How @0xPolygon is Building in AI

Recorded: March 5, 2026 Duration: 1:03:51
Space Recording

Full Transcription

Thank you. Thank you. . Hello, hello everyone.
Hello, hello, GM, GM.
So we will be waiting for a few more people to join in and then we will start. So welcome Polygon team.
Welcome James Smokey.
Awesome, glad to be here.
Great to see the invites.
So just for starters, James,
What have you been up to today?
what have you been up to today?
Like in general?
We just launched our agentic CLI.
I'm sure we'll get to that later.
Lining up the launch plans, doing the tech QA, bug back, testing.
It's been a fun first part of my day uh
live and now i'm getting some good feedback from the
yeah i have been seeing few announcements lately as well like regarding the latest list over update
um polygon just launched a subsidized gas for agent payments where, you know, these agents can send transactions and the gas will be subsidized by the Polygon Foundation.
So that's very interesting as well.
Yes, yes, for sure.
So it's really exciting.
We're actually getting quite a big uptick in X4O2 transactions, which I'm sure we'll
come down to in a little bit.
But overall, I think it's really exciting to see actually a lot more payments from agents
for these X4O2 endpoints.
I know Quicknode enabled a bunch of them, so this is really great to see.
And yeah, I think we just want to double down
on like Agenda Commerce.
100%, right?
Like, because everything is moving in that direction.
Everything, I mean, every workflow
which you see around is via one or another agent.
So I think it just makes sense.
And I think on that note, we can start
and we can start with introductions. And then we can move into questions, we can talk about what
we have been going on from each teams, what we are developing, what we are launching, what we
already have for these new generation of agents or for these new generation of workflows i would
say so i can kick this off james so hi everyone i am syl from quicknode's side and i work on the
developer relations side where we create a lot of educational content where we create awareness around blockchain around quicknode and just helping people how to build
in blockchain and build better in blockchain so that's what my day-to-day role is and to give you
all a little bit of context around what quicknode is doing around agentic payments and agentic
workflows in general so we have a mcp server using which you know you can access all the
quicknode functionalities like managing your endpoints managing your usage unlocking blockchain
operations like provisioning new infrastructure etc etc via the quicknode mcp you can just connect
that mcp in your lm and then you will be good to go and then on the completely 100 agentic side we also have a x402 proxy using which you can basically
or your agent i would say agent is the better term so your agent could make payments and get a endpoint url or rpc endpoint url and
even web sockets for that matter and then start getting data from blockchain start reading data
directly from blockchain that's possible 101% via agents you do not have to do any touch points
you do not need any accounts, etc, etc.
So we will talk more about that a bit later.
But that's just a little bit introduction or brief introduction from my side.
I will pass it over to James.
If you want to give a little bit introduction around what you do at Polygon,
what Polygon is doing around agentic payments and AI.
Okay, sounds like you're doing a lot on the QuickNet side. It's really cool that you're
I didn't realize you're also at MCP server set up. So, yeah, I needed to dive into that myself.
So, yeah, great to see everyone here. So, my name is James Laud and I'm the head of DevRel
here at Polygon. Fairly recently, so Polygon acquired a company called Sequence a couple months ago. And so we
were brought over an acquisition from there. And since then, you know, I've been using, of course,
AI personally for quite a long time now. And I think around December, everyone kind of realized
like this was just a massive explosion. And the models got a lot better.
So you could actually do these, you know, end-to-end workflows and one-shot prompts.
And, you know, OpenClock came out around that time.
So you had, you know, a lot of interesting things that were being built, like accessing
your calendar, showing stuff on Telegram.
This was a great time.
So we really realized that um you know agenda payments
and agenda commerce more broadly are going to be massive opportunities in the future
um i think you know not to be super bullish but like right now there is the volume is still picking
up like on chain but i find very very little people in this world that are building
agents or just interacting with claude um or you know codex on a regular basis don't really see
where the puck is going where you will be able to buy things with an agent it'll be able to predict
kind of what you want in advance and be able to navigate these sites. And so we're really building for this Agenda Commerce world.
And I think you hit the nail on the head where, you know,
I don't even like generally want to go to a developer dashboard anymore.
I just want to spin up Claude or OpenClaw and it pay for my indexer requests,
my node requests, and it can just do this autonomously.
And all it needs is a wallet.
So even though crypto and AI has been a meme in the past, I think this is a very legitimate use case.
And we see every day now there's new launches coming up.
And from our side, Yuri touched on it, we're building out our own X4R2 facilitator.
So we host our own. We sponsor gas in that facilitator.
So basically you can get the cheapest fees possible
if your agents are paying through that X402 endpoint,
which is really nice, especially for like microtransactions
and nano transactions that are pretty common.
