Latest / AI for Founders with Ryan Estes / Future of Payments: AI Agents, Micropayments, and Embedded Finance

Transcript

Future of Payments: AI Agents, Micropayments, and Embedded Finance · AI for Founders with Ryan Estes

0:00
Welcome back to AI for Founders. I'm Ryan Estes.
0:02
Today's guest is Alfonso Gomez, Co founder of Crossmint, former
0:06
Google and WhatsApp product manager, and the guy who
0:09
literally helped invent the I am not a robot CAPTCHA.
0:13
In this conversation, Alfonso takes us from building time
0:16
machines at Google. I got to have more details on
0:19
this to building the rails for a programmable economy.
0:23
We'll unpack why he thinks AI agents will soon be financial
0:26
actors, how micro payments could reinvent the Internet, and why
0:30
stable coins are doubling every few months.
0:33
If you're a founder curious about embedding finance into
0:36
your product, preparing for the rise of agentic AI, or just
0:39
wondering what it will mean when bots start paying bots, this
0:43
episode is for you. Big ideas, big numbers, and a
0:45
wild ride into the future of money and trust.
0:48
Let's go. But first in business, the
0:50
founder challenges brought to you by Kitcaster.
0:52
You know podcast guesting works. There's no faster way to build
0:56
trust and authority with your ideal audience.
0:58
But finding the right shows, pitching the hosts and managing
1:01
the logistics is a full time job.
1:03
At Kitcaster, we do it for you. We've booked thousands of
1:06
interviews for over 800 clients, from funded startups to Fortune
1:09
5 hundreds. You show up, you tell your story
1:12
and spark conversations that drive revenue, investment and
1:15
influence. So if you're ready to step onto
1:17
the biggest stages in your industry, head over to Kitcaster
1:20
dot com. We'll get you booked.
1:22
Apply today. That's kick caster by MO burst.
1:26
Hey everybody, you know what's crazy?
1:27
I've been sitting on a goldmine this entire time and I didn't
1:30
even know it. My Gmail inbox, literally
1:33
thousands of warm contacts who already know and trust me.
1:37
But here's the problem every founder faces.
1:39
How do you turn those relationships into revenue
1:41
without feeling like a sleazy salesperson?
1:44
That's exactly why I've been using Warm Star dot AI and
1:47
honestly, it's been a game changer for my business.
1:50
This AI platform connects your e-mail, analyzes all your
1:53
existing contacts, and helps you identify who and your network
1:57
might actually be interested in what you're doing.
1:59
Then it writes. Personalized outreach messages
2:01
that don't sound robotic or pushy.
2:04
The best part? You review every single message
2:06
before it goes out. No spray and pray nonsense.
2:09
This is relationship based selling done right.
2:12
My listeners know I only recommend tools I actually use,
2:16
and Warm Start has literally helped me reconnect with people
2:18
I'd lost touch with and turn those conversations into real
2:21
business opportunities. If you're a founder struggling
2:24
with outreach, check out Warmstar dot AI and use code
2:27
AIFF to get $50 off your first month.
2:32
Trust me, your existing network is worth way more than you
2:35
think. Alfonso, how you doing?
2:50
Great, thank you for inviting me.
2:53
Just grabbed my third coffee of the day, so pretty energized and
2:55
excited about the opportunity. Let's go.
2:58
I mean, hair looks great that you can't beat that.
3:00
Yeah, if a good hair day is is worth its weight in gold.
3:04
Now, I did a a deep dive on you coming into this interview.
3:07
I can't wait to talk about CrossFit.
3:09
But a couple things really stood out.
3:11
One, sounds like you're really smart.
3:13
And then two, you invented the I am not a robot capture.
3:18
Now being an expert in determining who's a robot and
3:23
who's not, which you are perhaps the world's foremost robot
3:26
expert in that I have a lot of conversations about AI and I
3:32
have been accused of being AI from some of the listeners of
3:35
this podcast. So I will let you determine am
3:39
IA robot. That's a great question.
3:43
Do you have a check box there that you can check?
3:47
No. I wish I had like a traffic
3:51
light then to confirm, but we're going to have to say 80% human
3:54
20% robot by the end. I'll see if I update my weights.
3:58
Yeah, you know, it's, it's getting kind of dicey, right?
4:01
The The the line between who's a robot and who's not a robot is
4:05
blurring. How?
4:06
How long do you think it'll be before you won't be able to tell
