Projects
![]()
No projects found
Investors
New
DAO Treasuries
DATs
Lend / Borrow APRs
Chains
Categories
Spaces
Brains for Chains: AI meets Aggregation
Recorded: Aug. 25, 2025 Duration: 1:02:39Space Recording
Full Transcription
Thank you. GM everyone happy happy Monday we will get started in just a few minutes I am very excited to be
talking to a lot of really really smart people on AI today so it should be a good one we'll give it
just a few minutes for the space to fill up and get going.
Until then, just sit back and enjoy the music. Oh I'm gonna be like a baby. I wanna hear U-CHA! From the hunt you never had to the now! I rode the marathon till they made me last five! If you battle me, I feel a little vibe!
People always say my style is vibe!
You got core, you got God!
Spell to me, I'm a ride for five!
If you wanna battle your engine now!
Moving from your 80s to the 25 style!
Go on, get my rhymes out, cha!
Cast you all into exile.
Jazz and a wall, that's our team. Step into the party, disrupt the whole scene.
When it comes to needs, well, I'm a beat.
I like my sugar with coffee and green.
I'm like, I keep it going, keep it going, full scene.
Sweet to be, so much and nice to be. We're in the fuck out style green I'm like, I gotta keep it going, keep it going, boozy Sweet to be so much and nice to be
With a talk-out style, I'm not too keen
Trying to change the world, I'm a clock and scheme
Mario T likes to keep me clean
Gonna shine like Aesha TV
Fun run cause that's my team
Change my defaults, ticket to me
When it comes to MPs, all is D
Jealous of the rhyme and the rhyme routine
Another dimension who shall exceed
Intergalactic planetary I'm going to go. to
me me me me me me me me
me me me me I need you to get mocked, I still find you in my rock You need to start shakin' if your fingers pop
Like a bitch, you neck, or Mr. Slot
Alright, welcome everyone. Happy, happy Monday. If everyone can just let me know that
my voice is coming to you, that would be greatly appreciated. Just give me a thumbs up. Okay,
there we go. Thank you, Deepin Jesus. Yeah, so we'll give it just another minute. I think we are waiting on someone from Sentient.
They were here earlier, but, well, yeah, Vivek, I sent you a request to speak.
So if you want to come up and get on stage and we can get going.
But without further ado, we will start with introductions. So we are talking all about AI and aggregation today
and what makes it all work. So we have a few really, really impressive projects that I am
very excited to talk to. So without further ado, Deepin Jesus, why don't you go ahead and introduce
yourself, tell us a little bit about
the project, what you guys are working on, what you're building, etc., and then we'll move it along.
Sounds great. Thanks, everybody, for tuning in. And obviously, thank you guys for having me over.
It's going to be a pretty fun space. Thank you, Timmy, for hosting this.
pretty fun space. Thank you Timmy for hosting this. So I'm DeepinJesus, I'm the head of
DevRel at IoTeX. IoTeX has been in the let's say physical infrastructure IoT plus blockchain space
for many years now, since 2019. All this to basically culminate with our let's say
culminate with our, let's say, latest expansion into the AI realm, where we are going to basically
use all the data, knowledge, and all the years of basically being in the physical infrastructure
space, in the deepened space, to basically provide now AI systems with real-time, real-world
knowledge that is dynamic and not static.
And we think that's going to be a super important evolution for the AI sector.
Besides, so obviously pushing the DevRel team,
we've been doing actually some interesting stuff with the Polygon team
actually some interesting stuff with the Polygon team as far as like accelerators goes
as far as accelerators goes.
we just launched a few super bullish Debian projects on the market
we have a new season of our accelerator program called
Traders Got Talent coming up in the next couple of weeks
so if everybody in the audience is a builder
and you want to reach out to me
feel free and or either maybe just follow the IVEX channels
where you will be redirected to maybe submit your project
and get a shot at winning a big, juicy prize.
So yeah, we'll maybe get into a little bit more details later,
but that was just a quick intro about myself.
Thank you. Awesome. Thank you.
Awesome. Thank you. Yes, I am very excited.
This is going to be a great space on a Monday.
Usually it takes us like 20, 30 minutes to get some help, but we got some in the first five minutes.
So we are doing great.
Thank you for that deep engine.
Best way to start the week.
Oh, yeah. Absolutely. Absolutely.
James, why don't you go ahead and introduce yourself
i'm james founder co of mode we're building a trading focused or ai trading focused l2
where agents can trade hedge coordinate across defy markets. And we're very excited about verifiable agents.
We previously built out some or worked with lots of
AI agent projects, building out early infrastructure.
