Projects
![]()
No projects found
Investors
New
DAO Treasuries
DATs
Lend / Borrow APRs
Chains
Categories
Spaces
Agents Unplugged: Your Bi-Weekly AI Deep Dive #17
Recorded: Sept. 17, 2025 Duration: 1:10:51Space Recording
Short Summary
In a dynamic discussion, industry leaders explore the future of AI and blockchain, highlighting partnerships with Google, innovative projects like digital twins, and the growing trend of community ownership in AI platforms. The conversation also touches on the potential for token launches and the importance of funding in driving growth.
Full Transcription
Thank you. you you you you you you you you you you you you you you you you you you you you you
hello ladies and gentlemen what's up um i think everybody's got a speaker slot that was fast and problem less effortless this time around for change. I think that's nice. Usually we have some issues, but I'm welcome guys. Thank you very much for joining us today. We've got a pretty exciting panel coming up. As usual, I'm here with my co-host
Marco. My name is Yannick. I am the founder and acting CEO of Swarm Network. This is our
bi-weekly agents unplugged session where we mainly have a very open format of talks with some of our partners and just general like interesting people in the space
we will probably end up in uncharted terrain territory during our conversations and to all
the speakers out here today feel free to just go in any direction you feel is reasonable and you have interest in and you know
maybe just like a riff
with us today
there are no rules
like life
there are no rules on this panel as well
so feel free to
take it away with introductions first
I'll pass that over to
Marco first and then
the rest of you guys can one by one just
take the mic and drop some heat on us. Thank you, Janik. Yes, pleasure to be back again.
And this is session 17th. 17th session on the 17th of September. It's going to be a fun one.
We had some really cool sessions in the past we always go
technical but also very philosophical from time to time so i am looking forward to this one too
and i'm sure it's going to be a mix uh i on my daily i'm head of ai at oasis where i these days
mainly work with like trustless agent projects and i I don't know who knows, but there is a new ERC 8004 spearheaded also
by the Ethereum foundation,
which is making quite a bit of waves in the agent space.
So lots of cool things happening specifically also with TEs.
Yeah, whoever wants to go next, feel free.
Hi all, really happy to be on this sex basis today. This is the first one I think with
this group. My name is Matthew Ainska, I'm the founder of Me3, we're an AI engagement
platform. I'm really looking forward to it.
Hi all, this is Celia from Revox. Thank you for inviting us here today.
So yeah, for those of you who don't know Reevox is enabling AI to AI composability,
we're very happy to be here and have this conversation with you guys. Can't wait.
this conversation with you guys can't wait.
Hello, everyone.
Thank you for inviting us, Yannick and Marco.
We're super excited to be here.
My name is Edweth and I'm Chief Strategy Officer at Open Gradient.
And at Open Gradient, we're a research lab
that's primarily focused on the intersection of blockchain and AI.
And we look at solving problems in the memory space. Most recently, we've been venturing into building digital twins and digital clones
of people, and there's a lot of exciting developments that are happening in this space
right now. Great to be here and excited.
Cool. Awesome. Marco, we have like an MCp specialist here on on the call i think we can we can touch
on this a little bit later but um yeah let's let's kick it off um i will chart the course
here for us today um let's see so where shall we start I think we can start with kind of like, as obviously the panel is all building in the AI space.
community ownership of these AI, whether it's an application or a model or just a tool or,
you know, infrastructure in what form have you. The community owned aspect of this is becoming
increasingly important in the light of the trust issues that we face on a daily basis when we do interact and we do use more centralized solutions have been the ones that have been proliferating the fastest because
of the simple nature and the um the simple hierarchy that these businesses run on right
so it's okay we have one boss at the top this boss is somehow in control of many of the operations
and many of the strategic decisions goals what have you
of the organization and if this person is visionary enough and can also act this vision out in a you
know efficient manner this company will grow very fast now the flip side of that is having a very community owned and democratized
setup right so according to our panel today like does the community ownership change what the
advantage or competitive advantage look like looks like in the general sense of AI business.
Let me know if that question is formulated clearly enough.
So maybe, you know, whoever wants to go first, like just put up your hand and take the stage.
I think we can use the raising the hand and we can just like whoever raises their hand first can go first like just put up your hand and and and take the uh take the stage i think we can use the
raising the hand and we can just like whoever raises their hand first can go first and then
we'll just let everybody speak one-on-one so keep an eye on on that as you as you open your mic
let me just start.
So yeah, the thing is, when AI platforms are owned and run by their communities, what
changes is the idea of competitive advantage.
This is transformed, like traditional advantage relied on control and access right
on the other hand community-owned data community managed models they spread value more evenly and they make collaboration the new advantage, like trust, meaningful connections,
and the ability to adapt quickly.
With shared learning, these things turn community
into a deciding factor for success in the AI era.
So yeah, we are seeing a shift in this field for sure.
So how do you guys like do you guys have things like governance? Like how do you guys involve
your community in the development of your tools? And what kind of edge does that give
you specifically? In Revox specifically,
you can just,
anyone can just use the Revox Studio
and create their own agents
and customize them the way they wish.
And they can also deploy customized models.
It's a very accessible tool
for everyone to use.
And of course, we have staking mechanisms
and the next one is coming very soon.