And then something we just released today
is what we're calling the Polygon Agentic CLI.
And that's more or less command line interface
if you're not familiar with what a CLI is.
But really the core value prop
if you're not in the terminal all the time
is you can just let your agent access
all of Polygon technology and services within
a single like skill file.
So it can really get access to a secure wallet with basically on chain defined permissions
so you can't get like rugged or prompt injections.
It has a bunch of other functionality like connecting to X402 endpoints automatically.
You can register with 8004 on chain so other people can find your agents and make payments.
And also is swapping and bridging and deposits into different DeFi protocols built in.
So that's what we've been focused on to kind of consolidate a lot of these other skills
and services into a good package that people can just install,
run wild and start transacting on chain
and go from zero to zero really fast.
Okay. That's a lot of stuff.
Are you sure?
I mean, in a good way,
because I meant that Polygon is working across different channels or across different verticals for the AI space.
So I was just going to ask, is Polygon still a blockchain company or they have transitioned to AI?
It's crazy how much you guys are building.
Yeah, so, you know, broadly as a team,
so we use AI across the entire team.
There's a strong emphasis to, you know,
accelerate your work and do shit more.
And this is really good.
And it's enabled us to produce more.
I think I wouldn't call us, you know,
basically a large focus of all as Polygon Labs is to accelerate the chain, of course, but also focus specifically on payments.
So ultimately, whether they're on-chain or off-chain payments, and how this ties into Agentic is we see that agents will kind of be a large portion of the total payment volume so
we want to make sure we support uh well in advance um but that's the yeah 100 percent
100 yeah i mean i get it and uh maybe a like you know ai ready or ai first blockchain company maybe
that will be a better way to put it and i i think yeah polygon
has always been as a company not just a chain but i would say as a company has always been in the
forefront of new pay payment trends in the industry or ecosystem so really love looking at what you
guys are building right now and you are already prepared for what's coming so
um james you mentioned polygons runs its own explorer to facilitators so a lot of people
who is who are listening now or who might listen later on they might not know what that means you
know what that means functionally so would you like to or would you care to explain
a little bit to the people who are new here to x402 technology or x402 you know as a protocol
or how the protocol and signing works that what does the facilitators do in this entire pipeline
or basically what is a facilitator why polygon is hosting one and what does it mean?
And how does it compare to the Coinbase facilitator,
which Coinbase runs?
Yeah, yeah, that's a great question.
And so, I wanna give kudos,
like, I mean, Coinbase also did a large push
for basically X402.
For those of you who don't know,
like really basic is X402 actually came out of a HTTP status
So, this code called 402 basically just sat unused for a very long time.
And basically, you would, you know, if you make a website request or an HTTP request,
it would basically say a payment is required and then
respond with a 402 error code. So that's where X402 basically comes from. And in short, this was kind
of, I wouldn't say hijacked, but utilized because it really wasn't being utilized at all in order to
basically capitalize it for like crypto payments. And so a facilitator basically sits in order to do payment verification and actually handle
a lot of the settlement.
So in short, you can basically augment any API endpoint with basically X402 handlers.
And what it allows you to do is whenever someone wants to,
let's say I want to make a request at Quicknode,
I want to get some blockchain information
or make an indexing call or whatever the case is,
I would basically ask one of those endpoints,
hey, can you get me my token balance
or whatever I would like to do?
And it would respond with payment
required. It would send it back to my agent. The agent could then detect how much payment is
necessary in order to access that resource. And then it would simply make the payment to
where the response basically said to make the payment to, and then it would allow access
once I pass the verification to that endpoint.
So in short, it's a really seamless way for agents
to unauthenticated make API requests.
And so this is typically a problem,
I'm sure you guys have experienced it,
where you can't have really unauthenticated endpoints
that people just don't pay for because that's
how you get ddossed and in short the the payment aspect basically mitigates that potential risk so
it's a really elegant solution and it allows you know agents that don't have potentially an email
and password or can't do two- two factor authentication or solve captchas.