And there are a number of agents live on mode already, including
live on mode already, including Olash's Modius agent,
which trades across Balancer and other DEXs on mode.
And yeah, very excited to chat more about AI
and the kind of intersection with on-chain activity.
That's fascinating.
Now, I love the AI trading L2s.
I actually think that, I mean, we're seeing a whole explosion there, and I think that
actually DeFi is really complicated, or can be, and what you guys are doing and just AI
in general in DeFi is fantastic.
I think there's a huge market fit for that.
So I'm definitely excited to dive in and talk to you a little bit more on that.
And last and certainly not least, Vivek, why don't you go ahead and introduce yourself
and your project?
Yeah, thanks for having me, Timmy.
So everyone, I'm Vivek.
I'm from Sentient.
So Sentient's main mission is to ensure that AGI is open source and not controlled by any single entity.
I guess to put it a little more simply, basically, we're picking up where OpenAI left off.
You know, they started this grand project to continue building AI, which is the most powerful technology humanity has ever seen.
And they said it's for the good of humanity and we'll keep it open.
And then they made a decent amount of money and decided they can close the town and make
a lot more money.
So yeah, we just picked up right where they left off.
I love it.
And yes, for those who don't know me, I am Two Cent Timmy.
I am on the Polygon and AgLayer marketing team.
But you're not really here for me.
You're here for our guests.
So let's dive right into it.
So I think we'll kind of start off setting the foundation going like, I mean, the first
question is why, I think, first.
So like, why does AI,
I don't want to say why does AI need blockchain,
but like what can AI be used for on chain that makes it so powerful?
And why is it a need in all of Web3?
And then what about blockchain actually adds to AI?
Because I actually do think
it can be a symbiotic relationship
and both make
each other better. But yeah, we will leave it wide open. If any of you three have thoughts,
you can go ahead and just unmute yourself and start going. And then if anyone else wants to
add on, just put your hand up with the little reaction button like that, raise your hand and I'll be sure to call on you. So yeah, why is AI needed for blockchain
and what does blockchain have that helps AI
is kind of the first open-ended question.
May I start?
Because maybe just like a quick eye level for listeners
and obviously we can just dive in.
One thing I always think is like uh i find it funny that um we think about all these like different technologies kind of like a separate
because we're kind of used to like compartment compartmentalize knowledge uh but at the end of
the day it's i think it's all gonna it's basically all one thing right so AI system can use
obviously decentralization it can use data from like real-world devices like
all of these like the IOT the AI element the blockchain element basically
eventually all get together and create just applications for whatever mankind right so in a general term all of this is
basically part of the same thing now having said that why do all these
compartments need each other if you think about for example the way that
deepin has like and will be even more uh propelling ai uh to basically you know decentralize for example
computation decentralized gpu and so forth and so on you can see and i think maybe was it uh you
know master like dropped something like that uh a tweet similar to this like a few days ago like
the the the decentralized the edge uh computation part in ai is going to be crucial because it's going to be impossible to keep it all
and offer an equivalent service in a centralized setting.
On top of that, you can see that knowledge, for example,
dynamic knowledge, real-world knowledge,
it's something that your general AI model doesn't quite have, right? So there is a great value in like insanely kind of like all-knowing models,
but at the same time, there is a ton of value into like specialized,
very granular data that knows the user really well,
that knows any sort of like, I don't know,
specific temperature in like specific pockets of like a certain city
or traffic
conditions or car conditions and so forth and so on so there are all these like little verticals that
for example with the ai uh iotex ecosystem we can we can tackle where basically we create like
pieces of knowledge uh for these verticals that can be used to enrich AI systems.
So, yeah, they're all really needed for scalability.
I think blockchain is a great way to scale AI.
AI is a great customer for any sort of like blockchain-based application because it's
eager for data.
And so it's like a perfect combination,
in my opinion, of supply and demand.
Now that's a great overview.
I do think it's interesting.
Like AI does, I don't know.
We, I would say we have a,
I don't know if it's a joke or tongue in cheek.
People are serious.
I do think a lot of people that say it are serious where it's like,
bring everything on chain.
And then you look at like,
well,
that includes AI,
but AI is also this.
I don't know.
It seems like at this point,
this infinite knowledge where like,
you can just come in and use it for all sorts of tools.
And it has an application everywhere.
And maybe some are better
than others but it actually seems like it's improving basically every single industry uh
vivekar james i didn't know if there was anything you wanted to add on uh to why
ai and blockchain seem to need one another yeah so we're very focused on defy and agents that execute transactions on behalf of
users and i think that you like humans need these blockchains can benefit from AI because it can automate a lot of tasks and it can do that with much more complexity than just running simple trading bots or bots like simple automations.