And many ways where the community itself
can participate in the whole process of creating
and participating in the ecosystem and building.
Cool.
Let me provide some context from my perspective.
Maybe we can go a little bit deeper on this.
So when you think about the competitive advantage
of having a, let's say, large pool of participants influence,
not just governance, but um also let's say the actual function and
utility of a network if we're talking about in our context it would be for a swarm for example
right like secrecy is is inherent to successful business in a traditional sense.
So in the Web2 world, but also in just the regular business world that has nothing to do with tech.
Secrecy is often the one factor that a company relies on to be able to make their money.
relies on to be able to make their money right and this usually is generated through information
asymmetry or just a plain large information gap perhaps that information gap is due to
other companies not understanding how something is produced or not understanding how a certain
technology works or just you know by mere brute force perhaps a company has you know
in the ai sense more gpus than another company so i believe that the competitive advantage for
a decentralized ai business and it doesn't really matter what it's in in my opinion like obviously some
categorization applies and some categories will thrive in what i'm about to say more than others
but i believe that it is um largely gonna come from the fact that you can organize not just people but also capital right so if you're
talking about ai agent networks like swarms for example like what they're going to really benefit
from is to be able to outline a certain vision a certain type of niche if if you will. And then for anybody, no matter where they're from, no matter their
background, no matter their ethnicity, no matter their education, no matter whether they actually
understand what's going on or not, they're able to rally capital behind the purpose of this
either autonomous or semi-autonomous business, right?
So now where you would be alone or where we would need a team of engineers
to make something happen
or a team of traders to make something happen,
now you have all of this AI power at your disposal
and we can probably expect
that the advantage is going to come from,
okay, hey, like there's a bunch of people
that have this great idea, they either use,
I like, sorry, they either use a setup
that's been made by somebody else,
they compose a bunch of agents together
and they run them and they tie
them up to whatever protocol you have maybe swarm and then you can rally capital behind it can have
its own asset and it can become its own business or whether you know it's it's truly like a unique
thing a community driven effort where there's a bunch of developers that are working on the same
repository and then there's a model
which allows them to perhaps create an asset
and have governance on the code base, etc.
These kinds of things are going to move so fast
and there's going to be a battleground.
What I think we can expect
is that there's going to be a bunch of people
with a lot of great ideas
rallying a decent amount of capital behind that idea, deploying things really, really fast, going to
market really, really fast, and just seeing if they can hit the vein of gold that they expect
to hit or not, right? And if they don't, that's fine, because they can just move on to the next
thing, right? Whereas you're in traditional business, right? You know,
like the whole fundraising, the whole starting up, the whole collecting capital, whether you're
doing it like a more institutionalized way, or maybe a more kind of like crowdfunding way,
it doesn't really matter. But it takes a lot of effort. It also feels a lot more serious for the people doing it right because the velocity
that crypto flows at is just like a thousand x ten thousand x maybe a million times faster or
at higher velocity than regular fiat you know so when you're a builder in the web2 space a dollar
holds a lot of value and this might be a good thing but i think to be
able to chase technological advancements in a fashion that is similar to how we you know do
it in crypto it is very important for those yeah for your dollar to, I don't want to say less valuable, but flow at a faster pace
and you can go to ground zero,
you know, like,
instead of having to set up
a corporate structure
and do the fundraising
and spend like six to nine months,
you know, all of that can be compressed
into just like a few days
or even a few hours
if you want to.
Yeah, Marco, go ahead.
Yeah, I just want to clarify with
you, does ownership
need to equal also governance
for you? Because with governance,
I see so much downsides
and I haven't seen a really good example
of it in production.
Yeah, you can speak first. Go ahead.
Advait, you can speak first. Go ahead. Advait, you can speak first. Go ahead.
No, actually, just go ahead with the answer to that question, because I also wanted to answer based on that.
Okay, okay, sure.
In my opinion, it doesn't have to equal governance.
I believe that governance should be an opt-in.
And I also don't believe that governance is always beneficial.
Like it depends, right?
Look, it's in the end for at least a moment still, it's about human nature, right?
So, and not just human nature, but also let's say like person-to-person interaction which is going to
largely influence whether a project is going to be successful or not if you have a team of five
people that is highly talented highly skilled with one leader at the helm you don't need the
noise that governance is going to bring you right because you're the expert when you're leading a team onto the battlefield, you know, with a certain vision,
and you will have your way with that vision.
Now, if you open every single decision that you're making on a daily basis, or at least the larger ones,
but if you're talking about true governance, true decentralized governance,
every single thing has to be voted on, or at least every single major thing has to be voted on. That would mean that there has to be
governance voting going on every single day. Even most of the time, that's not the case.
It can still become a burden because it's like, okay, sometimes I know something is correct.
And if we do a certain thing, then we will be successful in whatever it is what we're trying
to achieve i might have my entire team against me and my community against me nobody might
understand me but i still make the decision because i've been in this situation before or i read
the um the the the uh the variables correctly and and it plays out my way. It's kind of like a gamble but I'm certain about it.
With governance you cannot have those kinds of things and that is very harmful for projects
that want to develop very fast.
