Google launches AI money
Duration
24:00
Captions
1
Language
EN
Published
Sep 17, 2025
Description
The latest AI News. Learn about LLMs, Gen AI and get ready for the rollout of AGI. Wes Roth covers the latest happenings in the world of OpenAI, Google, Anthropic, NVIDIA and Open Source AI. ______________________________________________ My Links 🔗 ➡️ Twitter: https://x.com/WesRothMoney ➡️ AI Newsletter: https://natural20.beehiiv.com/subscribe Want to work with me? Brand, sponsorship & business inquiries: wesroth@smoothmedia.co Check out my AI Podcast where me and Dylan interview AI experts: https://www.youtube.com/@Wes-Dylan ______________________________________________ #ai #openai #llm
Captions (1)
So, Google just partnered up with
Coinbase to create a sort of a virtual
economy for AI agents. So, a lot of
people have speculated that this was
coming, that AI agents will need their
own currency, their own way to pay for
services so that they're able to acquire
resources and go out there and get stuff
done. You've probably heard this idea
talked about before, but it's here. And
I for one got caught off guard just a
little bit because I didn't expect it to
be here this fast. So Brian Armstrong is
the co-founder and CEO at Coinbase. Now
Coinbase is a crypto trading platform.
So why is it partnering up with Google?
Well, just a few days ago, Google
DeepMind published this big paper. What
is this big paper about? It's about
virtual agent economies. Here's the key
part. The rapid adoption of autonomous
AI agents is giving rise to a new
economic layer where agents transact and
coordinate at scales and speeds beyond
direct human oversight. So again, not a
new idea, but it just feels like it's
here a little bit ahead of schedule. And
notice for the keywords, they do mention
blockchain. Now, a lot of people are
probably hesitant to mess around with
crypto and the blockchain just because
of the sheer amount of scams and various
shenanigans and various shenanigans that
got pulled with crypto and cryptoreated
projects. But a lot of that had to do
with just the possibility of speculating
on prices going up and down. This, as
far as you can tell, is completely
separated from that. So, we're not
talking about any tokens that you can
bet on. We're just talking about some
sort of a payment system that can be
carried out away from the dollar, away
from, you know, needing Visa and the
credit cards companies to transact those
things. Something that can be done very
fast, done automatically by AI agents
and something that can be very, very
transparent. So, none of this is for you
to buy anything. None of this is going
to the moon, nothing like that. But
let's take a look at what it actually
is. So, this is Coinbase. They're saying
Google Agentic Payments Protocol. So, I
guess they're calling it X42.
So, if you're not aware of the
background, there's a lot of these
protocols that are being built out right
now for various AI systems to interact
with one another. MCP was the model
context protocol that was an open-source
protocol by Anthropic and just allows
the various large language models and
other AI systems to interact with other
things whatever those things may be
where it's banking or services online
services or other software. It just
helps them to kind of plug in and
communicate to each other and to get
stuff done. Then we had a much less
known one, but one that's sort of going
to be getting more and more visible.
That's the A2A.
That's Google's agent 2 agent protocol.
So, we've covered this in a video from a
while back when it got announced. So,
we'll come back to that. But just really
fast, Google has been quietly building
out an entire Aentic marketplace. at
least they briefly announced it at one
of their conferences at Google IO either
earlier this year or late last year and
it it seemed like a big deal but it was
there was not too much to find out about
it. So it's something that it seems to
be like they're working in the
background but I wasn't able to get
access to it or anything like that. And
so there's kind of that agentic
marketplace. We'll get back to that in
just a second. There's agent to agent.
So Google Cloud's agent 2 agent A2A
protocol AI agents across multiple
platforms can already communicate and
interact with each other to complete
tasks. So when they introduced this, if
I recall correctly, Google gave an
example of I think something that that
already happened like an example of how
these agents communicate. So let's say
you have an HR department, right? So
you're hiring people for your company.