AI can add a huge amount of context and at the end of the day that's to increase profitability on chain for users. That's
kind of like the core goal. So AI can definitely help optimize that. And as an example, especially
when it comes to trading, users often have views about the market, like, and their views can be
right or wrong, but they can express their view by like, I want to buy BTC because I think it'll be 150k by the end of the year.
And if BTC ends up being 150k, that user was right.
And oftentimes they don't actually capture all that value because their execution of that trade isn't
correct. They might leverage, they might put on too much leverage, they might not manage risk
effectively and they might not capture like the upside. Whereas the way I see AI working is AI
can, you know, you can use a user can express what you want to do and you can use AI as a tool to be able to execute your
view on the market much more effectively and manage that position much more
effectively than if you were just doing it yourself. And the reason is because
AI can bring in a lot more context around the position, around all these different data sources
and can help you as a user. So why does AI need blockchain? So recently there's been a lot of chat
in the Ethereum community about this agent to agent standard.
And I think this is a very good example
of why agents need blockchains,
or why AI needs blockchain,
is that blockchain is a foundational layer
for agents to interact that's trustless and verifiable.
And as agents, if one agent is interacting
with another agent, it needs to know
that that agent is legitimate
and it's doing the things it says it's meant to be doing.
And blockchains are great at ensuring
that you can verify different types of data on-chain
or transactions on-chain.
So that's why AI needs blockchain.
Yeah, you just gave me the biggest aha moment I've had
when thinking about AI and on-chain activity,
where I'm like, you know what?
The communication between agents is actually...
I mean, we talk all the time in this space in general,
and I think it's this big narrative of like, you have provers and intents and all of this stuff,
but like, it also applies to on-chain activity and still, I don't know, be right,
but not effectively be right because they lose upside. It's almost like AI can actually be there
to help protect people. I know even in traditional markets, like one example is just like people
don't understand why doing leverage
on stocks that rebalance daily may be a bad idea, and they would actually lose stuff to volatility
decay, or things like that, where I think on chain trading, it's, it's very much a needed source
of information. But we will keep moving on with the questions that I have. But yeah,
you just like, you opened my eyes right there. That was brilliant. So one thing that we think
about, and I guess this is kind of what you were getting at a little bit, James, was like, how imagining AI agents working across
multiple chains, without users having to worry about where their assets or data is. So like,
what does that look like? And I guess another question of like, who will, who do you see as
being the people who build these AI agents and kind of what mechanisms need to be put in place where there is that level of trustlessness that we're all working towards and why we think working, why on-chain is better because you have this trustless, decentralized aspect.
But who's going to be, I guess, building those AI tools and how does that look kind of out there?
So I think that the way we're going to see
users interacting with DeFi in general,
and it's already kind of starting to happen via
vaults is there's going to be a lot less of users moving assets between chains regularly
right and managing positions across chains they're more likely to have automated strategies
that manage positions across chains for them using like different types of intent layers
like tools like vaults and that's where I see like that's how I see users like human users
using like agents that run across multiple chains. When it comes to
I guess, yeah, agents running across multiple chains, I'm not sure if there will be one hub,
I guess that will be where the agent is registered And it will interact with agents which
are also registered in other chains.
And yeah, I'll probably use the same bridge layer
or the same intent layer to talk to each other
between different chains.
So I guess that's where AgLayer comes in,
as it can aggregate that trust between those agents
across different chains.
No, that's great.
I definitely think there's this like,
we talk about ag layer and we talk a lot about
a shared liquidity layer,
but I also do think that we're going to,
with all of these interop solutions,
we are going to move to not only liquidity, but
also this aggregated data layer where you can have these really large data sets and
taking inputs from a whole bunch of different chains where you can, I would say, kind of
react and put in strategies for the DeFi focus people of like, this is what I want to do. And if, let's say a large portion of Ethereum is being liquidated on Arbitrum, then maybe the Ethereum that you have on Katana, you may want to do something with when it's happening across, like on a
different chain, you can build your strategies that way. So that's, that is kind of where I
see it going. I guess one thing we see with AI agents all the time, and this is kind of an open
question and I'll, when we think about smart wallets and wallets that act on behalf of the user, but what, I guess, I'm thinking about the wallet experience.
What does that look like?
And will you have, is it possible, will you have like multiple agents running on a same wallet?
Or will it almost be like, this is my wallet and this is the agent that runs on the wallet.
I guess that's a question.
Like,
will it be singular one wallet per agent or one agent per wallet?
Or do you think we'll have multiple?
I'd love to hear your thoughts on that.
Yeah.
