis as to why I think that's going to be extremely incorrect. I think that crypto AI is number one breakthrough case is going to be a consumer APP in particular.

I think it's gonna an agent and taking IT even to a spicy level. I don't think it'll be a financial agent. I think it'll be a human based agent that attracts uh, attention and requires engagement from your average x user or far cost to user. Um I think it's someone are sorry. I think it's an APP that people communicate with pretty frequently in a chat interface, that there's a bunch of things for them and that doesn't necessary need to be financial.

I'm getting visions of just like an A I new pet, like a neo pet that's just like super smart and actually like sophisticated and like fun engaged with. Ah I kind of you're illustrating similar.

I mean, we can talk about A A few examples. So I I think the number one thing that drives a lot of financial value, encrypt an, has proven uh, over the extent of its uh, life cycle, has been narrative. A narrative can be distorted into something as simple as attention.

So then you could argue, well, the thing that can drive the best, uh, attention and narratives will be the thing that drives a lot of value into critical, a lot of investment into cyp to now we ve seen a few examples with a the gospel of gosa, which is this an example of a social, uh, A I agent that can talk to users and that can reference human slang and sounds completely like a human being. But it's it's just an LLM in the back end. Um I capture a huge amount of attention.

I mean, like this twitter account has gone from five k to one hundred and A T K currently in three and a half weeks, which is pretty insane. And all of those have, you know tweet regularly, twenty four, seven huge amounts of engaging. So I would say that I could come in the form of that new pad example, David, I think you'll get integrated into games.

You're seeing um uh the prime guys do this a lot. Um you're seeing this appear in forecasting agents, uh, with community, for example, where you can just kind of like chat to IT and is doing a little things. You have the financial element as well with the AI sixty z which is of a uh a play on a sixteen year, right? So it's like A N A I powered fund manager. So I think they'll be a few different instantiation of IT, but I think that's where the break through p is gonna.

There's a been a couple like themes in rapto over the last like eighteen months in the size like despair market that has felt um we others just like rebelling about like ming is been like how many years since we've invented crypto u where are our apps like we don't have a killer apps. Sure stable coins, but like honest, this is the dollar.

So like we know where's the apps has been a big theme over the last like almost two years now where all all the users has been also another big theme. We don't have any more users. We're actually like starting to like feels like I starting to leak users, like people quitting cyp to so you don't have any apps, you don't have any users.

Uh and then also like mean points have been a third theme, which has like kind of created the economy, like the attention economy. People have realized attention is actually just a hugely valuable thing. This is why I mean coin's have such a product market fit because they drive attention.

And so I am seeing like a lot of these themes that we've had like defined craped over the last two years, lack of users, a lack of apps, h and then the attention, economy and A I agents kind of check all of those boxes like, well, we get new users. They're just a uh, we get new apps there. The agents uh, and how do they work, they drive attention. So there's like a big void here that I think like the whole agent like thesis idea again, this this idea but allowed people pointing towards is like filling all at once. Ejaz has been great thinking we're walking us that through the AI agencies.

Thanks for having me think this nation.

you guys know the deal, crypto o is risky, meaning ines are risky no matter if there is a human or an AI agent shilling them. This is the frontier. It's full of robots and AI overlords since going to to look really weird.

But hey, at least is succeed. We're glad you're with us on the back. This journey takes a lot.