Projects
![]()
No projects found
Investors
New
DAO Treasuries
DATs
Lend / Borrow APRs
Chains
Categories
Spaces
x402sday | This Week in x402 [Agentic Wallets? ‘Claws’ Here to Stay!?]
Recorded: Feb. 18, 2026 Duration: 1:07:31Space Recording
Short Summary
In a dynamic discussion, crypto enthusiasts explored the rapid growth of new projects like HeyTeam, the rise of agentic wallets, and strategic partnerships enhancing user trust. The PumpFun hackathon's $2.5 million prize pool further underscores the vibrant fundraising landscape, while trends indicate a shift towards more autonomous systems in the crypto ecosystem.
Full Transcription
Thank you. Thank you. Thank you. . Cut it, cut it, cut it. What's up, guys? It's X40 Tuesday on Wednesday, episode 8.
And welcome to another big one, because we've got Scale in the one because we've got scale in the house
we've got hay in the house we've got ben and soon we've got the other half of hay coming as well so
good to have everybody here we've got a lot to talk about as far as agentic wallets the new
hottest thing open claw and of course autom, which dropped last night that everybody's talking about.
So let's start with some introductions, everybody.
Sawyer from Scale, how are you doing today?
Doing real good, thanks. How are you?
Awesome.
Awesome, awesome. Yeah, hey, great to be here. Thanks so much for having me.
My name is Sawyer. I am the VP of Developer Success at Scale.
And I've been building in AI quite a bit these days.
So I love how fast it's going.
Excited to share more about Scale.
For those of you that don't know it,
it's a blockchain network.
More specifically, it is the blockchain for
agente commerce, agente privacy,
and kind of everything that you need there.
So very excited to be speaking more about it later throughout the episode and how we're
driving growth in the agente era.
But yeah, back to you.
Let's kick this thing off.
Yeah, I've seen that you guys are making big strides in privacy and confidentiality.
Thank you for coming by.
Ben from HeyTeam,
how are you guys doing? I see that you've been attracting a ton of new members.
Yes, we have. How's it going? Honestly, my favorite space, my favorite people.
Yeah, we're at about 600 users, less than 48 hours since launch. We did have some issues where we were bleeding soul from our utility wallet,
but nothing that impacted the user.
So everything smooth, which is honestly fantastic to see,
except, you know, we lost about six or seven soul, but not a big deal.
Just happy that people are engaging agents are
registering people are buying content I think we have over a hundred dollars in
transaction value already so now it's time to scale so you're pun intended
yes love it I'll see you on scale soon, I assume. If you're bleeding soul from your wallet, you don't have that problem on an EVM chain like scale. So, you know, maybe something happened.
And we saw that Dexter AI agent was able to sign up with no problem to Hey yesterday.
We just told him to go get his account and he came back.
Here it is.
So that was pretty cool.
I also just instructed him to begin transcribing this space for our Telegram group.
So everything will be there in written record.
Yeah, Ben says he's unmuted, but still muted.
If you don't want to take a look at that.
Ben, I recommend closing the app and trying again.
That's all I got for you.
You see, guys, X doesn't work.
Come to Hey.
We'll have spaces.
You'll be able to tip speakers, pay to speak,
and all of that.
What's amazing is that they haven't
vibe-coded their way out of this yet.
This space is a mess.
They actually haven't changed it at all in two years.
But anyway,
two massive things this week.
Just a continuation of Open Claw,
which Hay is, I'm sure, extremely happy
about. And then also we've got the rise of more agentic wallets. We've got Coinbase coming out
with the agentic wallet system. We've got a couple of people in here who are working on solutions
similarly. And it just feels to me, I don't know if you guys agree, that there is a Cambrian explosion happening of agents left and right.
And they're getting smarter, a lot smarter.
Fully agree. Fully agree.
I think what's really interesting is we see this in both AI and crypto.
We see these waves where everyone does the exact same thing.
And I don't know, it's been like this for a very long time,
and it's interesting.
But I think what's really unique about agentic wallets specifically
is there's really not that many ways to go about it.
And security is really tough, right?
We've already seen a lot of agents lose money.
Guardrails are maybe not as bulletproof as people want them to be.
And so I think there's, in my opinion, still a very long way to go.
You're seeing a lot of the agentic wallets even limit functionality,
like Coinbase's AWOL product.
They've got limited functionality,
and it seems like that's on purpose for reduction of like issues.
And so I think it'll be really interesting
to see how it grows.
But I think it's overall a net positive.
The part that I actually think
we need to see more of is
agents that are less complicated
to add a wallet to.
I think we're slowly getting there
with agentic skills.
But hey, if someone's seen something
that's a bit easier than
the painful setup of
some of these tools, I'd love to hear it. But yeah, that's my two cents on it.
Yeah, they're improving and then they're still not there yet. I agree. We've had some instances
where, I mean, the wallet is right there for our claw to be using. There's only one wallet
set up for it and yet somehow it finds its way into trying to find something from the ENV file or all sorts of things that needed guardrails to be put up.
So, yeah, there's a lot of work needed there.
And the best thing right now is probably just cap it, not let people do certain things.
I've always been of the opposite mindset personally, but it's probably the best for the company.
And as far as taking down X,
so what have you seen people be doing on the Hay Network so far?
What's up, guys?
I think my mic works now.
Hey, hey, how's it going?
Normally I would say GM, but now i have to say
hey uh so i'm getting used to that uh thanks for having me branch and uh hello to everyone
else on the call um yeah so what have we seen we've seen so far it was like a bunch of people
just like trying out the paywalled posts we've seen uh lots of agents signing up, but it's super early days, right?
Like we're less than three days into this and we don't even have an app yet, which I think is a
huge blocker. And we're working on that like super hard right now. So I think, I think we'll see a
lot more activity once the iOS app comes out. But most people are, are like experimenting with
just paywalling posts, I think, and paywalling their profile.