Now I do believe that once you reach a certain stage then governance becomes more important
because you surpass a certain point let's say, point of critical
mass at one point where the weight is too much to bear for any individual. And then a project and
AI decentralized swarm really, really will benefit from governance. But the governance should not be
done by the community at large, because I believe that the community at large will not always, at least maybe there's something to say about it.
But most of the times, the community at large will not be better in decision making than a handful of people that are actually working on the product day to day.
They know the ins and outs.
They know the nuances. You know, like you you as a community member you're very far away from
what's actually happening like you see everything on on service level and there's value in that
sure but you don't have to have you know like like the the the governance so like if you are
talking about stock markets nowadays right previously if you would hold a lot of share
you could probably get like a seat at the board um if you're like a majority shareholder of a
company right you have something to say but nowadays there's a lot of people that trade
a lot of different stocks they are definitely never going to be a majority shareholder and
there's absolutely no governance going on and that model works you
know it works you do deploy the capital you do invest you do have a say in what's what's going
on on the financial side right you either buy or sell so you either commit or you don't commit you
either trust or you don't trust um and and that's a very easy way for you to maneuver in and out of positions of ownership, right? So I think the answer ultimately is governance is not necessary immediately. It will be at one point. But I think the next wave of opportunities and success is going to be generated by just a handful of people, but that do have a big community behind them in terms of ownership, right?
I guess it really depends on the framing of the problem.
of Switzerland. So in Switzerland since 1990 there have been 622 referendums that have been held to
a national vote and there are several issues where the Swiss population actually decides on what is
going to happen in their locality. It seems to work quite well for them. And then there's the opposite, which also works in Singapore. So
in Singapore, it is a select handful of people who make most of the important decisions.
And both of them actually do work. But most countries, when they do scale, are usually
better off having an elected set of representatives manage particular particular niche issues and then like those
people are delegated with the authority to make the final decisions instead of the whole community
being or like the whole population being involved and every matter being put to a global vote so
that's the way i see it it can work there can be switzerland's in crypto but not
every company will be successfully be able to be switzerland because in switzerland you would need
a certain level of education for all of the voters and it is actually provided and ensured
so that they can effectively vote in an educated manner and that is the problem that actually stems
from delegating too much to the community but that's what i think
right you do need to be hitler, right? If anything, the team always knows the best for the project itself. The community may sometimes not know this.
I wanted to also note, we have also seen many attempts,
many projects implementing some governance system,
some staking system to cover for other gaps in development.
Like, because the community sees governance
as a very important thing to have in theory.
Sometimes it is not organic. You have to only implement it
when it's the right time, when it's organic, when there is a real need for governance.
Yeah, that's exactly the reason why I asked the follow-up question, because i'm definitely pro community owned but also pro small committee or community
governed um reason being that once this community owned you can still give users or the community
a chance to impact decisions but it shouldn't be via governance it should be via however they are
contributing to this community this could be like really financial so if there's
for example a token if they are not happy with a decision they can of course just sell like they
vote with their wallet and what i really like are these approaches where you have these small
committees deciding things but the actual just kind of think of a board of directors but the
voting happens in public and before something implemented, you can have an implementation delay.
So let's say like there's a week to review any decision that has been voted on.
So it is going to be implemented on.
But if you're unhappy with it, feel free to walk away.
Feel free to withdraw your contributions, however they were, and then find a project
however they were and then find a project that actually aligns more with this and just kind of
that actually aligns more with this.
to add something new that i wasn't too aware of but i had a call today with someone that brought
it to my attention and that's futuaki i don't know if you guys know this like this new or new
it's this governance mechanism that was introduced in like 99 and also picked up quite a bit in
crypto and the idea is that you go away
from traditional voting on things,
but actually have kind of prediction markets.
And those decide, like you predict on,
is the project going to be better off
if this is implemented?
And then you can also benefit from this prediction.
So yeah, I don't think I explained it properly,
but there's many ways to implement this. And I know that in crypto, quite a few are working on this. from this prediction. So yeah, I didn't think I explained it properly,
but there's many ways to implement this.
And I know that in crypto, quite a few are working on this.
I know Optimism and Unichain are working on some things.
So fascinating things to improve on decentralized governance.
I do still prefer the kind of small committee style currently.
Yeah, so what I was gonna say is like,
it's imperative to identify what
should and should not be voted on right and that is um depending on on the context you know like
in what stage is this um swarm or or company developing right like are you still just in like the mvp stage or are we like post
product launch like like where are we on the on the road um because like for example doing these
implementation delays it's it's great it's it's great but it also slows you down a lot right
so if there's an opportunity for you to
capitalize as a company and get in like, let's say half a million dollars in profits, if you act
today, but then you have to wait and you cannot do so, right? So that's why I was saying earlier,
you know, like at a certain stage, the certain size, like it becomes more important but there are so many different types of models
that you could follow right so let's say you could go more on like a political um in a more political
way you could say like okay well what can the community vote on well they can vote on the board
members they can vote on the leadership and like at a certain stage in time you know like every
let's say six months or every year
there's a vote to see you know if we're still happy with with with the people that are running
the ship something like this I think could be very valuable and could be very interesting
especially now that we're talking about AI agents becoming much more capable in running a business right so it becomes much more about
strategy vision and technology and there are going to be probably less variables to control
and to take care of once we start really kind of like thinking about like autonomous businesses just purely run by
agents and and a set of people so there may be there is an opportunity for mass
governance so I think it also really depends you know like what did you start
off with and and and how do you run this because i can see also like swarms that are fully autonomous
right like there's nobody that really does any leadership or does any direction other than
the community and that can be done through some kind of complex model of governance yeah sure
there's a there's a lot to that and it's going to be very complex and it's going to
have many faults right because
we've all seen what has happened and the many failures of decentralized like really decentralized
organizations on smart contract level right there's always flaws there's always problems
there's always things that go terribly wrong there's always money that's lost and uh yeah
it's an interesting thing to think about actually i can
shine some sorry shine some light on the uh future arcane because we've been developing prediction
markets in the past and prediction markets in and of itself are a very useful tool not just just for the sake of having, you know, the prediction, but there's so much to it.