So just to answer briefly the previous question about AI agents working
across multiple chains,
I think that's the obvious path for the future.
So I guess for context with Sentient,
I guess in a way we're aggregating intelligence.
intelligence. So this partnership with Aglier works quite well. But if you think about intelligence,
So this partnership with Aglier works quite well.
everything is all in one place. Let's say the internet, or let's say we can say Google.
So the fact that we have agents that can only operate on one chain, for example,
on one chain, for example,
and this is something that we're doing with right now as we onboard agents onto our platform,
and this is something that we're dealing with right now
as we onboard agents onto our platform,
we have an agent that specifically analyzes base wallets.
We have an agent that specifically analyzes ETH wallets.
It makes a lot of sense to have just an agent
that can analyze across chains.
So I don't know how this happens.
Hopefully it happens through AgLayer.
Once you aggregate all the chains,
then someone will definitely make an agent to work across this stack.
But we've already seen it work in intelligence, and the next step is for agents to work across chains.
With your second question about the wallet-to- to agent interaction, what does that look like in
the future? My personal opinion is it'll be one wallet and many, many agents can work across it.
So actually we're working on this at Sentient right now on these complex workflows.
And an example I give a lot of time and actually one that we're working on in engineering is,
can I give my wallet to an AI platform, say, here's my wallet, here's like 500 bucks.
I want you to analyze all market sentiments, figure out the top five meme coins that grew in the last 24
hours, analyze my risk portfolio, my wallet, and then invest across those meme coins matching my
risk portfolio. So this is like five to six different tasks that I've asked this AI platform to do or this AI agent
to do. And our thesis is that each agent can be specially customized to do each of these
tasks. So one agent can do the market research, one agent can access my wallet and then check the funds. One agent can access my wallet and do the analysis. One agent can do the market research on the meme coins, etc. So I think the wallet is the grounding experience. And then each of these agents that are specially customized to do something can jump in, analyze the wallet and do whatever they need to do afterwards.
jump in, analyze the wallet, and do whatever they need to do afterwards.
That's interesting.
It makes me think, too, so as you were talking about, like, the meme coins being traded,
I'm just thinking out loud here, but if you have thoughts on, like,
one of the powers of on-chain activity of like this really really nominal small transaction costs
across most L2s when things are put on-chain but I think one thing that we've probably
underestimated the powers of it actually gets rid of a lot of spam and bots so like I it's
relatively easy and cheap to just spin up a ton of bots on Twitter, start posting about meme coins, and then if you kind of pull the data and scrape the data, it looks like certain tokens are doing well.
Where when that data is on chain and everything, every post about a meme coin now has a small cost associated with it.
It kind of adds up quickly.
Do you think the AI agents, they'll get,
like how much data will they pull from on-chain,
like public resources, and how much will be Web2 API pulls?
Do you have any thoughts on kind of where the data is going to be sourced from?
And then maybe even where the data is sour to be sourced from or and then maybe even
where the data source from now and where you see it in five to ten years yeah it's an interesting
question um so i can say let's start with with how data is being sourced now um so one i guess
i don't know interesting or like negative thing or maybe even positive thing about crypto is there is an insane spectrum of information available.
But no one has really solved the problem of getting everyone the best information across the spectrum.
And that's what AI is helping with a lot.
How do we go across 30 different data sources
when someone asks,
what's the price of XYZ token?
How should I think about investing in it?
That's like a two-hour project
if you really want to dive deep
into all types of data that crypto offers.
Because everyone has opinion and crypto offers a lot of information about this kind of stuff.
How we do it now is we've developed some very complex workflows to get the best data across the internet, whether it's like DEXs, sexes, DeFi Llama, et cetera.
These very well-known players, not well-known players, sentiment across social media.
But the hard part is it's not gathering all the data.
The hard part is how do you make them all agree? How do you remediate all these differing sources of data? So that's how it's done right now.
I would say I hope that someone figures out how to give good information without it being tainted by the incentivization layer of crypto. you know, you can incentivize and coordinate people to do essentially anything.
Because it's fast, you know, it's a coordination mechanism.
But I think sometimes you see very like negative side effects of people doing it for the sake of the incentive
rather than people being coordinated by the incentive.
And I think information has taken a hit with that.
Let's say like Kaito, for example, and the way that they've created this incentivization mechanism for people to inform others on social media.
In theory, it's a very, it's a phenomenal idea. But like, now we can see
the recent updates and how people feel about it. Like, there's still some kinks to be worked out
in terms of how do you keep information clean? Because at the end of the day, AI can gather
everything, but it's very hard to clean it. So the most ideal scenario is you just keep information clean.