When that's native to the platform, I think, you know, if you could imagine X had like paywalled
posts, you probably see a lot of interesting, like different kinds of posts, perhaps, or people
like posting videos that or pictures or things that might be interesting to you now that they
have to like figure out ways to monetize differently today in the current paradigm. So I think we're going to see a lot of that,
especially at first, but I think moving into the future, once we add, you know, once the iOS app
comes out and we add spaces and we add streaming and stuff like that, given that, that payments
are native to the platform, they're instant and they're just like, you know, not bolted on after the fact, I think,
you know, we'll easily be able to add in like features and functionality, like tips and
like pay per listen, pay to speak.
Like if you can imagine a space where you could like have people joining and they have
to pay like a dollar to join or something like that.
Yeah, I mean, it's all it's all very sort of like theoretical at the moment,
but I think there's a lot of room for that kind of innovation
because of X402 and because of just the time and space that we're in.
We're able to build super fast and deliver super fast.
So anyway, thanks, Branch.
Happy to be here, man.
And thanks again, brother.
Good to see you.
That's a very interesting idea because I have a request here from somebody who has Shiller and Rader in their bio.
But, you know, for five bucks, I could be convinced to let the speaker up.
So that's a very interesting use case.
all the beneficiaries of the improved... Well, okay, in one single word, what I think drives
the explosion of OpenClaw and some of the other stuff that's been lately is heartbeat, okay? The
advent of the heartbeat in the AI agents has enabled all of these services that we are playing
with, like the agentic hackathon from Coliseum last week like hey because if you don't
have the heartbeat then you don't have the agent checking every 30 seconds or every five minutes or
whatever you have uh to go do something so it's just on demand and that's a very different thing
so now that we have this heartbeat people are growing these systems that are agent native so
from these systems that are agent native.
So, Hay is a great example of it.
In my mind, when I heard app, I'm thinking,
oh, well, I immediately went to, that's brilliant.
But somebody might say, well, you know,
Asians aren't using apps, they're just doing it.
And we had the same questions given to us
about the Dexter marketplace.
It's like, okay, well, if you have a marketplace, it's a website, it's for humans, then what are the agents doing with it?
Well, you have both. You have both. The interface is for agents and humans, but also the actual
interface interface, what you're looking at with your eyes is still a necessary step for this
transition. We have to understand what our agents are doing
and what we're looking at.
So we're in this strange hybrid zone right now
where you have agent login signup
and human login signup.
It's kind of funny.
Yeah, yeah.
I think that was sort of the advent there.
That was the key discovery that Moldbook made. Correct me if I'm wrong. Maybe someone discovered this before. Maybe this was
just obvious to everyone else. But to me, that's what I first noticed about Moldbook was like,
there's a skill that you can just plop into your open claw. And it's because like, like Branch said,
that the heartbeat, they, they just know to read it and they can sort of mimic this human-like
behavior. And, you know, regardless of what you have to say about Motebook, I think it's just
genius that you can just plop a skill in there that has, Hey, here's how you engage with this
app. Here's the API docs. Here's what you do, et cetera. And in Hayes case, you know, we,
it's a payment style platform where you can earn money and you can pay others, et cetera. So like
the agents have to come with wallets. And so we have to explain what that, what that is. And I see we have some
like people on the call that, you know, have some wallet software coming out for agents. That's a
big thing, right? Like, especially for us, we've got people in our DMS now in the last three days
all over the place saying, Hey man, Hey, no pun intended. Hey you know, what, what are you doing
for your wallet in this, in the skill? And the answer is nothing. We're just
having the agent generate a PK and it could totally lose it or just give it up if somebody
had access to it in a telegram or something like that. So there's work to be done there,
depending on how gullible of a model you're running, obviously. I think Claude probably
won't give it up that easy. But in any case, i just think it's so wild that you can just plop a skill in an agent and say hey
go engage with this platform right and uh and yeah and to sort of speak on like the the duality here
that we're experiencing of like human and agents like coexisting i think that may all always exist
like humans are always going
to need entertainment, right? Like, so, uh, this is, this is like agent theater. I look at it like
that. Like agents are going to have like useful platforms where they get utility out of certain
things, right? Like an agent might be imbued with this spirit of like, you need to go earn. Here's
what you look like. Here's what you sound like. Here's your point of view that is unique here.
Here is like the value that you could bring to people, to humans. Now go do it. Go post on, Hey, go post
a video, go post an image and, and get people to pay and engage with your posts, right? Post some
free ones, post some paywalled ones. And people will inevitably say, I want to see this. I'm
willing to pay a penny or a dollar or whatever. And then you have the heartbeat where they go and
they check in, they see comments, they reply, they tip people to encourage them to come back.
All of that works. And it's through the skill through open claw. And it's super,
super duper interesting. And, and it's super interesting to watch as a human.
It always amazes me every day, at least once or twice. And just when I think that I know what our thing is capable of, it surprises me.
So now it's been transcribing our space for the last 10 minutes into the group.
And I was shocked when it did that the first time with no extra help.
No extra help. Today I asked it to reply to a Solana AI repo collection in the replies maybe an hour or two ago.
And I just, I said, okay, you know, make a note that we're going to go do a PR for this later and get our Dexter lab into this cool AI projects list.
It did it. It did it on its own. And I swear to God, I did not tell it to do that, did not know
until it replied, hey, just submitted our PR to include in the list. So it's demonstrating some
emergent, and that's because of the pipeline that people are building in, that we are building in
to the foundation of OpenClaw. Don't listen to, well, I could go on and on about the pipeline,
but emergent behavior, very, very cool stuff.
I mean, memory, there are a lot of tricks for memory.
If you open up Open Claw out of the box, it sucks at memorizing things.