So it's kind of like a vote of confidence.
If you're talking about, let's say we create a prediction market within a project that talks about,
okay, if we're going to make XYZ business decision,
what do you believe the likelihood of this decision becoming like the
decision that leads us to the successful end of this project for example right like maybe you can
be more specific let's say if we go for car design a or car design car design b which car design is
going to yield better sales results right so now if you create a prediction market within your
organization let's say you are for it create a prediction market within your organization, let's say you are Ford,
create a prediction market within your organization, you give your employees funds from the company
or you oblige them to use a certain percentage of their salary, for example, to participate
on these markets, they're going to be that serious about it.
They're going to be super serious about it because it can mean them benefiting perhaps in a substantial way
if you are one against a million right if you're one against 10,000 people in
your company and you win out the like the benefit the upside of that is it is
great and so you give people essentially an anonymous voice in a very arbitrary way, in a very non-direct way.
So you don't know why people are voting, but you do know that there are some wisdom in it because everybody is serious. where you can pose questions or pose governance proposals that are directly related to a certain
type of outcome that the project is interested in, the community at large is interested in.
And then you could have the community use their tokens to participate in the prediction market.
use their tokens to participate in the prediction market and then if you set it up in such a way
that the leadership of the project or the agents have to listen or have to um have to adhere to
the majority or the highest priced outcome or option then you could run a project that way but yeah so it's a very indirect way of
of governance in a sense because you don't know what why people vote in a certain way
and you don't have a line of communication you don't know who's voting you just know
that there's you know like the value and the size of the uh liquidity like how many people are
voting yes how many people are voting
no and it also doesn't really matter like if it's one guy that puts like a billion you know tokens
or there's like a hundred people that put like a hundred thousand it doesn't really matter um
because you only look at the value of a certain outcome let's say like what is the what is the
value of yes what is the value of no at this time
but adding to that like you cannot see it as a regular prediction market because it has kind of
it has to be like a perimutual prediction market at maybe not but like at one point the market has
to close pre-outcome because otherwise people will just keep readjusting their you know their their um
their bets essentially as time goes on and then you still have no verdict
yeah i think we're all quite the core on governance mechanisms and also what is preferred
at what stage of a company a project uh just kind of to paint the picture a
bit more regarding what I was thinking it's like I think most of you saw this Albanian AI minister
that is now supposed to be ending corruption I'm sure we could talk just about that one for an hour
but the idea is you have an AI agent that is now interacting with these grant applications to distribute government funds
and yeah since it is ai supposedly it should end corruption and what i was talking about before
like i don't want governance for every single decision of the agent that's of course horrible
i'm gonna delay everything but anything that impacts the actual decision making so a new model different weights that should be
governed publicly and i should have some form of implementation delay so that i can react to like
these major changes just kind of to have it a bit more tangible to the listeners
yeah yeah i agree i agree and the prospect of having a having a AI governor is very interesting too, because you remove the emotional dimension for how much that's possible. From this chain of decision making. But you also introduce a new type of bias.
You know like model bias obviously.
But arguably the emotional bias.
The native human bias if you will.
Is more harmful to the process than hopefully the LLM bias.
But we don't know that.
But that's what I would put my money on.
Yeah, I actually met with Vitalik at least two times.
And we talked about the importance of prediction markets.
And especially AI-driven prediction markets.
especially AI-driven prediction markets.
Behind the scenes, we've been working on this for a while.
Behind the scenes, we've been working on this for a while.
And I see a very interesting future also in the governance for AI systems themselves,
where it's not the community that uses prediction markets to make decisions,
but it's AI that uses prediction markets to make decisions but it's ai that uses prediction markets to make decisions and we actually have run
tests on this and we might publish something about this in the near future but it's super
interesting to see because you know you have these hyper frequent and hyper rapid um decisions that
frequent and hyper rapid decisions that like a swarm of agents is making on almost like a
like you know like a at a cadence of let's say every 10 to 20 seconds and it's interesting to
see how LLMs and agents react to them having to use a asset as a vote of confidence in an uncertain situation,
it changes a little bit about, it changes how they interact and how they reason,
especially if it's part of like the initial prompting.
So yeah, it's an interesting segue that we've made because we're building the truth protocol
and everything is going to
be AI agent powered. So we've been struggling, to be honest, with the governance side of things and
the decision making things. We wanted to make it as transparent as possible for our community
for them to understand, like, okay, if a claim is verified or falsified, like, what are the agents
actually doing? Like, how are they reasoning? But the reasoning alone is not enough.