No, that's great.
I mean, yeah, I think the incentive that you touched on
is very much like the, I don't know,
what Prometheus and the fire from the gods,
like it's a blessing and a curse.
It can be really dangerous and it can be really really good
um it just depends on
figuring out i would say how to harness that power
deep in jesus we'll hand it over to you
yes i mean uh
honestly i
i don't know i just wanted to start with like a crazy statement like everything is going to be
on chain in the future so that I think again all the system are going to basically work together
the on chain the blockchain whatever provides an incredible level of like verifiability
in terms of determining you know what kind of execution is done
for maybe certain agents, where they're pulling the data,
where is this data coming from,
can I trust this agent rather than another agent,
what kind of score they have, where are they sourcing their data.
All this stuff could be verifiable.
And I think that's kind of where the on-chain element
provides an incredible solution all on top of obviously scalability for like
sort of like decentralizing the infrastructure and I think that's
basically the that is it you know it's kind of like the same when you think about how blockchain works with like
iot systems you know it's a perfect verifiable layer for for iot system to basically just be
trustless it's going to be the same thing for agents and i you know, back to your initial questions on the various wallets,
you can have like, you know, X, like I think it's a many to many type of relationship.
And for example, I mean, we can see it with the X402 protocol, for example, it's like
easy for agents to transact in stables and different coins.
It can be across various ecosystems.
As in the ITEX ecosystem, we basically started this little swarm of agents
that provide each other with a certain amount of data.
So for example, if one of our agents wants to know something about weather data
that the other weather
agent in the storm is aware of, then it just makes a request and then the weather agent
gets paid a little bit more, a little bit for providing the information.
All of this is like automated.
So at the end of the day, on the user side of the equation, there is just a simple language request.
And yeah, so we're planning on basically just amplifying this
to the point where each specialized agent is able to query others
for the information that it doesn't have
and then provide whatever service to their clients.
So I don't know, maybe a long way of answering
your set of questions, but that was my two cents.
No, I like that answer.
It was very, very good.
I think there's, I don't know,
there is a lot of I would say
open questions and it's not easy
stuff to figure out. So I do find that answers
tend to go on and on because if it were
obvious it would have already been done and this is certainly
technology that is not
the easiest to, like I said,
work within control.
Really, like a fire, it can spread very rapidly if you don't know what you're doing.
So I guess that kind of brings me to the next topic I wanted to cover about risks and guardrails and that sort of problem.
I think you have people all the time you see, I don't know, AI.
I don't want to.
Yeah, basically, AI is bad.
AI is dangerous. There's a lot of questions about how much AI we should use.
And then when you start talking about financials aspects where people will be operating on chain, what risks do you guys see of AI agents controlling assets across aggregated chains, right?
So like, yeah, it might be on one chain, but if we get to the point where you do have AI agents operating on the ag layer, like across the canonical bridge with controls to all the assets a wallet or an address actually holds.
What are those risks?
And then what are the guardrails that we should put in place
for things like that?
And I'll just open it up to anyone who wants to jump in.
I can jump in.
So I think that many of the risks are still with like underlying technology of
blockchains or DeFi less than the agents themselves right now, because there are
still risks to using like decentralized technologies.
So as a first step, I think you still
need to make sure users are aware that using an agent, like
the agent is reliant on it's as secure as the underlying
technology that is interacting with, right. So if it's
interacting with different lending protocols, you need to
ensure those lending protocols are as secure
as they can be. If it's interacting with different DEXs, the same thing. And when it comes to,
yeah, moving across different chains, you have the same, yeah, the same questions you need to
answer, like how secure is every chain? How secure is the bridge? Are there anything that any checks that
should be done when it comes to like understanding bridge liquidity, things like that. So there's
still like a lot of I would say like DeFi native security best practices that need to be thought about when building out agents. And then yeah, like the key thing when
it comes to building out agents that run across chains is again like it's this verifiability,
it's understanding, it's like making sure that agent is doing what it says it is meant to be
doing and not doing something different. I think everyone's experienced LLMs hallucinating
at certain points and giving them weird answers.
You know, there's not that much risk
when you're asking a question, you're getting a weird answer
compared to when millions of dollars
are moving across chains, right?
So there needs to be like a lot more done to ensure that, yeah, agents
cover like all these different edge cases and they don't hallucinate
and essentially keep user funds safe.
So that's something we spend a lot of time thinking about.
And I think really the best way is for agents that hold capital to scale up
slowly and, yeah, to just be tested rigorously.
On that too, what you made me think of,
like, as you were talking,
is thinking about it needs to be verifiable,
but I'm wondering, especially in something like DeFi,
where there's an element of PvP to DeFi that I'm thinking of, right?