You have your little HR agent and it's
tasked with doing certain things in
order to hire those people. In this
case, let's say we needed a background
check. That was this was actually the
exact example that Google used. So your
little HR agent would go out there on
the Agentic marketplace and try to find
other AI agents that provided background
check services. So it did a bunch of
research. should interact with a lot of
the other agents and found out that
this, you know, agent X, let's say, was
the best background agent checker for
USbased people and then agent Y was the
best for international background
checks. So, let's say you had a number
of applicants, some of them local, some
of them international. So, it's
interacted with agent X, right? So, it
sent them the information about the
applicants and they said, I need a full
background check. agent X. So that's
from a different company that owns Agent
X, the background check company, right?
So it, you know, did the background
check and it send it back to our HR
agent and then also the HR agent gave
the um international applicants to agent
Y. Agent Y did a background of them and
gave it back to our HR agent. Right? So
all that happens seamlessly without
human interaction. Those agents do the
research, get information, make sure
that everything is what they need.
Right? So the background check meets the
specifications that we need and they
complete the task. So now our little HR
agent was able to get the background
checks on our applicants. Question is
how does the payment take shape, right?
Does our little agent pull out their
Visa rewards credit card and and then
scan it or whatever? Well, maybe that
could have worked, but there's tons of
problems with that. Number one, I mean
the fees for using Visa and stuff like
that. We might need there to be tons of
microtransactions,
right? A few cents or maybe even
fractions of a cents that these agents
can exchange quickly for small
microservices, etc., right? Like if
we're able to create something that has
no friction, that's super fast, doesn't
have huge fees. That would be a massive
unlock of course. So they continue here
now with the addition of X42
with an extension of Google's newest
agentic payments protocol AP2. These
agents can do something entirely new,
handle payments. As one of the first
extensions to AP2 and the only stable
coin facilitator, X42 enables agents to
monetize their own services, pay other
agents or handle micro payments
automatically on behalf of users. A
stable coin is just basically means that
it's backed by let's say a dollar,
right? So if I wanted to give you a
dollar and you needed to give me a
dollar, right? We we would have to go
through let's say the credit card
company, I'd pay you a dollar. They take
some insane fee, right? So basically
every single time we move that money
around, somebody gets paid. That's great
for Visa and Mastercard and all those
companies, but tons of issues and costs
and slowdowns arise with that. But if we
are able to take that dollar, turn it
into a stable coin. So basically
something that's worth a dollar, but
something that we can split into a
million pieces and send quickly
everywhere in whatever fractions that we
want, that would just make things a lot
easier. And that's what a stable coin
is. So what does this mean? That means
that certain tasks which previously
required manual oversight like paying
for data crawls, services or microtasks
can now be executed seamlessly on behalf
of users by the agents themselves. They
do have a demo here. Here's this demo
that they're talking about using X42 and
Google's A2A to buy a new refrigerator.
So, looks like this is Coinbase. They're
using some sort of an agent development
kit. So, here's two agents talking
together. Looks like our agent or the
human is talking to a Lowe's
merchant agent and they're trying to buy
a refrigerator. They get several options
for a refrigerator and it looks like
once they confirm which refrigerator
they want, they're saying, "All right,
I'll process that right away with our
secure X42 payment technology." So, it
gets paid a certain amount of USDC
units. So, that that's that stable coin,
I assume. And so, this allows you to do
various purchases, you know, without
leaving the agent window, without the
chatbot window, you know, you don't have
to bring in your credit card. And also I
mean eventually of course this could be
100% automated with just the agents
making those decisions doing various
micro payments. Maybe you're having some
threshold at which you need to approve
the transaction right but like if it
needs you know 75 cents to get some
computes or whatever to do something
like I don't want to have to approve
that. Just go ahead and get it up to a
certain amount. So why this matters?