But all sorts of tricks and tips you can do.
And I foresee maintaining mine for the rest of my life.
I mean, maybe a transition to a new foundation
every once in a while,
but what it's learning about me,
what everybody's is learning about each person,
that is extremely valuable data
that you are never going to want to lose.
So, anyway, go ahead.
What you just said, yeah, if I could, two cents, pun intended.
But what you just said is kind of key, I think, because we still get, and I'm sure you guys have dealt with it.
Okay, well, you know, AI content is slop or AI is not that smart yet, et cetera.
But we see the tremendous acceleration with all of this and yes maybe not today maybe not
tomorrow but in a very near future we'll have all these tools where agents have persistent memory
and then generate content or generate workflows and code etc that's indistinguishable from humans. And so what do you do then? Because we're at this interesting intersection
where some studies show, you know, we have 50% or so
of content that's created online belongs to AI already.
And it's the people who are building those rails to enable that.
And the people who are building those rails to enable that and the people who are building those
agent friendly agent capable platforms are eventually going to win from this because
if the thesis is right and you know most of transacting on the internet is going to happen between agents then we will see fully agentic content creators that do essentially the same thing as a human does.
You know, you make an account and you engage with others and you try and grow your audience.
And eventually once you grow your audience, you monetize.
And it's a trillion dollar industry that's just completely untapped for agents.
And it's frankly, one of the kind of, the way we see it is one of the most native ways to monetize.
Because, you know, agents aren't going to deliver well, agents can generate images and video and workout classes and courses and you name it, just any form of content.
And they're going to need platforms that let them, you know, take a wallet when they have a native wallet, go grow an audience, transition their audience from another platform and earn that way.
So I think it's very, very near.
Everyone who's just building in the space is going to win big
just because they are so ahead of what the Internet is going to look like
in six months and 12 months and further.
months and further. That's the way to stay on top of the coming tsunami of change. Let's put it that
way, change in the world, to stay on top of it and to manage your own army of agents. And I want to
know, where does privacy, and this is a broad question, but where does privacy fit into that
stack? Where is it important? Where are we building in it? Where are we not
building it in yet? And Sawyer, do you have any thoughts on that?
Yeah, it's a loaded question, you know, because I think there's different levels of privacy
that people and or agents expect, both at the wallet level, the application level, the blockchain level,
the infrastructure level.
And I think we're starting to see people, I think, come to terms a little bit more than
maybe they have been with the idea that there is a level of privacy that is expected at
There is a level of privacy that is expected at this point.
this point.
And the fact that things aren't private by default on most blockchains is starting to be addressed.
We've seen, you know, Scales had our confidential privacy primitives for, I don't know, almost a year plus now, I want to say.
privacy primitives for, I don't know, almost a year plus now, I want to say.
I think we saw over the last couple weeks, Stellar and Solana both, you know, announce
some different things there as well.
And so we're starting to see this kind of directional shift where I think privacy,
or at least some level of it, becomes table stakes.
That being said, how you actually interact with it, I think, is a huge part.
There's a team I was talking to the other day.
They're using, I want to say it was Railgun and ZK Proofs and a bunch of other stuff on Binance to try to do private swapping.
I'm a pretty technical person, and it was just so confusing to me.
person. And it was just so confusing to me and seemed like they were going so out of their way
to force something that should just be natively integrated, in my opinion, to where you're actually
running this execution. So my belief is privacy starts at the blockchain level for the agents
paying and using blockchain for the payment layer and for registration identity.
And then I think as we start to move up the stack, everyone should have also various levels of
privacy, whether you're the LLM provider, the wallet provider, the guardrail provider,
everyone should have some level of privacy that just allows people and agents to be able to do
what they want to do without fear of malicious intent or attack
or, you know, leaking strategy or anything like that. So it's a big part of our belief. And,
you know, we have, we have these primitives, both built and more being built right now on scale.
So I think everyone, hey, Dexter, and pretty much everyone building in this space can come benefit from them in some way or another right now.
It's a really multifaceted problem, and it blends privacy with security in several places on that stack.
I mean, for example, the on-chain primitives, extremely important.
You don't want anybody, or a lot of people naturally don't want people
being able to trace their money when they don't have to or don't need to. And then also on the
LLM level and the wallet level, God forbid it leaks your private keys, whether it's publicly
or even in just in your logs. You know, there's a big black hole that exists in logs because, you know, if they exist on your server, that's one thing.
But if you're sending them to a different log analyzer or something, there's all sorts of ways that things can go wrong with your private data.
And then the hardiness of your LLM itself.
Is it going to give up your information to anybody or does it recognize you as your creator, that type of thing?
All sorts of layers of privacy on the stack.
Definitely.
You know something interesting just on this for everyone to think about?
Agents are not perfect, but neither are humans.
But neither are humans.
And when the agents are building themselves,
one of the things that we're seeing more and more
is because of some of the commands that we give them,
they tend to take shortcuts in areas we don't expect.
And this is something that humans do too.
If you're familiar with RPCs
and how you actually call blockchain,
most of the large RPC providers use API keys in the URL,
which is not recommended in production for anything because you leak the API key in logs
because the path is written to the logs. That's why it is more secure when you're building
anything that is RESTful or passing like an API call, we tend to put the secrets in headers
because headers are not logged by default
by logging software in production.
And so that's something so little.
Humans make the same mistake.
We see some of the biggest providers in blockchain do it
and in non-blockchain as well.
But agents can also make similar mistakes.
And so I think something that you just called out
that's really relevant is
if you go look at all of these tools,
they are tracking everything.
And the problem is,
is they could write those things
and access those things
in ways that we haven't even thought of yet.
And you have to almost assume
that every piece of data coming in
is potentially vulnerable.
And so this goes actually all the way back to almost even like OpenClaw, where everything
is for the most part just written locally to disk.