You know, that's actually being worked on now, but in short, that allows them access to these API endpoints to get data. good that you mentioned that we obviously cannot have unauthenticated agents or unauthenticated
you know second party using us for any kinds of things because you know you have to have
some sort of authentication to use an endpoint so in that way you can manage their usage in that way
you can manage their traffic wherever the traffic is going to etc etc so um on that in fact we have
i'll just speak a little bit on how we have tackled it so what we do is in our x402 proxy
what we do is whenever a user or an agent they want to you know sign up for an endpoint they send a request let's say that they want to get
latest block from polygon so they send get block for latest and then they send the request to the
proxy and then the proxy sends the request like they check if that particular second party is
authenticated or not and if it's not it sends as you mentioned in earlier that it
sends 402 payment required error and then basically the party or agent will just send that request
again but now with payment and then what they do is they sign in with ethereum in this case and in the message they also accept quick node they also
have to accept quick nodes terms and conditions because you know in a way they will be using the
platform's data etc etc so they do that and then they send the request along with that signed in
with ethereum message and of course payment so once that's done then the proxy settles the
payment with the facilitator in this case it will be the polygon facilitator and then get once the
payment is settled it will start serving the request to actual quicknode infrastructure and
the agent will start getting the data and they'll keep on getting the data for whatever
credits they paid and once that credit exhausts they again get a 402 payment required and the
loop continues and that's how it works it's if you look at it you know on a very high level it's
very straightforward and can look very simple but it's very much useful for anyone who, you know, wants to use blockchain data completely via agents without any account, without any email whatsoever.
every user or every agent even gets a JWT token assigned to them so that, you know,
we know we can make sure that this agent is authenticated and they have initially generated a JWT token and they can only access that.
So I guess enough of that, but maybe I should move on with the questions so james i mentioned that polygon
recently launched a subsidized gas program for agents so would you like to talk a little bit
more about that but before that if you would like to cover that what does polygons near instant
finality and you know low transaction cost means for these agents?
Because these agents a lot of time makes thousands of micropayments.
So why does that matter?
And what does that mean for these agents?
Yeah, yeah.
So I can talk a little bit about like the mindset behind this and why we decided to do this.
So essentially we have priority fees on Polygon
and a portion of these are like we're being burned
via 50, 59.
And so we basically wanted,
we identified that like agent payments was a key priority
and we wanted to accelerate this on Polygon.
So we decided that a portion of those fees
would be basically allocated to the addresses
associated with the XROR2 facilitator that we run.
And so in short, this allows you to actually make payments,
XROR2 made payments specifically,
at extremely at no cost for the agent.
And why that's so important
and why also Polygon in general is so important
for agendic payments is because in general,
the chain has extremely low cost.
So very low transaction fees, which agents tend to love.
Because as you mentioned, a lot of these payments,
like they're high volume, but they're low in value.
So like they can be micropayments or nanopayments.
And this is quite common with agents in general.
And so you really need a chain that can settle at low transaction costs.
And what we see time and time again is a lot of the other facilitators that are
hosted, they will have like a free tier, which is fairly low in most cases.
And then the second you kick in over 500 requests or different thresholds, all of a sudden it becomes
economically unviable for a lot of use cases. And so we really wanted to identify that this is a
strategic imperative and that we needed to accelerate this specific
vertical. And so what this basically means is you can have the highest profit margins
for your agent, whatever you're running on Polygon. So that's quite important is by leveraging
our X402 facilitator, you can kind of the the highest value add so businesses
like quick node um or if you're running your own you know agent you're also paying the lowest cost
and i think something that's very interesting and that's unique to agents that hasn't really been
socialized yet is agents don't have the same switching costs that humans do.
So as an engineer, if I integrate with one provider,
I generally have to be convinced to make a big move
and spend resources to actually do this.
But agents don't really have this problem,
and they can analyze at scale,
okay, if I'm making payments
over here versus making payments over here, what is kind of the delta between these two?
And so their volume naturally flows to the path of least resistance. And so what we see is
they will gravitate to where fees are the lowest and therefore it's important to do this at scale.
And so I think that's something to note that we see a lot more of this.
You also need to build for an agentic world where these kind of switching costs are quite lower.
And you can also capitalize on kind of the scale.
Yeah, 100%. And I can attest to the point where you said
they will move towards the path of least resistance.
And it's not just Asians who love those lower fees.
I think we humans love those lower fees more than them
because at the end of the day, it's our money they are using.
So on that note um
james would you like to would you like to dive a little bit deeper into
the latest uh polygon agent cli where it supports uh
wallet creations wallet allowances session limits etc so how does that looks like for an agent and
do you give a context to an agent and can you manage the level of allowances and level of
permissions you give to an agent so basically how does that look like for me who wants to make an agent, let's say who does perps on Polygon, and I want this agent to handle my wallet end to end.
So how does that look like for me in this example?
Yeah, yeah.
So great question.
So this has been pretty widely recognized. And I see a lot of people, you know, they have their agentic wallet. But I think something that I've noticed working a lot with agents, whether it's OpenClaw or Clawed or Codex, is that typically I had to, you know, cobble together a bunch of different services
in order to have something worthwhile.
And it's pretty hard to, like the discovery layer of like agentic stuff
still hasn't really been figured out.
Like I have to know where skills and where different services are.
And so we wanted to kind of aggregate a few different services into a single CLI
to make it really smooth to actually build applications or have your agent actually transact on chain.
So that was the focus of the agentic CLI.