So like when people start using leverage,
there's obviously their risk of liquidation
and you'll have people try to drive the price up to liquidate
and things like that.
Do you think AI agents will, everything will be put on chain
or do you think it'll be more like a zk tech like proving that
it's verifiable like proving that it's doing what it should um but not actually having like the
actual code or mechanisms publish so does that make sense yeah i think like the attack vector that you're talking about is essentially
like if people can see what agents are going to do um won't they just manipulate them
and i think that's pretty interesting um like attack vector because if you can see loads of
agents are buying eth at a certain point like
wouldn't you just front run them and buy eth before them and then sell eth straight after them
um and we're kind of seeing that play out on hyperliquid right now where like everyone's
trades are public and there's this big debate between like big traders saying that if you have
like a position of 100 million dollars you shouldn't be doing it on chain because everyone can see what you're doing.
But then it creates this game where like, well,
maybe that's what these people want.
I think with, yeah, when it comes to like manipulating markets,
manipulating various things, I think they're all again,
like attack vectors of like the specific market, not necessarily
the agent. But yeah, we will definitely see people trying to do things like sandwich attack,
agentic transactions and things like that, because they might be on chain.
When it comes to different trading strategies, I think most of them will like on chain. When it comes to like different trading strategies,
I think most of them will be off chain
and the execution will be on chain.
So a lot of like the reasoning
behind the agent's decisions will be done off chain
and then yeah, that'll be verified and executed on chain.
So it should prevent some of that.
No, I do appreciate that answer because that is definitely something where i would say it definitely like you said like let's just front run everything um if you know
what the agents are going to do you can figure it out bovek you have your hand up yeah i was just
going to add in a quick point that what James mentioned related to an agent's connection
to the underlying technology being the model in this case.
So theoretically, the model should be the utmost source of security
for an agent.
The most ideal scenario is I just tell a model,
I'm building an agent on top of you,
and this agent should never rug me and should never lose me money.
And that's the perfect form of AI ever created.
But I don't think models are there yet,
but they can get there.
And honestly, they should get there.
We should try and figure out how can we perfectly align an AI model
so that anything built on top of it
will act in that manner.
So to share a piece of research that we've done,
back in April, we tested a lot of open source
AI agent frameworks.
And we found that the biggest attack vector
was the model, agnostic of any model,
ChatGPT,
Quad, et cetera.
And, you know, the sense of security that these models have,
like even today, like I think last month,
we also go crazy and say some insane stuff.
Like that is also a form of security in the sense that,
you know, the alignments that they have are not fully set.
So we're a long ways out of...
Obviously, if you think about the end user,
how do we give an end user the most safe agent?
It's about giving them the most safe model
safe model and the most controllable model which I think we're a far ways out from.
and the most controllable model,
which I think we're a far ways out from.
No, that's actually awesome. That is, I would say it's kind of scary when the biggest attack
vector is actually the model. I would hope that eventually we get to a point where that can go
away and we get like, I don't know.
It really does feel like a cat and mouse game where it's like,
this is,
this is how people broke the model.
Let's fix that.
this is how they're breaking the model now and kind of go on.
We have about 15 minutes left and I do want to give some time for people in the audience to ask questions to you guys.
So we'll kind of end in this.
What future looking,
where do you see this going,
especially for the average user, right?
So like, will the average person
be interacting with AI
when they're doing on-chain activities?
Will it be fully abstracted
so that they don't even know
that they're doing
it? Where do you see the industry going? Or will it be more geared towards, I don't know,
the people that I think are doing DeFi looping strategies now are doing more than most people
on-chain? Will AI help them, but will it not really permeate into your average mother and
father, grandfather going on chain and using
their activity and we'll kind of leave that open to all three of you if you want to go through
we'll start with deep in jesus but kind of where you see everything going over the next five years
oh difficult question i don't know i want to think that it uh i don't know how do i approach this i
i honestly want to see a world where the barrier with any sort of like defy protocol or on chain
is just pretty much non-existent where, like you were saying,
the mom-and-pop or the older generation, the non-crypto generation,
could interface with blockchain seamlessly,
even work with some sort of more complex DeFi protocol
because it's just going to be one obviously beneficial for
with the Web3 ecosystem in general uh it's going to make just you know let people access this all
like crypto world in a way that is super smooth and simple for them obviously there is the whole
like traditional finance route and you know the all the institutionalization of crypto at the moment,
but I feel that the AI, the agents,
could be a great way for people to get into it
without having to learn about how to use these protocols.
So I see already that the user experience
is getting better and better,
but that's what I would love to see.