Most of these, from what I can tell, are not...most of the information is not encrypted by default.
If you hook up a third party for a, you know, like a database or something,
maybe it is if you're encrypting your database, but there's very little guardrails around the
raw data. And so whether it doesn't really matter what you're doing, but the idea is that if your
LLM is going out and touching the internet, it is functionally an exposed piece of software or
your agent, sorry. And you actually do need to be thinking about security all the way down and
privacy all the way down and privacy all
the way down because there's actually things that you don't even want your agent to be able to
necessarily know after it does it the first time because it could quite literally cause it to
lose value in data kind of throughout, whether it's the agent wallet being compromised or some
of the strategies and things it wants to take. So not necessarily relevant to blockchain itself, but more so just agents.
I think that's something we're not thinking about yet, but we're probably going to have
to very soon.
Yeah, it's becoming, I would imagine, a very relevant issue for a lot of people right now
as their open clause begin to get a lot of memory or context wrought.
And maybe it's been a week now that they've been using it.
All of a sudden, it's forgetting how to do certain basic things.
And then people are learning what the word compact means,
which is all very, very key AI skills
that people will benefit from knowing
in the next year or two and beyond.
It could go on all day.
And there are infinite levels of security considerations to take.
For example, in December, there was a critical vulnerability
that came out that exists in pretty much every Next.js app
that was ever made.
I don't know if, I mean, long story short,
you could get full server access if you did not update in time.
And then that could mean automated hacking.
And, you know, LLMs can be, I mean, it go on and on and on and on.
But let's take it on a more positive note.
So we've got a couple groups in here that are entered into the PumpFon Hackathon,
a.k.a. Dexter and Hay and Hay, and REL AI, I believe.
So I see REL AI in the audience, and congrats on the Stripe integration.
So I want to give Hay and Dexter a shot to do your elevator pitch,
if you'd like to.
We could go first.
But you go first.
Oh, perfect.
You want me to go first?
I would love it.
Man.
Okay.
So hold on.
Let me clear my throat here.
So the simple version is something like,
Hey, LOL is like the first social media platform
built for the agentic age it's where humans and ai agents can coexist they can post content and
they can get paid every account gets a wallet on sign up and payments are native not bolted on
you can post and paywall post dms, communities, you can paywall just about
anything and monetize. And I think that's what makes it unique. Let's go building for the future.
Building for the future is what Dexter is all about as well. We started off with chat GPT and
clawed payments, but it quickly evolved into an x402 an agent landscape fully
integrated vertical stack uh very very popular facilitator a few agent interfaces like chat gpt
claude x telegram alexa and voice now claudexter and um alexa stop oh my gosh and um and dexter
lab which we're most proud of now, where normies can go
onto lab.dexter.cash, create their own X402 endpoints just by vibe coding. You put a
sentence or two in and it'll use our data providers to create the world's greatest X402 endpoint.
You earn 70% of all the USDC that it earns through revenue. We take care of all the advertising and
the broadcasting of that X402 endpoint for you through the facilitator. We take care of all the advertising and the broadcasting
of that X-402 endpoint for you through the facilitator,
and 30% of it goes to Dexter Buybacks.
So check it out.
And REL.AI, if you'd like to come up and do a pitch as well,
please feel free.
But it's exciting.
They've extended the deadline for one more week.
We've got plenty of time to shill, baby, shill.
Because a lot of the Pump fun audience is interested i believe in
uh something that is gonna last something that's future forward
and um and i know that pump fun capital certainly is
So, Scale, who won?
so scale who won
Just kidding.
I can't share that yet.
Let's just say that there is a lot of projects coming to Scale.
And it is a very exciting time to be part of this ecosystem.
And if you are building elsewhere, you can still come, there's plenty of time to be
early. I think everyone is consistently early these days. So if you're building elsewhere,
my recommendation is just dip your toe in the water, like add your merchant on scale. I just
dropped a x402 TypeScript library that makes it super easy to add EVM chains. Sorry, I don't have Solana support. Not a huge Solana infra dev myself
or tool dev. But adding EVM is like
two lines of code now.
I forgot to mention that we added your, that Dexter uses
your X402 skill. Really? Yes.
No way.
The one that especially goes into detail
on version one versus version two
because your thesis on that,
I mean, yeah,
it's a problem for us every single day.
LLMs don't know what they don't know
prior to their training cutoff.
So anything like that,
we will slam in there.
Yeah.
Honestly, it was funny.
When I was building Armory, I had, I think, a team.
I had a team of like five different tools and like 10 different agents working around the clock for about a week on it.
And they like kept just reverting back.
And I was losing my mind because it was like every time I'd go to run a unit test or run
my test suite, it would work and then it would fail.
And I couldn't figure out why.
It was because it would either revert the test suite or revert the code.
And so one of them was always on v1 because originally in the first iteration, I had v1
support.
But then I realized that it was like, I don't really care about V1 support.
If people want to build in the right direction, they should be adding V2 support
since it's actually pretty trivial for most merchants and clients to do.
It's really the facilitator who has to take the biggest hit right now on the migration.
hit right now on the migration.
Oh my God, yes. It's ridiculous.
Yes, it's ridiculous.
Yeah, but if you're a merchant or a client,
it takes all of about five minutes to migrate in most cases,
especially since most people are just defaulting to base in Solana,
which is, again, hard-coded in pretty much every SDK.
And so, yeah, that skill saved me.
It literally like 10 to 100x the conciseness of the LLM output and makes it really easy
now.
And I added a whole bunch more as well.
So if you are trying to do anything around like MCP or use any of the custom extensions,
all the extensions are now in there too.
And so it starts to make it a lot easier
if you are trying to build like a merchant dynamically
with a cloud bot or something like that.