This is actually kind of a big deal
because it's going to unlock a lot of
things we couldn't do before. So, for
example, I mean, this solves the
challenge of agents needing to interact
with each other. We're setting the stage
for agentic commerce. So, they're saying
here for developers, this introduces low
friction pay-per-use payments, ideal for
microp payments, per crawl fees, and
other agentdriven scenarios. Now, this
is huge. Just yesterday we were talking
about a situation where a number of
companies are for example suing Google
because their AI search overviews they
claim sort of takes away from their
traffic. So if they're doing
investigative journalism and Google you
know crawls their website and just kind
of summarize them for example that might
take away from their traffic. Lots of AI
companies might crawl Reddit to use that
training data for training their own
models. oftentimes artists are not very
happy with AI models, you know, using
their artwork to train the AI models. We
currently don't have any good solutions
for dealing with that, right? Even if
you wanted to pay like it, how do you
negotiate those payments? You would have
to talk to the company or talk to the
artist. You there's tons of headaches
involved and it just does not work at
scale. It doesn't work fast. Right now,
whenever I do deep research with Chad
GBT, it might go to 20 to 30 different
sites. It gets information from those
sites and organizes into a report for
me. And I love it. I can't complain. But
the reality is all those sites, all that
effort that they've been putting into
building and having that information. I
was able to get around whatever
monetization schemes they had, right?
So, I don't see their ads. I don't see
their pop-ups. And again, while I'm very
happy with that, I mean, at scale, that
does have negative impacts, right?
Because now we're disentivizing people
from building, putting that information
together, from journalists going out
there doing investigative journalism,
and then writing about those things,
like we're cutting them off from their
ability to pay for that content to to
support those efforts, etc. There's tons
of publications where I wouldn't mind
paying them to read their stuff if it's
high quality enough, but it introduces
various headaches. One is it's usually
going to be some monthly subscription
that it's going to be hard to get out
of. You got to put in your credit card
details. You got to deal with a whole
bunch of nonsense. Imagine having an AI
agent that can take care of all of that
while keeping everybody satisfied. Let's
say there's some, as they put here, per
crawl fee. That could be a few cents. It
could be a fraction of a cent. So my
agent goes through, checks everything
out, goes, "Oh, I think this would be a
good thing to include in the report. I'm
authorized to spend a few pennies to
acquire that information or whatever
that number may be." And if the
publication agrees, that thing gets, you
know, paid. I get the information and
gets put into the report that I need,
everyone's happy. Same thing if I wanted
to use somebody's artwork, for example,
on some product or whatever, right? This
could make it a lot easier to license
those things. Instead of me sitting
there going through their terms and
conditions or whatever, the agent's
like, I want to put your artwork on this
thing. This is my budget. Yes or no? And
they handle all the nonsense needed to
get that thing to happen. Cheap, fast
micro payments let all of those things
happen. It allows all those use cases to
become reality. And as they say here,
this marks a big increase in agent
autonomy that will only accelerate as
agent discovery expands through tools
like the X42 Bazaar. I'm always going to
have a hard time for some reason saying
that X42. It almost seems like it spells
something out, but it doesn't. So, they
have the X42 Bazaar. We've already
enabled agents to automatically discover
and interact with any service that's
added to the ecosystem. So, by the way,
here's the X42 Bazaar, right? So this is
self-improving AI agents kind of going
out there and being able to interact and
conduct various financial transactions.
So they have one of the first projects
is this or maybe it's pronounced pre a
stock price API allowing agents to
create up-to-date financial report and
various image and video generation
endpoints. Right? So instead of signing
up for some SAS, some software as a
service for $15 a month, you can just,
you know, if you need one report, you
can negotiate with this company like,
okay, for 25 cents, generate this
specific report or something more
complicated for $5. The point is your
agent should be able to maybe negotiate
that price for you knowing your budget,
your needs, etc. Really gets the wheels
moving, doesn't it? I mean, there's a
lot of possibility here. They give you a
great sort of example of how this would
work, right? So maybe you want to have
some outdoor event. So help me find the
best one. Book it for me. Right? So the
agent could pay less than a penny to get
the weather for next month. So this is
the thing that I'm kind of talking about
because there's a lot of these little
services that you might need once and it
doesn't make sense to sign up for it.
You don't want to see their ads, right?