And so in short, you hit the nail on the head. So one innovation is basically using a smart wallet and the smart wallet is basically unique to agents in particular because you can
create what we call smart sessions. So you basically your agent will ask you to like
authorize this wallet and it will also ask you like what are the limitations that you want your
agent and you want to give your agent. So as an example, I want to only do 100 USDC per day.
I only want to interact with these contract addresses.
Like you mentioned, a perps protocol.
I only want to interact with perps.
And this happens over some timeframe.
So what it allows you to do is you can actually like,
you sign as a human and you give permissions for your agent to do this
over whatever kind of time frame you like.
And then your agent can go off to the races,
can transact, can do anything on chain.
You can let it run free,
but you have all the basically security permissions
like set on chain.
So that way people can do things like prompt injections
if they find it or highly mitigated.
It also enables you to have more security that you can give more funds to your agent to transact
autonomously. Because I found like when you're using an EOA or you're using other kind of wallets
that don't have these permissioning setups, it's really hard to trust your agent to use quite a lot of funds other than like
play money to do anything worthwhile or to have it act autonomously.
Like I have an agent set up to automatically find the best yield bearing protocols once
per day and switches between these different ones.
And so this is something that's easy to set up with the Agendix CLI,
and you can just easily analyze this.
It can call different functions to get like the amount
that it holds and balances,
and then also automatically swap or bridge.
So if these DeFi protocols are in different chains,
it can also take advantage of that.
So that's a little bit about what we build
is like consulting all these services into a single CLI,
having the smart wallet in those on-chain permissions,
and then we also built a few different skill files
that walk you through the different flows.
So if you have OpenClaw,
you can just install one of those skill files,
it'll read the whole thing,
and then literally just run off to the races and you can start
transacting on chain.
So it's really cool.
And I think James, on that note, it will be a good segue into asking what are the early
apps which are you seeing right now?
What are builders gravitating towards?
So I read somewhere that openclaw had successfully
done ai to ai commercial transaction on polygon so that's one example but what are you seeing
like what's new being built because we already talked about what polygon has to offer
for to help these builders get started on the chain, get started with these AI toolings and how they can easily use these tools to use these agents
interact on Polygon.
So what are these new things that are being built you are seeing?
So a few things that come to mind.
So of course, Polygon is the home of Polymarket.
So of course, of course, like Polygon is the home of Polymarket.
And so a large amount of on-chain activity,
I would say actually the vast majority,
is centered around prediction markets.
And a big reason for this is because LLMs are uniquely positioned
to analyze social information and actually capitalize on that.
So LLMs are still not really used in quantitative trading firms
to actually like, oh, what is the price of Bitcoin in the next hour or so?
LLMs are still really bad at this and they're not better than chance.
However, they are really good at parsing all the information that's currently happening on X, looking at who you should be
following and then making like bets associated with that on prediction markets since it's really
reactive to social media and basically like what people are talking about. So that's something
that's very unique to Polygon where, of course, Polymarket is the biggest prediction market.
And so you can have agents that are interacting with like X or other social media in order to get a great advantage.
So we see a lot of people building agents to interact with Polymarket and then doing different strategies there.
Another one is also like, as I mentioned, kind of evaluating or evaluating or having autonomous yield finding opportunities.
So a lot of DeFi use cases where agents can autonomously get the best savings rate for
This is really cool.
And I use this myself personally.
And it's a lot of cool to see your agent move around and float around between different DeFi protocols autonomously.
And then some things that I'm also excited about is the more innovative things that are happening in the space.
For instance, someone released Rent-A-Human, where you can have your agent rent-a-human to actually do things in the physical world for you,
which is hilarious.
So yeah, we had someone like, yeah, we rented someone to actually like hold up a sign
and like Polygon is the home for agentic payments and things like this.
So you can do like these innovative things, but I think that's where I see a lot more opportunity. Of course, like,
yeah, the prediction markets and DeFi, I think are very clear use cases. And then more innovative
ones. And the last one is like, simply, agents are going to want to pay each other. And I'm not
sure how much you've looked in like 8004, but I think as we build these better discovery layers of how agents can actually find each other and then like what services these other agents are offering, it's a natural extension that there needs to be payments between them.
And so we're seeing a lot more innovation on like specifically like the discovery layer.
How do you find agents that you actually want to interact with,
like Quicknode for instance,
and then also how do you make those payments
more efficient automatically?