There also has to be some kind of governance going on.
And that governance should preferably be a network-wide governance,
if we're talking about AI agents, because that makes it less subject to bias.
If you have five agents voting versus like a thousand agents voting
or predicting a certain thing, yeah, it makes a big difference.
But anyways, going a bit on a tangent here,
here i want to switch up the topic because we are halfway already i want to talk a little bit about
I want to switch up the topic because we are halfway already.
um custom versus off-the-shelf agent solutions so we've seen a lot of good
off-the-shelf solutions for building agents and off the shelf i'm also talking about frameworks
etc it just it doesn't have to be you know like a UI-based drag and drop or canvases.
But we have also seen that the majority of the success that companies are seeing in general,
also in Web3, comes from very unique, creative, and custom-built agent chains or agents or models.
So I wanted to know what you guys think about this.
And maybe you have some alpha from your end
or you've seen some cool things.
Just feel free to speak your mind.
Elia, why don't you go?
Because you guys are handling MCP.
I think you will have some insight.
Right.
So, yeah, custom agent solutions are tailored for specific needs, right?
right? They offer deep integration and flexibility. Now, off-the-shelf agents are
They offer deep integration and flexibility.
surely quick to deploy. They are cheaper, probably, most of the times. And they are
generally thought of more reliable, or they at least were initially thought of more reliable or they at least were initially thought of more reliable because this
is changing as we speak we we are seeing many developments in this field right now
we are also working towards this
this in this field now
in this field.
Now, when it comes to traditional software,
it is indeed fading somehow.
And this is something we will be seeing
over the next few months more because agentic networks
use autonomous systems that can adapt, that can connect across platforms,
that can work without central control. This is very important for many reasons.
Like the user's data is safer, the agents are faster, they are smarter for sure. And yes, they're often more resilient
than old school software tools for sure. Yeah. So yes, they are more flexible, they are easier to
use and customize. I totally believe we are heading towards a future where autonomous agents will surely be centralized software as a service services.
Let me take the consumer side of things since I've been thinking about it for quite a while now recently.
side of things since i've been thinking about it for quite a while now recently um for me this was
always the fascinating part and this i mean like customization whenever i was trying these like
fun ai tools like just think about song generation like actual music songs or poems like that's what
i initially used uh llms or any kind of interface with AI for.
And the customization part made it like really unique.
I think what, and now I'm giving advice, a very good chance to shill a bit.
I think a big challenge for these customization setups is that you somehow need to give them quite a bit of input.
And this is super annoying.
Whenever you start from scratch you have
a new tool and you kind of have to explain them tons of things about you so what is emerging
recently and i've i know now like quite a few projects working in the space it's a portable
memory like how do i have this one memory bucket or one kind of place with multiple buckets that hosts and control or not control
just like store data about me and then i can selectively share this with new applications
so that i have a very very custom experience and not just like it knows my name but actually it
also impacts the quality of the answers so i think the more custom it is the better from a
user perspective and for corporates like similarly like the enterprise solutions will always be
customized and very similar to how sas already is for smes i mean it's gonna depend on the cost of
customizing something at the moment it doesn't seem very high, specifically with all the LLM agent things.
But let's see.
Thank you, Michael, for giving the opportunity to show my work at Memsync.
So what we learned is that all of the agents that you use as a consumer are eventually going to end up being super personalized.
to end up being super personalized.
And essentially what we're building is,
we're building a backpack which you can carry
that stores all of your memories
that you build across all of your interactions
across ChatGPT, Claude, Grok,
and every AI agent that you ever speak to.
So the goal here is to make sure
that you have one unified memory layer
across all of your AI applications
so that you don't have to constantly keep explaining
and giving context to these new agents
that you interact with.
The way that we are seeing the world evolve
is first, we saw a lot of development
in the general agent space.
And now as a progression of the industry,
we're seeing a lot of verticalization.
So you already see this
in the most recent YC batch with an N number of startups that are building cursor for X,
cursor for Y, like for example, cursor for video analytics and so on and so forth.
Every time you interact with any of these new agents, what you want to do is you want to
make sure that they know who you are so that
you can pick up exactly where you left off. And that's exactly the future that we are building.
And we truly believe that you will see not just customization in the personal level,
but also you'll see a lot of customization at the enterprise level. So as we speak, actually,
we are partnering up with one of the biggest blockchains to build
digital twins of their employees. And so essentially, what we're doing is we're taking
the workflows that all of these employees are responsible for, and making custom agents that
actually represent each of those employees. So the goal here is, even if let's say your
coworker is on holiday, then you should be able to get all of the contacts that you require to be able to progress on your project.
And you shouldn't really have any hindrance in terms of context transfer, which is actually a huge pain. putting developers or new developers into an existing project. The biggest problem arises in the first month or set of months,
where they require a lot of context transfer and they require
a lot of explanation of what has happened already in the code base.
This is the problem that we're actively looking at as well.
But yeah, as you move forward,
there'll be a lot more customization and personalization of
all of the agents that you interact with in your everyday life, for sure.
Awesome. Awesome. Cool. Let us, let us, let us, let us, let us move on to another topic, perhaps.
Marco, was there anything that you wanted to bring up before I switch up?