So the the agent is going to be able to
pull that information without you going
to your website watching all the stupid
ads. If they charged you a dollar to
check the weather, you'd be like, "No,
that's insane." But if it was a fraction
of a penny, I I certainly wouldn't mind
it. If that request at the end of the
day cost me 5 cents, I feel like I'd be
okay with that. Meaning the the actual
completion of, you know, planning and
booking the entire event with this
fraction of a penny thing being as one
of the line items on there. And we've
talked about this before, but this idea
of building autonomous businesses,
right? So this idea of you being the
only sort of human overseeing a bunch of
AI agents that are going out there on
your behalf doing various things sort of
saying in the near future there might be
a self-driving taxi that owns itself and
starts to run its own business. It
starts simple with the taxi having
access to a wallet and being able to
give people rides and using its earnings
to pay for its own maintenance and
operations. It uses this payment method
to learn how to create a website for
itself, find people to hire if it needs
cleaning, or post on social media
advertising and services. Soon, we could
have entire businesses and storefronts
operated by AI. Now, again, like we've
we've been talking about this for a
while now, and this is indeed where
things are going. We're still very
early, but as you can see, serious
companies, I mean, a lot of the Google's
behind a lot of this. They're building
out the infrastructures. They're
thinking ahead and going, "Okay, this is
happening. We know it's happening." And
they're trying to sort of skate to where
the puck is going by building out the
infrastructure on which this whole thing
will likely run. So, I think we're going
to need to do a separate video about the
X42 bazaar because that is the discovery
layer, right? So, it's the discovery
layer for that whole ecosystem. So
notice we already have at least two sort
of marketplaces of this nature. Google
and now Coinbase. Okay, we have to read
this little note that they have here cuz
it's hilarious. They're saying the X42
Bazaar is in early development. While
our vision is to build the Google for
agentic endpoints. We're currently more
like Yahoo search functional but
evolving. Features and APIs may change
as we gather feedback and expand
possibilities. So like we want to be
like Google eventually. That's the goal.
Right now we're kind of like Yahoo. So
work in progress. All right. So also
keep in mind that few days before this
got released, Google DeepMind published
their virtual agent economies. So this
is the Google research paper that they
published. By the way, quick side
tangent. There's this article about it
saying Google researchers warn of
looming AI run economies. These sandbox
economies could emerge that magnify
inequality, monopolize resources, and
create systemic market risks. Google
researchers warn. Risks include systemic
crashes, monopolization, and widening
inequality. And while those things are
sort of mentioned as potential risks,
that not not really what the paper was
about. The paper was really about this
idea that hey, we we do need to create a
separate sandbox economy and this is
what that would look like. In other
words, the point wasn't all of this is
coming and this is in our future. The
point was, hey, just to avoid issues
like this, we need to have this
solution. It needs to be separate so
that the agents and AI, they can kind of
do their own thing off to the side
without interacting with the regular
economy. That's why they're building all
this stuff. But so just FYI, if you see
scary headlines, understand that that's
not what was in the paper necessarily.
They said, "Yeah, these are risks.
Therefore, we should do this. If you're
interested, here's the point. They're
saying, "Our current trajectory points
us towards, you know, this big shift
that's coming, presenting us with
opportunities of awesomeness as well as
significant challenges, which they list
here. I'm just concerned that all the
newspaper articles will just focus on
this part and just ignore the rest of
the paper." And the whole point is this.
We discuss a number of possible designs
that may lead to safely steerable AI
agent markets. To boil down the point of
that paper into one sentence is that
these AI agent markets, they're coming.
They're coming in one way or another.
Accept that as fact. And Google saying
we need to figure out the rules of
engagement like how these marketplaces
will function. And they propose some
things that they think will make it run
better. This is part of the effort I
think of coordinating with Coinbase. But
my point is focus on the first portion
of that statement. These AI
marketplaces, they're coming. Assume
they're going to be here maybe a year
from now, maybe 5 years from now.