Yeah, I mean, the reputation layer,
which is being built with the A4EIP,
I think that is that is the key for these agent to agents marketplace where
they can find and along with MCPs where they can find these other agents via MCPs and learn how
or what kind of services those other agents are providing and having a reputation score
because of that AIP I think is very crucial so this kind of solves the building block which is
required to make these agent to agent marketplaces and service exchanges possible so 100 agree on that and i think just the poly market api that the launch
of that api has just accelerated these use cases where people are building
workflows and use cases around prediction markets so yeah i mean even at the even at the decent hackathon i think it was each denver
most of the projects i think more than 80 percent of the projects which people are working which
people were working on were something either related to prediction market or ai or both so
i would say 70 percent were prediction markets plus ai then other 30 percent were
somewhere around scattered between either prediction market or either ai so i think that
is one of the biggest if not the biggest use cases what people are using these workflows, AI plus agentic payments is for.
But yeah, so I just wanted to just sorry, just wanted to mention that as sorry if I went.
I'm seeing something very similar is like, I mean, the hot trend right now is, of course,
agentic stuff. And I think people are are seeing that the
obviously ai is very hot even outside of crypto um and so people are trying to you know how how
can we tap into kind of the ai narrative within um but it's i've seen a lot of other industries
it's a little tangential but in crypto crypto, there's a very natural fit.
And crypto especially is very ideal
for payments actually transacting and making payments,
like because you have this unpermission layer,
which naturally fits into kind of like a agent's workflow.
And so I think we'll just see this like space explode
over the next year or two.
I think we still need to land a few good applications that are unique, especially to agents.
So I don't think we've hit on the biggest use cases just yet.
We see more and more being built out and I would expect this to continue.
And I think when it does actually land, will just go exponential oh 100 i mean we even
haven't seen like a hundred percent full potential use cases for crypto itself and the ai layer just
adds or widens this scope of use cases so i think it's gonna just get exponential from here and we
are gonna see a lot more innovative and even even novel use cases of both of these technologies so james another question i had
was i mean this is a bit of a different question not really related to people using these
using these protocols or tools but something where if people are willing to contribute to
this so most of the stuff which even polygon is working on and even what quicknode is working on
for example we have two ruby like two ruby libraries one which people can integrate into
rails and accept micro payments like ruby on rails apps and accept micropayments like Ruby on Rails app and accept micropayments and one
package we have for Ruby where you could sign payments in HTTP headers and send it to the
X402 protocol and of course we have the X402 protocol proxy as well all of it open source
so I wanted to ask you being the head of developers relations that what is the
willingness you are seeing from people to contribute to these kinds of projects, to projects
which are open source, not just using them, but to contribute to them? Are you seeing any interest
there or uptick there?
I mean, of course, we see a lot of I mean, the open source is actually booming now because, well, everyone is a developer nowadays.
So I think actually, like, almost ironically, now it's actually hard.
It's much harder than it was previously because if I wanted to make a pull request to your repository, I had to, you know, go and do the repo,
your repository. I had to, you know, go and do the repo, get a somewhat general understanding
of like the structure of the code and syntax and how to actually contribute to the repository and
make sure it's structured well. And now I see a large avalanche of like people just submitting
PRs for everything and they're just using AI and
then it becomes actually more difficult to maintain over time.
So I would expect that we have some incentive layer in order to do this.
So I know there's a few people from the Ethereum Foundation and several others where if you
want one of your PRs included into an open source repo, then you should pay the maintainer
or something, or you should put up some kind of stake.
And then if you just have a bad pull request, then your stake can be slashed, which I find
somewhat interesting.
But I suspect that we'll have to, like,
as more and more people become developers
and start contributing to open source repos,
this will probably be something that we need to maintain.
But it is also nice because a lot of,
if you're only an open source protocol,
it wasn't so clear how you actually could create a viable business model
and i think now with agents you can create a much better one um by doing things like what i just
mentioned i've got it and on the use on the user side as well a lot of people are using it because
i can see that uh recently i read that polygon has surpassed any other chains in terms of USDC activity.
And most of the people who use these authentic payments, of course, uses USDC.
So, of course, Polygon's market share in X402 transactions is also very high.
So, what do you think is driving that adoption?
Is that the fee structure or the stablecoin liquidity or something else and
i think like the ipo and before you answer that question i personally think that this is just
going to get increased from here because the catalyst catalyzers are uh catalysts are these
polygon cli and the gas gas subsidy program you guys are running?
Yeah, definitely. So, I mean, as you mentioned, like, we are one of the biggest stablecoin
platforms and now the biggest. So we have billions of dollars flowing through Polygon all the time
in stablecoin volume and payments volume. So, I mean, a few things that are kind of driving this there is we have a lot of enterprises
that are building real use cases, real payment use cases on Polygon. And this is for a few
reasons is because it's an extremely stable and scalable chain. It's been live in production
for a number of years now. So people rely on it. It's been battle-tested.
It can serve large amount of volume, like Polymarket.
So payments companies know that they can rely on it at scale.