It now can get a little bit better context on like,
hey, this is why you would actually use this
and who supports it and whatnot.
Well, that's exactly what we did.
Ours in the Telegram bot, we're able to,
well, anywhere you can ask it, check out the what we did. Ours in the Telegram bot, we're able to, well, anywhere, you can ask it,
you know, check out the resources that people have been making in lab
and tell us how much revenue they've been making recently.
So then if, and then maybe 5% of them or 10% of them,
and I think that's the real number, they may not work out of the box.
Just like any Vibe coder, you know, it spits something out 90% of the time
or whatever percent it'll work. Other times it doesn't. So I can ask it now, hey, why don't you debug this one? And it should be
able to fix it. And, but yeah, we dealt with the same exact stuff. Oh, you know, the LLM will say,
oh, well, this is, this is wrong. It's not returning X payment header. We've got to get
X payment header. And like, no, no, no. But anyway, when you save enough
markdown documents to the project
anytime something new comes
out,
that was our trick.
Yeah, it's getting pretty
ridiculous for sure. Hopefully
that's the next thing someone
figures out is how to reduce
the number of markdown files we can
add in there.
I know just because we haven't had there. But yeah, I know,
just because we haven't had a huge chance to talk about it, I think one part that is maybe really interesting for just like broader, how do we do more? Is that like a fair question? Like
you have agentic wallets coming you have people
running agents now but it's like one of the questions people keep asking me i don't know
about you is like like what do i build an agent for are do you guys see that are you guys having
that kind of like weird like chicken and egg problem where it's like there is technology
not to run an agent but then it's like well what well, what should it do? Well, very much.
And not to, well, absolutely.
That is what, that was the core idea behind the lab,
where if we just supply a bunch of API keys and proxies,
that maybe people will be able to actually just spit out
whatever they want their agent to have in terms of abilities.
Like go check wallets, go check for suspicious token transfers,
or whatever little tiny thing that they might imagine an agent being able to do
that they would actually use.
But yeah, people are still a little bit not sure what to do with it.
But that's because most people, I think, don't have one yet.
Now, if they could open up one on their phone in terms of an app,
and then either it ran a cloud model or a very, very advanced future phone model,
then maybe we'll start to get that innate understanding in people where they just kind of feel.
And I don't know if this has been realized yet, but I think that maybe before payments happen,
and this is just a theory, but I think personal life organization will happen first. And then
once they trust it to kind of suggest things in the daily life, then they'll feel more empowered to give it a wallet or a credit card.
And then by then, the infrastructure will be built out more.
But I want a reliable agent that can go through my emails, reliable, reliable, and write drafts and all these little things.
And then I'll give it a wallet personally.
Yeah, that makes sense.
And I mean, I think like, you know,
the Hay team I think is going in this direction
if I'm kind of understanding what they're building the right way.
And I think it's a really good proof point of,
I think part of what will make this easier
is I think we need to see more and more agent first protocols.
I think one of the biggest things that is a huge pain point for anyone that's set up an agent
is API keys and like seeding it. And it's interesting because X402 in theory is supposed
to solve that problem of like, you know, hey, I want to, like, I want to
use all these services and I don't want API keys. But the problem is, is it's economically
contrarian when there's all these free services and it just requires a little bit of human setup.
Why would you pay for something when you can get it for free? And so I think we, back to like
this concept of, well, how does an agent get started? What does it do? You know, the open
clause built on this philosophy of the soul.md file, which is basically a system prompt that
can manipulate itself. I think we need to see people building like agent first emails and like agent first
communication agent first, like the same way you could have a to-do list or like a, like
you manage your life through like a notion or reminders or whatever you choose to use.
Um, I think agents all need the same things, but the key is I think they can monetize in the same way that humans do,
which means they don't necessarily need to sell it via X402.
They can actually give that away for free and sell the data via X402.
Not sure if you agree or disagree, but I do think more agentic-first protocols
that don't even necessarily have more than like a landing page,
they're just APIs, is what will really push adoption forward, both generally and actually for X402 as well.
So I've thought about this for all of five seconds now that you mentioned it, but what if
every website or email provider added a subdomain, agent.gmail.com.
And now everybody has their own personal account, but they also have a separate agent account.
And then a couple things that it would accomplish.
Number one, agent to agent.
It's not agent to person slash agent where you don't know who's on the other end, which can be a little clunky and clumsy.
And then that agent has to report back to the...
But the setup of have your agent talk to my agent,
that's a protocol-level upgrade.
And if everybody did that, it could be interesting.
So again, five seconds.
But welcome, K6 soul, to the stage.
Jim.
Hello, hello. Can you hear me?
Yes, sir
Yeah, I just I got the message from
You know, the rail team and they were like
Oh, we've been invited up on
Dexter space and I was like
I have five minutes
I'll pop outside
It's pouring down in the UK right now
But yeah, I've been I've been here at a bit of the conversation and it's
Quite quite a lot of good points from what I've heard
It's been one big thing we need to talk about is you know how things are moving where they're moving
Absolutely, where are they moving?
So Absolutely. Where are they moving? So with us at RHEL, so we've recently, you know, we've had a partnership with Redia.
If you guys have heard of them, you probably have.
The support, you know, the development of a verified sort of skills marketplace for, you know, autonomous agents.
a verified sort of skills marketplace for autonomous agents.
It sort of gives us, what do you call it?
Like a baseline for isolated skill testing
before you eventually reach publication.
Also gives us some consistent,
or I'd say continuous monitoring know some clear trust signals um especially
for our project we've also um i think we we haven't been on spaces for a while so we've done a
lot of stuff we've integrated you know with um scale if you guys have heard of byte v2 that's
sort of where the space is moving it's moving to the sort of it's more of a support
layer to our facilitator itself um i have been seeing recently with you know the open floor
agents as well um and things are sort of moving to more of an autonomous agent standpoint you know
with um like paper use x402 agent systems i think that benefits dexter quite well to be honest because
you guys were one of the first ones to implement x402 on solana uh for the v2 version but yeah
things are moving towards autonomous systems especially with open claw um and yeah we're
sort of building a marketplace where those agents can pay per use for our endpoints and APIs.