Knowing that, how do you approach things
differently moving forward? This is
going to be a massive change in society,
in the economy. Are you setting this one
out? And if you are, that's totally
fine. No shame in it. Just making sure
no one is, you know, a year, five years
from now going, "Shoot, I wish I knew
this was happening because I could have
been in the right place to prepare for
it." Don't be that person. We knew this
was coming and we're seeing the first
sort of examples of this being showcased
here. You're going to watch it being
developed alive by very capable tech
companies. Very soon we're going to be
hearing stories are on fully autonomous
businesses being run by people with AI
agents. And just like people made
millions in the dot era by creating
websites and online businesses, this is
like the next sort of iteration, the
next evolution of that. By the way, in
my experience with a lot of these online
businesses as the whole thing was
starting off, the people that are on
there in the beginning tend to do
extremely well because a lot of these
marketplaces or whatever they are,
right? So, Google's search engine, if
you were one of the first people to
figure out how to do SEO, you're under
you were doing great. Same thing with
YouTube and Twitter and Amazon and
Facebook ads and just about everything.
initially these companies are super
duper cool because they just want users
and they'll do whatever it takes to make
that happen. They want you on there,
right? So, in the beginning to kind of
jump start it to build that network
effect, which is extremely important.
It's also extremely difficult to build
it if you're not the first person to to
build it like to try to create another
strong business off of a network effect
that somebody else already captured is
extremely difficult. So, don't be
surprised if Coinbase and Google pour a
lot of money and effort and marketing
into building this thing out. What are
they going to need? They're going to
need agents that can do stuff. What kind
of stuff? Whatever stuff you can think
of. Background checks and financial
reports and anything. Anything that
people are searching for, they're going
to need some agent that's able to do
that thing for them. That's kind of like
asking, "What do people search for on
Google?" The answer is just anything and
everything. And it will likely be a
while before the big companies, the slow
movers will figure out how to use it,
how to be on there. So, there's going to
be an AI agent on these platforms that's
the first one offering legal help or
offering tax accounting help or help
with graphic design or copywriting or
building e-commerce stores or just
anything. By the way, the agent itself
doesn't necessarily have to be the one
building the ecom store. It can just be
sort of a a sales agent. It should be
there to interact with other AI agents.
Give them quotes, right? Negotiating the
fees, the timelines, kind of getting the
project specs. So, if there's something
that you yourself can do well, you can
have an agent on there just selling your
services. Then you get an email saying,
"Here, I've secured a deal for us." Then
you go to work, you complete that deal,
you hand it off to the agent, and you're
done. I mean, yes, that's not
technically a fully autonomous business.
But initially, people will just use that
thing for the sheer novelty factor just
to find the deals that they're looking
for, etc. When the various app stores
launched on on phones, you know, keep in
mind that there's people that have made
millions making fart apps. If you don't
know what that is, it's an app where you
you push a button and it makes a noise.
Was this innovative technology never
before seen a musthave for the modern
world? No, not really. It was just
there. It was one of the first apps when
the thing was just getting launched. It
was very novel. The app marketplace was
brand new and a lot of people that were
first movers made a lot of money.
Somebody will build something as huge or
maybe even huger bigger than that, you
know, for for this for AI agents, might
be bigger than Google, might be bigger
than Amazon, might be bigger than any
other business we have right now because
you can literally have automation for
anything you can think of. And these
companies have tons of resources. They
can do pretty much anything. The only
thing that they need, the only thing
that they can't provide is well that
network effect. They need people to come
in and build apps and provide their
services and put out the AI agents just
like Google needed people to make
websites for it to organize. Amazon
needs sellers on there to sell their
products. X needs people to post their
stuff on there. So there's YouTube and
Instagram and every other content
platform. That's the game. You're
witnessing the birth of a brand new sort
of platform type, a new medium, and
Google and Coinbase are the first two.
Let me know if there's any other ones
that I'm missing, but as far as I can
tell, these are the first two AI agent
sort of network effect platforms, or at
least they're trying to be. Anyways,
this is obviously big in 5 years. Don't
tell me that nobody warned you. If you
made this far, thank you so much for
watching. Please hit the thumbs up. That
really helps. Right now, Google is doing
something weird with the YouTube views.
Hit thumbs up. That helps to show that
you are a real person and real human to
real human. Uh, that thumbs up would go
a long way. Thank you so much for
watching.