And then the second thing is we are constantly iterating and expanding the amount of throughput
that the chain can handle.
So I think over the past few weeks,
we've almost accelerated 1.5x.
I would have to look at the exact number,
the amount of throughput that the chain can handle.
And so this is also something that we can scale extremely
fast, and we have a good team in order to do this.
So that's what we see driving a lot
of the volume is um cheap fast reliable scalable is really a home run and then you know we work
closely with a lot of enterprises to bring more of that volume on chain and i think more broadly um
we also releasing uh the open money stack.
And so that's really a philosophy of like,
how do you move money anywhere in the world?
It doesn't matter if it's on-chain or off-chain.
I just want to make a transaction,
and we can find the best route for you.
So this is a large focus of ours is accelerating just money
And we've already accomplished that.
And so we're doubling down there.
That was a really good answer, by the way.
Agent side is just stretching the story of what we're doing.
So, James, before we move on to the Q&A session,
I would just like to ask you a last question
like uh apart from what already in works is there any new things which are on the roadmap
or new benchmarks or new metrics or new goals to achieve for your team and the foundation in general
for these ai and agent payment stuff for the rest of the stuff?
Yeah, I'll speak to the AI stuff. Of course, we have goals across the whole org and the
foundation. But on the AI side of things specifically, really, we want to get more
agents acting autonomously on chain. So our core goal of ours is to get more people using the CLI,
making it easier to actually interact
with interesting applications.
I think now that we have this infrastructure in place
and kind of an opinionated flow
to get someone started from they know nothing about agents
to deploying something that adds value,
we're going to double down there and then connect with like these various applications,
services such as QuickNode in order to amplify kind of their development process and make it more useful for agents and they can buy more things.
So this is really a core focus over the next few months is working with partners such as
yourself to get
more people actually using these different protocols. I think that's what I see as the
missing layer. We have a lot of infrastructure in place and now it's time to build useful things or
make it easier to build useful things and to iterate really quickly. So that's a core goal over the next few months is onboarding more X402 sellers, creating
more interesting applications, onboarding more X402 buyers so people have unique and
interesting things to buy.
So that's what we'll be doing over the next few months.
That's amazing and that just gives me i mean i i just thought of uh doing uh content together as
well so we do these videos we do a lot of these videos so i was just thinking you or someone from
your team we should do a video together where we build something completely of course by using ai
together where we build something completely, of course, by using AI, but
and also for AI, where the AI agent will be using that whatever we've built.
Yeah, for sure.
Definitely always down to, uh, I, I think that's what I love about, um, about
the engineering side and about how much flawed and open AI has accelerated the
dev process is like,
if I come up with any idea, you can have a working prototype in minutes, basically.
And then you iterate and it's like just wild from what I,
it's, it's ridiculously fast and seamless right now.
Like you just have to be good at explaining and be thorough in your prompt and you can just
one-shot everything if you were very comprehensive or if you were very on point with what you want
you could basically one-shot everything right at this point yeah i used to i haven't used it in a
little while but uh i don't know if you're familiar with Ralph at all, but it was like this more or less like a loop that you just put like Claude in.
And it would break down, like if you wanted to build, I don't know, a remittance app or something, it would break down this app into like 100 byte size, like context windows.
And then it would literally just do a for loop and like loop through every
use case and like produce it but now like lms are so good most times i can just one shot
more or less anything as long as i describe it properly and then i'm off to the races from there
and then i iterate uh yeah 100 because i haven't used the one you were speaking about but i just use cloud code now and
i mean it itself runs or runs multiple agents and one will do research one will start executing it
and one will make sure that whatever the other is trying to write is like the basically an auditor
so they can even audit i mean of course you cannot trust or make 100%
production grade softwares but it's great for creating sample apps and great for creating poc
apps which you know you want to create on day-to-day basis to show people how things work
etc etc yep for sure so that's um that's something that has been really helpful. I think it's also really helpful to, you know, if I build an application or I want to show a new piece of technology, I can have a working application very quickly and just showcase it to a potential user. So I think I love this. Yeah, this really fast iterative process that you can do now.
fast iterate process that you can do now.
with that being said,
I guess we can open
the stage for people if they
have any questions.
James, you can even go ahead if you have any
questions for us or QuickNode team.
I mean, I'm curious what's
on the roadmap for you all, especially
on the Agendix side.
I'm sure you're building stuff across the board, but yeah, what's next on the AI side and how you plan to expand there?