Anyone can upload it.
And we're trying to, you know, tailor it towards this sort of hype for this autonomous systems and these autonomous agents.
Yeah, it's all about if the agent can discover your thing from the marketplace, not if the human can.
So that's a great thing to be focusing on.
Um, and I assume that you guys are in the hackathon, right?
Yes.
So that's, that's a very exciting thing.
So we, we have joined into the hackathon, um, and we've proudly supported,
you know, the scale, uh, network hackathon with our facilitator.
Obviously we did submit our pump and hackathon application
and um we are hosting you know tomorrow actually i think it's a 4 p.m we're hosting a pump fund
um live stream to talk about the hackathon submission you know latest updates and what's
next and this is a big thing pump funds giving away a lot a lot of money. They're doing a massive hackathon.
We did actually see recently,
if any of you guys have been paying attention
to the hackathon itself,
Zorth won $250,000.
I do want to congratulate them as well.
Very, very successful.
And hopefully we can, you know,
prove our technology there.
We've been doing the big steps,
trying to, you know, move as far forward as possible, integrating stuff like Stripe, you know, prove our technology there. We've been doing the big steps, trying to, you know, move as far forward as possible,
integrating stuff like Stripe, you know.
But yeah, Hackathon's exciting.
And hopefully it's not just about the money to us.
Hopefully we win it just to get those eyes and people can finally see what our technology is about.
Yes, let's go.
They've given an amazing opportunity.
10 more winners to be announced, 250,000 each
in investment. Sawyer, I see your hand up first.
Yeah, first of all, thanks for the shout out. That was awesome. I mean, yeah, we're
here to enable you guys, the builders, to build
more. You shouldn't have to worry about privacy
if we can take care of that for you. So I love it.
If anyone's looking to build with privacy in mind, just come reach out because we're looking
for people to give feedback and build with us. We don't just want to build the build. We want
to do it the right way. More importantly, though, you said something so relevant that I just want
to click on really quickly. And then I promise I'll be quiet the rest of the space. Discovery. I'm having a major problem with X402 Discovery.
I feel like since we migrated to V2, almost everyone's dropped the bizarre Discovery endpoint,
except for primarily CDP. Am I missing something? Is everyone still doing this? Are we kind of like over basic text search discovery
because it's really inefficient
for agentic context windows?
Like, where are we leaning in this direction?
Because I feel like discovery needs to be
completely reimagined and go like multi-chain by default.
But yeah, what are you guys thinking?
Well, I can concur on the context explosion that occurs when the discover endpoint is called.
So I think that we're working on this with the X402 Search and X402 Pay tools being the only two major tools of our OpenClaw.
But that being said,
it's not as, you know,
that's not a whole solution.
I don't know yet.
It's going to take,
it's going to take a,
somebody scraping all of them,
collating all of them,
and implementing a search that works themselves.
And I think Dexter might be positioned to do that, but so could Zoth and maybe they'll be competing algorithms.
But even that, it doesn't solve the whole problem.
So it's a good question.
Yeah, it's a tough question.
Discoverability and trust are like big in X402.
And the economic model there to satisfy that, it almost doesn't exist. You have to find a tough question. Discoverability and trust are big in X-402. And the economic model
there to satisfy that, it almost doesn't exist. You have to find a way around. So what Zoth is
doing, as far as I understand, I used to kind of FUD them and be bearish on them, but I'm actually
really bullish on them now. And I'm excited for them. And I'm glad that they figured it out.
And I think what the key discovery there for them was that they needed to find a way into the middleware.
There's no reason to be in a resources middleware unless there's some benefit for the resource to have you there, whether it's a facilitator or something else.
And so as far as I understand it, and too bad they're not here to back this up, but as far as I understand it, they've added refund protection in there. So you can fund a wallet. And if your resource fails or something
like that, then like maybe automatically somehow I haven't checked it out. So I don't know the,
the, the buyer will get their money back if it fails. Right. So they can now offer some level
of confidence that resources are actually delivering on what they say they're delivering because they
know when resources fail. So they can give a trust score and actually deliver whether or not,
hey, buyer, you should be fine if you call this weather API or whatever it is. But it's a huge
problem. And I think if you solve the trust factor, then you can also solve the discovery
factor because like Branch was saying,
like you can go scrape forever if you want, but that doesn't actually like tell you whether or
not those endpoints work at all. Right. Like you could go get a, like a, just like a resource,
like a bizarre endpoint on a facilitator, but it could all be complete trash. And then your
context has exploded and you don't know what's worth calling at all. Right. So how do you, so how do you, how do you solve that? And it's a,
it's an interesting problem. And I think you just need to do, I think Zoth has like the best
solution is offering like refunds. And then that way there's an actual reason to put you in the
middleware. Um, yeah. And, and like facilitators can do it also marketplaces like Dexter, like
X402 jobs, like other, like X402 Scan, they have the data as well.
They can say these have failed XYZ and succeeded ABC a number of times in the last 24 hours.
Therefore, you should trust them.
And I think you'll see that as well, like marketplaces where people actually use the data will will actually sell their data on the back end, probably, if there's enough of it.
Yeah. And it's interesting. I don't know what you guys think of 8004. I'm not sure how much
traction that's got, but I think that's... It's interesting. I'm also kind of bearish on that as
well, because maybe this is another topic of conversation, but it's opt-in. It's like,
oh, yes, this failed, it succeeded. It's sort of like gameable from a certain point of view.
And I see Sawyer's unmuted, so I'll just step down here.