I'm also happy to get on topics from everyone else here.
But as I said in the intro, kind of what interests me for the past few weeks that I've been looking at is this ERC 8004,
introducing kind of add-ons to A2A from Google, which we've discussed a couple of times,
and it's building on top of it.
And what it does is pretty much it just introduces three on-chain registries,
one for identity, one for reputation, and one for validation.
The idea is to make agents more discoverable, cross-organization,
and also, of course, targeting more the Web2 use cases,
but they should just use any chain pretty much.
I mean, it's an ERC, so any EVM chain to register the agents
and provide some form of trust and validation that the agent is
executing for example in a te but it is very cool because it's such a simple and open standard so
really everyone can participate contribute you can just kind of adjust it to your needs
and it's fascinating to see like how many people are actually building in the agent space, because this was, yeah, it was spearheaded from the AI lead at the Ethereum Foundation and the AI lead at ConsenSys.
So there were some bigger names behind it, but the community just picked it up.
Like everyone in the agent space suddenly got interested because everyone is working on some form of these three registries, identity,
reputation, and validation. So it was easy for everyone to contribute their opinions and now
build example agents, demos, et cetera. And I'm just loving the community-driven building process.
And yeah, it's been a very cool experience the past few weeks.
Yeah, I think it's great. Like we've talked about this on some of the very first spaces where we
were talking about yeah we definitely need some like open standards that it's just foundational
things like how the internet was built back in the day right like all of the standards and all
of the protocols that the internet is built on are super open they're um they have just one thing in mind,
which is just like free interaction
between the units on the network,
the nodes on the network.
I think this is a great start for the AI agent space
and I hope that people can take it seriously
because I've seen that there's a lot of kind of like
PVP mentality in Web3 when it comes to like, hey, we're building this solution for that. that people can take it seriously because i've seen that there's a lot of kind of like pvp mentality
in web3 when it comes to like hey we're building this solution for that you know like you actually
have people that are building like a chain specifically for whatever ai agent discovery
and you know they are trying to um to build these registries and protocols in a silo and then that's just never gonna work
it's just not gonna work because that's first of all it's so against the nature of blockchain
right and then and then beyond that it's just it's just stupid because like how are you going
to convince all of the builders to just come and build on your chain right it's just another grift so yeah like to have the ethereum foundation work on this and now you see like the you know google
is doing a bunch of stuff um i think slowly steadily we're gonna we're gonna get somewhere
but i do think that the ai agent capabilities are going to exceed the ability for them to be able to
organize themselves in a more decentralized and a more peer-to-peer
fashion just simply because people are building more siloed solutions faster
and it's easier to do so as well.
Right.
Yeah, and I can just...
Go ahead, Elia.
Sorry.
I just wanted to add that on one hand, all this overbuilding also creates a bottleneck
because actually people have to take more steps to get there.
people have to take more steps to get there.
Often we often make things more difficult
instead of easier for people to use these products.
But on the other hand, all this building, I think,
will eventually lead to a few good things that will stay, that will manage to stay in the
long run here. Those who, of course, do the best work in all those fields that you guys
mentioned. And yeah, so it's like two sides of the same thing, which is together a bad and a good thing.
But I think it will lead to enough developments that are worse.
Yeah, and I just wanted to double down, Yannick, on the necessary standards. I was on a panel last week with the technical director of Bosch, so a massive corporate in Germany.
And we were discussing agents and kind of the use cases.
And for him, like the whole compliance and standards part was crucial because he needs to be able to explain this on a very deep level on what does this agent do?
Where does it come from
what what can it do what are kind of the restrictions it was so different to like my
normal crypto use cases where i kind of just want to limit the impact of bad actors um so yeah this
is definitely a step in the right direction no yeah for sure i mean like if you're if you're talking about you know large corporations
and institutions being able to even start to deploy agents um at a serious level you know
where there's decision making going on and there's autonomy in terms of um what actions can and cannot be taken. This is like very, very risky at the moment, right?
Because it's impressive that a guy like that, you know, from Bosch is able to think about
this because most of the people that are running large businesses, they're not even thinking
about these kinds of things, right?
Because it's just too black box for them.
They're like, yes, AI agents, like high potential, but, right, the risk's just too black box for them like yes ai agents like high potential but
right the risk is just too high for them now because if you inject one agent into a crucial
pipeline other than customer support or other than like non-vital, simple non-vital automation,
like automated emails with personalizing those more
or perhaps handling some kind of registration
of like if somebody orders in your web shop
and they want to return the product
to automatically send like the labels
and this kind of stuff, right?
Like not much can go wrong there.
But once you start talking
about you know maybe even things on financial level right um or compliance level or these
kinds of things it's like there's just so much risk for them because one small mistake could
lose them millions or maybe even billions of dollars and they just don't want to take that
risk and they can also not justify taking that risk
to their board of directors and investors
because indeed they are unable to explain
like what, okay, you know, how is this safe?
Like why is this safe to use?
How are you sure that this is better
than using a person and so on and so on.
All right, so there's still so much room
for really like agents to develop
on the internet um as how websites did you know what i mean it's like before
before commercial internet that was just like people internet right people were on the internet
they were organizing themselves it was basically like a free-for-all
massive forum that was compartmentalized into smaller networks that hosted people and that
had like a central central point of of of information that is dial up a pbs like bullet
board system and modem infrastructure that the that is the true origin, PBS, like bullet board system, and modem infrastructure,
that is the true origin of the internet.