Obviously, to get there, like we've been saying,
there's going to be a ton of testing, verifiability.
There is going to be a need to ensure these people
that they're not doing something crazy or something buggy
or something that could ruin them.
But that's, yeah, that's what I see.
I see AI is pulling data from real world,
like super granular, super specific,
like user data from all sorts of sources
coming up with like best data set
that it's a blend of like some Web 2,
maybe, I don't know, gated, high quality data
versus Web 3 community contribute data.
AI basically runs or models run on distributed networks.
And yeah, that's kind of like what I would love to see.
I don't know if that's going to happen in the next five years,
but I would be quite happy if it did.
Awesome. Awesome.
Yeah, I mean, we can tinker towards utopia forever.
James, what about you?
Where do you see happening in the next five or so years?
So I think that we're going to see lots of different tools
and apps and interfaces emerge that allow
people to express their views and have AIs execute those trades.
So whether it's something to do with prediction markets, whether it's something to do with assets like stocks, cryptos.
I think there's going to be like,
so like the design space for new interfaces,
which are enabled by AI is going to be just insane.
And we're going to see loads of new apps launch.
It could even be personal apps and agents that you tell it,
what you think will happen.
It aggregates hundreds and thousands of different data points for you,
millions of different data points,
and then it goes and executes on that vision for you.
I think we're starting to see the earliest versions of this emerge on chain,
which is very, very exciting.
Yeah, that's the vision we're building for that mode right now.
Yeah, I definitely see us moving in that direction.
Vivek, bring us home.
Where do you see it going in the next five years?
Yeah, sure.
So I think James summed it up pretty well.
I think at least in the next year, just speaking about AI broadly,
anything execution-based or not requiring that much brainpower to do,
AI will easily automate.
easily automate. I think we're seeing that right now. And something that I've been thinking about
I think we're seeing that right now.
a lot recently, and something we talk about a lot at Sentient, is the most powerful person
right now, at least for the next two years, is the idea person, because AI makes it so easy
to execute on any idea. And 30 years ago, the idea guy was nothing.
It was the guy who could code was everything because they could actually make it happen.
So in the next year, I think anything that can be automated will be automated.
I guess focusing specifically on what that means for blockchain and crypto, I think this is maybe well known and agreed upon that one of crypto's biggest weaknesses is the UX.
nightmare UX story.
I think I still hold tokens that I don't want to hold because I wasn't able to figure out
how to get some extra gas to get rid of them.
And now they're just sitting in my portfolio.
But AI, in that sense, this UX, it's a very easy, but it's like a very, I wouldn't say easy, but it's like a very concrete problem to solve that AI can solve.
So I don't know who mentioned generative UI, if someone didn't mention it. But I'm really hoping people do take a very heavy look at, you know, how do we improve the UX and the UI for people using crypto?
Because people do want to use crypto.
It's just, you know, just make it as easy as possible.
And I think AI can definitely make that happen.
Awesome. Yeah, I would say very frequently on the space, we talk about the really, really bad UI and kind of what we at AgLayer really want to do and see fixing it.
Because you're right, everyone has those horrible UX stories.
And I think everyone who's been in crypto definitely has tokens that they don't want in their wallet just can't get rid of. So we will start with the audience questions kind of quickly.
But GZ DeFi, do you have a question for anyone specifically?
Yeah, GM, GM, guys.
Can you hear me?
Yeah, we can hear you.
Yeah, so my question is, apart from using the agentic AIs for trading,
what are the most important aspects of crypto that we can use them for also?
That's my question.
So like after trading, where do you see ai basically in crypto what's the most important
um yeah we'll we'll start it can be a a fast answer vivek you just unmuted so we'll start with
so dad from from my side um you know we've been focusing a lot of a lot on information
um like i mentioned there's many many sources of data that you know like think
about if you want to do research on a topic or like even say like let's say do you want to research
eth um you know you're googling stuff you're finding you know you're going to different price
sites and figure out the price and you're going even on x reddit looking at
all your favorite influencers to see what they think you know this is like a very complicated
workflow that you're doing manually right now um so on our end we've already figured out how to
you know simplify this workflow through ai do all the hard work of aggregating all this intelligence
into one very good answer for you um so yeah collecting all this information and then piecing it together in a very good answer
deep in jesus sir james do you have any yeah quick thoughts yeah yeah i mean to to build up
on what was said basically uh if you think that your agent can use, let's say, funds, crypto, in order to retrieve as much information as possible, very granular information, very specific information to the task.
the agent interacting with machines to basically, how do you say, like complete certain maybe
physical world types of tasks. And it can be as simple as changing your thermostats to maybe a
little bit more complex as to sending a car, whatever is needed or things of that nature.