So I think just going more deeper with the integration, what we currently have is on our roadmap right now.
have is on our roadmap right now for example right now we do support rpcs via via x402 but
the quicknote products tag it goes much beyond that so we do have some other comprehensive
products as well which is streams and there's another new one which is coming which is xql explorer so i think just exposing those products via x402 will
be very interesting to see because with streams you can basically set up an etl pipeline entire
etl pipeline without setting up each portion of the pipeline for example extraction transformation
loading part separately but everything is just handled by streams so i
think that will unlock a lot more potential for ai agents to work with blockchain data at scale
or i would should say tremendous scale and for example for a chain like polygon if an agent is
just reading data from the chain continuously and just reading fresh
data it could be a lot of data to work with because the chain moves so fast so yeah so new
products and then just adding a bunch of support and error handling features around the X402 SDK what we have
and adding SLAs to it because right now the X402 endpoints,
they do not have SLAs because, you know, I mean,
how is an agent going to open a ticket?
I mean, if it's an OpenC law agent or if it's an agent agent,
it can definitely do that.
But just these two things,
adding more product supports and
adding better support in general
for any kind of errors,
any kind of things that might go wrong so
that we can have autonomous guard
rails to support
whoever are using those apis
oh that's very interesting um yeah admittedly i haven't thought so much about like the sla
side of things so that's really cool that you're working on that so definitely would
love to hear more about it especially in the the streaming stuff. I think that would be really common with agents
and they'll need to tap into that data
for very fast reading and transactions.
So I guess we can open the floor
to the people who are listening.
So if anyone has any questions,
just raise your hand
and we'll make sure to
bring you all up to the stage and you can just ask the questions be it for quick note or polygon
feel free to ask them
and in the meantime james i think if people anyone doesn't have any questions I have you so I would
and this is for both of us
what do you think
will be that one category that defines
X agentic payments in
2024 what would that be
is it DeFi agents
data marketplaces computation
or content monetization what do
you think that's going to be oh you know if i if i knew i would probably invest in it
so i mean i have my ideas i think i think there's a massive opportunity for um this is like
I think there's a massive opportunity for,
this is like not so wild,
but you take basically existing services
and you make them more agentic friendly.
Yeah, I think that's a pretty common one.
Like someone did this with like agent mail.
So just email for agents.
I think there's a large opportunity to identify
what are prominent use cases that need to be like modified for
agents and make them more compatible. And I think the reason for this is because then you can just
have what humans already want to do, but agents can do with 1000X capacity and at scale.
So I think that's a very interesting thing.
I think we do need the models to be slightly cheaper.
We're already seeing a lot of models coming out now that are filling this niche.
I'm already hitting max plans with a lot of these.
And if you have something that's really autonomous,
then you're like, you're just chewing through tokens.
So I think a little more innovation on like the model side
and we'll be actually in a good place,
I would say in the next six months.
But in terms of services, yeah, I mean,
I can talk about the on-chain things,
like I already mentioned, DeFi and prediction markets.
I do think an autonomous agent that can monetize my finances or optimize my finances is very attractive.
This is something that I don't have the time to do, even though I would love to do.
have the time to do even though i would love to do and i would love to just point my agent um
whether it's and by the way whether it's even crypto or yield opportunities for investing in
the stock market or whatever the case is i would love to just have an agent kind of like optimize
my finances i think there's a big opportunity there once you have things like in a secure state yeah yeah and on
your point of uh on your point of making these models more cheaper i i i would say that there's
always tensorflow we all could go back to tensorflow and start training our own models and create our own LLMs.
So yeah, yeah.
It was crazy, right? Like the times when it was all machine learning, not, not even AI.
I know I missed, I missed the days of support vector machines.
And you had a recurrent neural networks, which was like before transformers.
Now, now everything's.
Yes. neural networks which was like before transformers now everything's like yes um so coming back to the question um so i personally think that x402 along with the eip8004 i think it unlocks
a possibility of i and i mean i have spoken about this earlier as well in the
space that this unlocks the possibility of ai to ai services and i think the market we are going to
see a lot of these we are already seeing but we are going to see a lot more marketplaces where
ai agents can just transact with each other get service from one another and
then you know do end-to-end things without any human interventions where you just have an agent
to set up your day-to-day stuff let's say that i mean this is just a vague very vague example but
let's say that you have a agent set and you have that agent setup to do your
intraday trades you have that agent setup to order groceries for you um let's i mean
it's possible right now right for example these refrigerators what you have these days they have
ai in them they send you notifications on what groceries you are running low
on um you can have cameras uh in your kitchen and using those image feeds you can understand
or your agents can understand or these models can understand what things you are running low on
so a world where automatic refrigerator that's what I want
yeah so imagine like you know
your agent sees that hey
I'm running low on milk so it
orders it goes on and
orders that
from maybe
Amazon MCP
and then Amazon MCP
and there's an agent on that side
and it accepts X402 and keeps track of
their inventory etc etc so yeah i'm sorry i digress so just to be to be very to be very short
and simple i would say a mark the evolution of these ai to AI marketplaces and services is what I think we are going to see a lot.