I think we will see a similar fashion of development
when it comes to AI agent interaction,
composability, discoverability,
and that is just going to slowly and naturally evolve.
But the cool thing in a way is that i think we just need to make
the standards and then ai will kind of do the rest you know it's like okay we have a new standard
here's how you how you just you know like make sure your your agent adheres to that and then
if that becomes a standard and it's a, just like a, you know,
like a very simple add on, you know, like of add on type of thing, or like just a, um, a code base
that you can easily maneuver or that the agents can easily maneuver themselves. It will grow so
fast. And we will, we will see so many issues with the foundation of these things that it's very easy to
progress and develop them so that's pretty cool um yeah go ahead bro go ahead that's right
the way that i see it is uh so if you were to replace people that are providing a lot of useful input by parsing a lot of context themselves. That's going to be a very hard problem
because these are people who have been delegated
with the decision-making.
And so essentially, like, if you want to replace people
and automate their processes or what they do,
the best place to start is obviously, for example,
in supply chains or construction or places where you don't
necessarily have to choose between like X or Y too often, but instead there is an exact
approach or an exact formula that you can follow. So those are the best places for automation today.
In the future, the way that I see it is it's more likely that we see AI-driven
organizations that are built from the ground up instead of business units being replaced in
corporations. As you mentioned, they're never going to get the right permissions to completely
eliminate people from, let's say, the sales team or the marketing team. But instead, we will see like
AI driven organizations rise where there's probably just one person or like maybe a maximum
of 10 people in the organization and thousands of agents that are working
with them in different forms to fill different kinds of job functions. And that is what I'm
super excited about.
So the number of people that is going
to be needed to create massive value
is going to drop as we move forward.
We'll see a lot more plenty of fish outcomes.
So for example, plenty of fish is a dating site that was
built by a single individual that he scaled to 100 million
ARR and then was able to sell for
965 million dollars alone.
So we'll see a lot more of plenty of fish outcomes where one man is able to orchestrate
an army of agents that are each fulfilling different kinds of job functions in his organization.
in his organization and these will grow immensely valuable as we move forward
And these will grow immensely valuable as we move forward.
yeah that is just some insane metric you drop right there right there's like a one-man business
that is sold for a near billion dollars that's just that just seems so far out there but
it's true that you can capture so much value with a small insight and a set of hands.
And then now you can use that set of hands to direct, you know, like what is virtually like 10,000 hands.
As long as you control the quality and the outcome and enter to your vision and you do the thinking.
You know, like you can go so far.
You can build apps. You can do so many things but what i'm super excited about is is once this
like hyper proliferation happens in the information space all nooks and crannies are literally going
to be filled with solutions and everything is going to be super cheap democratized and there's
going to be very little competition but then where there is going to be a lot of competition is like ai agent
powered autonomous organizations that actually operate in the real world think of like fleets
of delivery trucks um so like autonomous um like uh autonomous logistics, autonomous shipping logistics,
autonomous air taxis,
autonomous drone deliveries, etc.
These are the things which I'm super excited about
because they're also just going to be meant
by just a few people behind the laptop or behind the computer,
probably even
using natural language uh for them for most of the time but they're going to make true impact
in the real world in our physical world you know like uh fleets of drones that do window cleaning
for for for skyscrapers this kind of stuff i think is just gonna be so fucking like booming and if you're early if you're early there
and you can you know make yourself present um i think that's you're gonna do some serious damage
in terms of like people losing their job obviously but it it is going to make way for
a very efficient and bright future. I'm sure of that.
I know we're almost out of time,
and I don't know if otherwise it has a point relevant to the topic.
I just want to remind you, Yannick, too,
let's do some viewer questions maybe today.
I do need to run in a second,
but I know I promised someone in the comments on your swarm
post that we would again have
some listener questions so
maybe we can do at least one
sure yeah anybody that wants to
come up on the stage and ask a question
just request a speaker slot
and I will let you speak
it should work today
don't be shy we still have a few minutes
meanwhile Adria do you maybe want to as you speak. It should work today. Don't be shy. We still have a few minutes. Meanwhile,
Adria, do you maybe want to
speak your mind a bit on
something that interests you?
Actually, the topic that we were talking
about is super interesting.
So over the past couple
of weeks, we've been building digital twins
of people and we've been
trying to see what kinds of problem spaces are they more suitable for.
So you can actually check it out at memsync.ai today and you can have interactions and it would create a digital twin of you.
So essentially when I ask my digital twin, what kinds of things would I like? It actually has a fairly good understanding.
And even if I ask it fairly complicated questions such as like, what do I feel about gun laws in
the US? It can mimic what I would think incredibly well, even if it doesn't have context on the same
subject. So we've been exploring like using these twins in different kinds of settings. And what we found is that digital twins of people actually perform incredibly
well in environments where there is no fixed answer. So let's say that you have a problem
space where like multiple possible solutions could be the right answer. So for example, in fundraising, in hiring, in operations, like there's a lot of like
nuances and a lot of problems where there are multiple possible correct solutions.