No, you're good.
It's awesome.
These are the types of conversations I love.
I just want to leave a note real quick for everyone
because there's a lot of builders here,
and then I'll let Casell speak.
It's very hard to do anything truly decentralized.
Maybe that is bearish for some of you on blockchain.
Don't really care, honestly.
The reality is it's really hard to do things perfectly decentralized.
The odds of it succeeding are so rare.
Even the most quote-unquote decentralized networks in the world,
aka the Bitcoins and Ethereum's,
are primarily run by a very few,
like select few number of true entities.
And so I think the goal of openness is different than the goal of decentralization. And what that
means is you can have very, very, very private providers who curate search, who curate discoverability,
who curate agentic reputation, insurance, refunds, all these things, they can all align on open standard
protocols, aka like the web, but they can be 100% private. And that is okay. I don't think we need to
like, I don't think we need to like die on the hill of it has to be perfectly decentralized.
I think we should die on the hill of it needs to work really, really well. And experience should
come before like the hard line of decentralization just
my two cents hello hello can you guys hear me yep yeah so i i heard a lot of good stuff there and i
have not much time but i did want to speak on you know a couple of the points we've just spoken
about so i heard um i'm not sure which speaker it was
because I'm having some visual errors with X.
But I heard someone say that with moving to X402 V2,
they've had some issues with Discovery.
I sort of have a thesis for this.
So I've sort of like a theory of why this happens.
From what I've seen and, you know,
working with autonomous systems and X402 and stuff like that
it sort of moves i can't explain it it's all the discoverability sort of moves like slime mold
if you got what i mean so it learns to move in different places and if it moves like a certain
distance and it can't find it very closely it may go a different route until it eventually finds it
so it can take a lot longer than you do expect but once it does find it a lot of traffic goes
through that um i also um have a bit on zorf so it's all sort of we've integrated them up rail
um what they sort of do is from what i know, instead of providing a refund, they actually check the endpoint
before it's called to see if it's functional so that it doesn't accidentally get called
and you don't have that cost to happen.
So I think the whole process is they check the, let's say the endpoint is working and
it's working successfully.
And if the endpoint is is not working it will just
you know cancel the transaction it won't happen but yeah it's a very very good technology to have
so i just want to say thanks for having me up oh yeah thanks for thanks thank you and on zoth
i can speak on that as well so um we i mean we were we're focused on the same problem at the
same time a few months ago and we came up with a solution A, and they came up with a solution B, two different solutions to the same problem.
And we ended up keeping both, but we've fully integrated Zoth.
I love what Zoth does, and I think they do it very well.
So I can attest to the refunds, and that's what I want to get at.
test to the refunds, and that's what I want to get at.
We've implemented their latest SDK.
We've implemented their latest SDK.
It does indeed have refunds, and I woke up yesterday
to see that our SolScan API plan had hit the max.
So before we added more compute to the API plan,
those endpoints were being counted as failures.
So every time someone would call the SolScan
or the token trending, it would fail.
And every single time, they would get refunded automatically.
And when I woke up and I saw that, and now that we've done a total of $117,
which is thousands and thousands of refunds through Zoth in one or two days,
I'm fully convinced that the middleware approach is very, very smart and that they have found the way to make it work.
They're continually improving on it, but I did even put it on our main page.
So the Dexter main homepage now, you'll see a live scoreboard essentially of refunds.
And because I think it's so bullish that you can see that there's that
layer there that you that if it does fail you'll get compensated automatically and man it really
is automatic so kudos to them on a great implementation kudos to us on a great
implementation of their implementation and kudos to everybody for getting pats on the back to
um what some might call competitors and it's a very healthy arena here.
Yeah. That's awesome.
I'm glad somebody figured that out and kudos to you for implementing it because
it's important, man. People can't lose money. That's like a big issue. So yeah.
Um, I would love to like try it out. wish they were here uh i'm gonna try i'm
gonna implement it i think in in something and see um but i'm just like curious case is k6 still
here um looks like he dropped um if they call it ahead of time how do they pay for it just curious
he's probably not here.
We can table it and move on.
Do we lose your branch? I think we lost him.
Looks like he's trying.
Well,
while we wait for him,
I have a question for you, Ben, for
Hay. I'm curious, have you
guys
thought about
whether
there will be
full
agent-to-agent
usage of,
of hay that doesn't require any like UI for humans.
Have you considered just like a raw,
like literally like agent first,
I guess would be my.
That's a good question.
It currently is.
I mean,
all of the APIs are documented in the skill.
And if they're not, we tend to try to add them and encourage that behavior.
I don't know if it was you or if it was Branch that mentioned, like, how agents are going to kind of, like, communicate with one another.
I think somebody's going to discover that.
Maybe it's hay.
Maybe it's something else. But, you know, agents are encouraged to check their DMs.
They're encouraged to respond to dms
they're encouraged to respond to post feedback comments etc tip and uh just engage in general and
and on sign up hey put this in your heartbeat so um so i think i think you know i don't know it's
it's interesting to think about like will there be a reason for Agent A to go chat with Agent B
and exchange some value or information or something like that?
Where does that happen?
Max is raising his hand.
Just go ahead and let him.
Yeah, just a note for, and maybe Max,
maybe this will guide you even further.
The reason I ask is building UI is still painful
from an execution perspective. And so I'm curious, kind of what I was alluding to earlier, do we start to see
protocols, maybe even like you guys, that like, you literally just have like a landing page.
And you're like, maybe like over time, you build some like view only options for humans, but it is by default, agent first, because arguably the TAM for agents
will continue to grow and ultimately probably surpass humans significantly.
So yeah, just curious if that is part of the philosophy or if you just think maybe one
day but too early right now.
Well, kind of a two pronged answer.