I would like to see a lot in the second half of the year at least
where everything will be handled by these agents
and they will do agent-to-agent payments,
they will do agent-to-agent service exchanges.
Yeah, I hope that's the case so um
i saw a good post where someone put out basically like yeah right now we're prompting agents to
do something for us um but eventually agents will also learn your context and they'll have
your own history of what you asked for and we'll be able to like
preemptively decide okay you need x y and z or you asked me to do like a much more general task
and it can find out exactly what you need like oh my son is going back to school next month um
okay he needs a backpack he needs pencils whatever case is. And the agent will just realize this and do it without you even requesting it, assuming you give it permission.
So I think this is very interesting of how it can, like, automize a lot of different things.
100% and I think that's good in a way and even scary a bit.
Just a bit.
But yeah, more on the good side, I'd say.
And another question is like, so X402 currently relies heavily on USDC, right?
So what happens when more tokens and more chains come online?
Does this fragment the ecosystem or strengthen it?
What's your take on that?
I don't think it fragments it.
I mean, I think people should be, in general, my philosophy is people should be able to
pay in whatever they want.
Of course, stable coins have, especially USDc has the biggest um you know kind of synergy
with x402 right now but regardless of whether it's stable coins or the native token of the chain or
whatever other tokens it wants to pay in um my thinking is basically this should seamlessly
convert into um whatever token that the merchant wants to receive and just settle.
And that's one thing that we produce is called Trails, which basically does the swap and the bridge underneath the hood.
So a user can basically pay in anything they want and Trails will automatically figure out what the merchant wants to actually settle in and
does this under the hood.
So that happens at the blockchain layer, but I suspect a lot of like facilitators and merchants
will want to do this automatically because that just opens up.
It's similar to like any kind of payment method.
If I'm a retail store, I want to be able to accept credit card and debit card and check and bank transfers or whatever the case is, as long as it's not too much of a burden.
And that just enables more payment options for your customers.
100%. Okay, that makes sense.
all right james so i we are almost at the top of the hour and i think we have
been speaking for almost 58 minutes so beautiful so if anyone has any questions this will be the
time to ask those and if not like you know you can always dm james or you can always dm me or even polygon or quicknote accounts
if you have any questions around the blockchain in general or even using the tools whatever
the support
all right so looks like no questions so i would just like to say if you are looking to build on Polygon,
you can go to the docs linked in the pinned comment.
You would be able to see those on top of the space.
So if you're looking to build on Polygon,
those docs could be a really good place to start.
And yeah, we have been working with Polygon team from a long time.
It's been years now.
So yeah, I think you guys are going to see a lot more of these,
if not continuously, but definitely occasionally.
So yes, if you are building on Polygon, that's great.
Go to those docs still and learn what we support for Polygon.
And if you are thinking on starting on with Polygon,
again, go to those docs and you will find 0 to 1 guides as well.
So that's my call to action.
And James, if you would like to share anything with people,
what they should be looking out for,
where they can find you and your team next,
any event, any online stream, workshop,
whatever you wanna share.
Yeah, yeah, I would think, you know,
my call to action is, you know,
check out the Agendix CLI, let us know your feedback.
If you'd like to have anything incorporated there,
we're more than happy to collaborate on that.
So, yeah, I think that would be the main thing.
And the next few months will be, you know,
at a bunch of different events all over the world.
So my team and others from Polygon.
So I'm sure you'll find us out there at different fintechs,
crypto conferences all over the world.
So we'll be there.
And yeah, if you're looking to build on Polygon, crypto conferences all over the world. So we'll be there.
And yeah, if you're looking to build on Polygon,
if you want to use AI,
more than happy to get in touch and would love to chat more about it.
So I always like talking AI and Agendix stuff.
So more than happy to help no matter where it is.
And for people who are wondering,
where can I find Polygon Agent CLI's link?
So you can just go to the comment section of this space and I have linked the GitHub for Polygon's
Agentic CLI or Polygon Agent CLI to be precise. So you can just go to the GitHub, learn what it is,
how it works. I can see they also have a really easy to understand
architecture diagram or flow diagram.
So yeah, highly recommended.
And I think I'm going to look at it.
I'm going to check it out myself as well.
So yeah, please check it out.
It's linked in the comments of the space.
And if you have any questions
run into any blockers, you know whom to text.
You have to bother James.
Always worth
a bother. Definitely no problem.
So thanks so much for having me.
Really appreciate the feedback and would love to hear your feedback
as well on the CLI.
Thanks for having me.
It was a pleasure having you, James.
And thanks Polygon team. For sure. For sure. For sure. It was a pleasure having you, James. And thanks, Polygon team.
Take care, everyone.