So I think that as we move forward, we're going to see a lot more development into this particular area, which is being able to solve not just problems with fake solutions, but also open-ended problems by mimicking characteristics of humans.
And I'm actually really excited for this world.
That sounds very interesting, too. I also have some papers that I read that are about scientific discovery using AI agents, but also using digital twins of academics.
academics, more so as kind of like sparring partners in terms of discovery process.
So like every scientific breakthrough stems from a observation and that observation turns
into a discovery, right?
And there's just so many discoveries to be made off of human realizations and usually
the cadence of major breakthroughs is relatively slow and we're now in this cycle of like okay
somebody discovers something nobody believes this person maybe 40 50 years later this thing that he
or she discovered has been sufficiently peer-reviewed and other studies have been
done to be able to replicate it and now it starts to make its way into our lives right like this is
a an interesting trend and there's this kind of like bottleneck that academia poses because
some people will defend their objectively incorrect ideas and ideologies to death.
And only after they die can academia move on.
So I think it is very interesting for us to be able to use these open-ended,
yeah, like navigate the open-ended problem space
and have like sparring partners that are
just much faster than us and yeah it's it's very interesting but uh we have a wise hey hey bro
how are you doing welcome back on the space like go ahead and ask your question hey what's up
i know i just had some comments on some of the things you guys were talking about earlier. You were talking about AI and government. And I believe things that are mathematically based, like some things are calculatable with math. That should be easily delegated to AI. When it comes to like morality, opinion-based things, there definitely should be some human consensus.
But there should definitely also be AI involved in things that you can calculate like the economy.
Because there's a lot of fear about AI taking over and this and that.
But AI does not have temptation temptation it doesn't have greed
it doesn't have the flaws that humanity has and yes there is the training data but if that was
made open and open source and every person can review that then that can be a huge difference because not everything is just based on opinion some of it's
just easily calculatable and then another thing that um out of it was saying about having a memory
pack that i believe is very important as well because i have have been using ChatGPT for like over a year now.
And I actually made a database out of all of my chats
and turned it into a SQL server, a JSON file,
and my current ChatGPT can access
all of my previous conversations
to get data from directly a lot faster.
And when your AI assistant knows you that well, knows what your goals are, it really
helps change how you interact with it.
It's very helpful when the AI actually knows what your goals are, what your situations are, what your current projects are.
So those were just two things that I wanted to touch on that I heard.
And definitely excited for Swarm Network
because I heard the message from Google and Sui Network.
It seems like everything's coming together.
Yes, thank you very much for your insights.
We got Elon Muskie.
Elon Muskie, you may ask your question.
Actually, Yannick, there is somebody that you know
that I need to speak with you about
as far as a partnership goes.
I don't know if this
is the appropriate time. We can talk at the end of the space if you'd like, or you can message me,
but I didn't want to take up too much space because you guys are talking about a specific
topic at the moment. I just wanted to let you know I do need to speak with you.
That's all right. You can DM me on my personal account and then I'll reply to you. That's fine.
me on my personal account and then i'll i'll reply to you that's fine thank you sir and then
we have a crypto biz um crypto investor analyst and games and web go ahead ask your question brother
or sister
Hi, it's a bit.
Yeah, it's, it sounds like you're talking to a broken phone.
So I, uh, I, I, I, what security is that your project used to protect the user?
How secure your smartphone is against you? Yeah.
That's too bad. We don't have any other requests and we're already eight minutes over time.
But I want to thank everybody for joining this space today.
It was a delight to speak with you guys.
And I hope to see everybody soon
on another space maybe in a
few months we can do a catch up
and we can talk about what
has changed since we last spoke
which would be today
and to all the listeners here
in the space today
I see a lot of you guys are commenting
when TGE
obviously we're not able to let you know when TGE happens until it happens.
So there's no use to keep asking this question over and over and over again.
The exchanges will make a post first and we will repost those posts and it will happen soon enough.
Yannick, I have a question.
Besides the TGE, do you have any swarm network
alpha to give the people because i'm excited about it as much as i said yeah i can tell you a little
bit of alpha it's actually regarding the topic that i spoke about earlier um the team is experimenting
with some sort of prediction markets for agents,
which we hope we can bring to you guys to try out
as part of the agent builder.
And yeah, we have a bunch of great partnerships in the pipeline.
We're doing some things with Google,
which we will publish probably a little
bit after TGE. But yeah, we're part of the wave, if you know what I mean.
Yeah, I definitely saw it coming together the way that Sui Network just posted about Google.
And you guys had originally had the Google, not like a direct grant, but it was part of the program,
like startups. And it looks like it's all coming together. So definitely excited. And, you know,
most people are asking, like, is the team like really uh making things happen and i'm seeing things happen so
you know it's great to hear that those things are coming along and adding that for the builder
program for prediction markets sounds amazing that's definitely something i would want to test
out for sure yeah i see that we have a bunch of like core community members listening uh to us
today i also want to just mention that a lot of the people that are complaining about TG and so on,
they're airdrop hunters, and we don't have to spend too much time worrying about these kinds of people.
We're here to build the long term.
But anyways, guys, thank you for joining and see you on the next one.
Thank you, Yannick.
Thank you, everyone.
Thank you so much. much bye thank you for organizing