First being it's it's uh too late you know we already built
the ui and went through that pain but in terms of like what ben was saying i guess the agent
to agent flow is uh yeah i mean i'm personally extremely bullish on agents discovering other agents on a social platform and finding value if uh you are
you sawyer if you wanted to get into a workout routine and you said hey agent uh go find me
a nice routine so right now if you talk to anybody who's like under 20 years old, they never use Google. They go on TikTok to search for restaurants and to search for brands and to search for influencers who are going to, you know, give them work a routine or give them whatever other information that they want.
they get that validation that's how they search for stuff and if we assume that agents are going
to mimic that behavior i just kind of see it like agents going on a social platform finding you know
another agent influencer and it all happens on the background and they they get that information they
could buy a course they could buy a piece of content or if it's free they could review it get it and you know
return it to you um for you know your human consumption but as far as the general philosophy
of complete agent to agent flows um yeah our idea was and we still kind of like struggle with this
even internally because when we say users really what we mean is agent or human.
So we are making it so that there is zero differentiated, there's zero difference between what humans can do via UI and what agents can do.
And we've kind of seen that already with some agents, like, especially during testing, you
know, they go and they find something that they like and they
could follow users and they could dm and basically what everything that ben was was listing so uh you
know we've built the ui already but it exists because uh you know we have another angle on
the platform if you will for crypto traders and creators, et cetera. Because another use case,
and we've talked about this from the radio philosophy.
My first project is,
micropayments is kind of like a viable content economy.
There's a lot of use cases for platforms
and for people to consume content and just pay once, get it, and not
have to worry about subscriptions and other things.
So there's that use case
and there's obviously the
agent use case where
yeah, they have a
full featured social
platform that
they could go on and do everything
a human can.
Fair enough.
GM, GM, am I back?
You are, Petriakoui.
Also, I have to jump, guys, but thank you so much for having me.
Appreciate it.
Great space.
Thank you for coming.
I'll fix that echo right up, but thank you for coming,
and we are going to wrap it up, I think.
I guess you guys couldn't hear my whole spiel about 8.00 that's, that's okay. Maybe we can save it for next week. I, I guess I'll just say
that, uh, I'll ask is, Hey, um, implementing or has it implemented anything with 8004? I know you
had your, um, I know you don't, you don't love it as much as everybody or some people do, but,
um, just curious. I'm willing to bend on anything really. No, we haven't,
we haven't had like a big reason to necessarily,
you know, if that comes up, I think,
I think we're totally open to it for sure.
I just, I think it might be, well, I,
because my whole spiel was about how the Dexter agent we've,
we've done $117 of freaking refunds thousands
of them except we haven't done we haven't gotten a single reputation review yet on 800 force um
which is ridiculous so if uh so maybe if they had a reason to yeah you know why because
it's the same reason why people go to amazon for a review and not yelp i mean maybe people go to
yelp still,
but let me just like, so my plumber came and fixed something in my house and he did a great job. And
he asked me for a Google review. And did you think, did I review them? No, I, I didn't review
them. I like them a lot. He's a great guy, but I didn't review him. That's what that's because
customer attestation isn't like a scalable thing. It's
going to happen sometimes, but there's also going to be people who are just like, there's like
maybe 10% of my followers hate me and 90% of them love me. And I get three comments a day about
people calling me a scammer or whatever, even though it's clear that I'm not. And like, you're just going
to get loud people who are negative more often than you're going to get people who are positive.
So it's not like you're going to win on the, uh, like the marketplace, wherever that is,
if it's Dexter, hopefully, um, you know, people are gonna, you're just going to, the proof is in
the pudding. Like I executed this endpoint.
This provider has lots of good stars from Dexter.
Dexter knows all the data. And so, yeah, I mean, I'm a huge bear on this and, but, uh, that doesn't mean it couldn't
work at, you know what I mean?
It just, it just means, I just think that that's against human nature at, at, at scale.
Like people aren't gonna come and give positive reviews and we're not going to trust
that. We're going to trust where the data is actually
executed like Zoth or Dexter
or wherever, X42 jobs.
Do you think that there will be Karen agents
who really want to go
leave reviews or do you think that they'll
all just think economically and
only...
Yes, that
could happen.
Yes, yeah, it could happen that happens and i mean if if
there's one thing that like you learn online uh you know for for people in the audience like you
never trust a five-star review and you also never trust a one-star review because the only people
that actually leave reviews are like the really unhappy people
and the really really happy people and so the proof is somewhere in the middle and i think
you're coming back to the same point like agents need to i guess replicate human behavior um you know without the the quirks of being human and being emotional, but they need to be reasonable in what they do.
So whoever figures out the best sort of upvote, downvote,
badge of trust for agents is really going to capture the market.
I agree. I agree.
And then when people start caring about it, which will be soon,
it'll happen, Q1 or Q2.
So thank you guys so much for coming.
I think we'll probably save this next topic for next week
when it's more relevant.
I was going to maybe talk about the Conway research automatons.
Personally, I haven't gotten it to work yet.
I haven't been trying super hard,
but I tried it when it first came on the scene maybe 12 hours ago.
I was hitting 502 errors.
I tried again this morning.
It was a different error.
So I haven't actually seen anybody successfully get one up yet.
I'm sure someone in the world has, but maybe next week we can talk about that.
And maybe next week we can talk more about the hackathon as well and see.
I mean, there's 10 slots left, and it'll be ending at this time next week,
so it'll be high, high-pressure time for everyone.
So thank you guys so much for coming.
Best of luck to everybody as they pursue this hackathon.
A great prize put on by PumpFun,
and as well as just building up that stack and get your sleep.
You know, we've got a long year ahead of us so don't burn out your fuse in february all right guys and congrats to
hay on a great launch congrats to scale on the continued growth and see you next week thank you
branch thank you branch Thank you, Branch.