So the read part of the equation is great.
An agent can use its own funds to basically retrieval this information. And then maybe that can also use funds to interact with the world around it to
complete the task that you suggested.
And I think crypto is the perfect railway for this to happen.
Awesome. Yes, thank you.
The engineer.
Oh, James, you unmuted. I'll let you answer.
And then we'll go to the engineer.
Yeah, I was just going to say agents for everything.
That's kind of where I see everything going.
I said last year, I think that the majority of blockchain transactions
can be run by agents in the future.
And I think that's still very likely to happen.
So yeah, exciting times.
D Engineer, your question,
if you can make it quick.
We got four minutes left.
Thanks, Jimmy. It actually goes with the James comment.
AI agents and CDK, how do you see that kind of integration going forward?
On AgLayer CDK? Like the CDK chains, you're asking?
Well, yeah, CDK kits and stuff like that like what does
deploy it on friday how do you see uh ai agents integrating into that kind of stuff
i think for me answering that question i think they'll
it's they will integrate i see them almost operating as they do and the goal for cdk
is actually making the communication between ai agents um especially on siloed chains work a lot
better so um there could be i mean there's a lot that goes on, but there's, I know, like, all the speakers basically touch on, like, you have different levels of information, different agents working. And I think when you getX is building, and then how is that going to impact what is happening in the physical real world?
How does that impact what you want to do with trading on chain with Moda or Mode?
different levels of information.
And then when you're all connected to this aggregated,
not only liquidity layer, but information layer,
I think that's where AI agents will integrate
and just start pulling to make decisions
that from information that's happening on one chain
and executing on another chain,
all in the chains will be connected through CDK.
That's really what I think.
But I don't know if any of the other three speakers have any other thoughts.
I got the thumbs up.
Looks like I'm good.
So we have a minute and a half left.
So we'll start with you, James, and kind of work the way through.
Tell us where people can go to find out about your project if they want to learn more and any final words.
Yeah, you can go to trade.mo.network.
You can check out the AI quant, which will do analysis on trades, screenshots.
It's pretty cool.
You can put trades on there as well.
You can DM me on Twitter or follow the AppMode network account.
And yeah, thanks so much for having me.
Appreciate it.
Yes, I definitely learned a lot from you.
It's always great doing this.
Okay, DeepinJesus, you are up next.
Yes. So again, yeah, you are up next. Yes.
So again, yeah, thanks for the moderation
and yeah, for having me.
It was great.
Also, you know, great chatting with you guys.
If you want to learn more about what IOTEX is building,
I will just say two things.
One, follow IOTEX on X.
So that's IOTEX underscore IO and Iotex.io if you want to just basically
redirect it to the whole ecosystem and as i was saying in the beginning guys if there are any
builders in the audience and you want to build in the ai space we are running the second season of
cryptos got talent first season with the polygon team was an incredible success. We got some insane Debian projects on board.
This season, we're going to work on AI.
And if you want to participate, please let me know.
Get in touch.
Either send me a message here on X or through the IonTex platform.
We'll be super happy to have you guys on.
And yeah, thanks again, guys.
Yes, it was great.
And then vivek
you are last but certainly not least yeah like again like like james and deep in jesus said
thanks for thanks for moderating to me um i guess to learn more about sentient um at sentient agi
on x feel free to dm me if you have any questions. We've got a couple big launches planned over the next two months.
Next one is coming, let's see, tentatively next week.
So yeah, we're pretty excited.
If you guys are interested in checking out Sentient Chat,
which is our AI platform, feel free to DM me.
I'm happy to throw some codes out.
But yeah, this was a great space.
Thanks for hosting.
Yes.
And with that, we will wrap it up.
You all, all three of you are welcome anytime
we have the space if you want to go,
get up and just talk and share your thoughts.
But we typically run the space every Monday at 12 Eastern.
Next week, we will be running it on Tuesday because it's a holiday in the US.
But yeah, stay tuned, everyone.
Big announcement for AgLayer just came out yesterday.
Or sorry, Friday.
Last day that I worked for Polygon, I suppose.
But we just launched Enterprise CDK,
which is really exciting
because now we are building tools
to get bigger players in the industry,
basically what the big enterprises
and institutions want to be on chain.
So we are working with that
and hopefully connecting lots of great teams to AgLayer.
But thank you all for being here.
Make sure if you're in the audience, you give all of the speakers a follow.
Really, really smart people really will share a lot with you.
And we will talk to you all next week.
Take care, everyone.
Thanks so much. Thank you.