4:10
anymore? Yeah, that's a great question.
4:14
I mean, I would say there's, there's already places where we
4:19
cannot tell, right. One of them that's increasingly
4:23
getting more and more complicated is just when someone
4:25
is browsing your website, it's hard to tell when someone is
4:30
communicating. Social media, even images and
4:34
video generation are starting to get to the point where it's
4:37
indistinguishable. So I, I would say within the
4:40
digital realm, we're probably within a few years of like
4:44
impossibility to distinguish AI generated this versus human
4:48
generated content. And of course, in the physical
4:50
world, that's going to take longer.
4:52
But it is, I think going to be a drastic change for how we think
4:57
about the Internet and eventually how we think about,
5:00
you know, society, trust and so on.
5:03
And something that, you know, I think we're all going to have to
5:05
figure out or we're going to wake up to.
5:09
Yeah, it's, it's interesting when it comes down to like arts
5:14
and entertainment because you're you're we're getting closer to
5:17
the point where anybody can make anything at any time.
5:23
And So what will rise to the top?
5:24
I think it's probably a net positive as far as like the
5:27
creative experience, because it's democratizing creativity in
5:30
a way that people that didn't have the resources to create now
5:33
can and they have that talent. So you'll see like the very
5:37
interesting people rise to the surface.
5:39
But that's all kind of determinate on taste.
5:42
So figuring out how in arts and entertainment that you could
5:45
just say that, like, the people with the best taste will rise to
5:49
the top. But now if we're talking about
5:51
casual conversations when it's not arts and entertainment that
5:55
like people, we still have to talk to each other in some way,
5:57
I feel like that's where it's going to get squirrely.
5:59
And we'll start to have like agents talking to other agents
6:04
as like a precursor to an actual conversation with a person and
6:08
an agent. And it's going to get pretty
6:10
wild. What do you see kind of the
6:12
short term future as far as that goes?
6:16
Yeah. I would say first is that the, I
6:20
think 1 characteristic of when we approach let's quote, UN
6:23
quote called the singularity is that it is almost impossible to
6:27
predict that what will happen and in which order.
6:29
But I do think on the realm of art, I think what we are seeing
6:40
and the way the way I see it, this I see tools being replaced
6:44
by AI. So in in some way, you know,
6:46
water colours or oil are getting replaced by AI.
6:51
But still you need, as you were saying, a human curator that has
6:54
taste to actually generate this. I'm not sure taste can be
6:58
replaced. It is a great philosophical
7:00
question. I don't know if that's something
7:01
inherently human or that can be trained or understood by an LLM.
7:05
What an LLM can do is given certain instructions, you know,
7:10
I think follow the instructions of the of the tastemaker and
7:16
yeah. I, I forgot who said it, but
7:19
somebody said that the artist of the future will just be a person
7:22
who recognized when art happens, they're not necessarily creating
7:26
things, but rather they're just framing it.
7:28
And, and it, and it was always interesting because it reminded
7:31
me of like, like just if for some reason I had just an image
7:34
of an old man sitting on a, a bench and just being like,
7:37
that's art. That's art.
7:40
But there will be obviously novel ways that we'll be able to
7:42
share data with each other and, and different ways to frame art.
7:46
And it's, we're kind of getting introduced to it by our, our
7:49
language now. You know, I got two kids, 2
7:51
teenage kids, so I know all the slang.
7:53
If you need to know the, the new Gen.
7:55
Z slang, I probably know it. But one of my favorite is aura.
7:59
And they talk about everything having aura or not having aura
8:02
or gaining or losing aura. But when you really press for
8:05
like, hey, what's, what's the definition of aura?
8:08
They're just like, dude, you know, just like vibe.
8:12
You're like, yeah, I do know exactly what you mean.
8:14
You know, so we're like starting to move to that place where like
8:18
the the need to actually create is not necessarily as valuable
8:22
as the need to just observe. Yeah, I agree.
8:25
And there's another analogy, I think in the world of
8:28
programming. But we're starting to see with
8:31
like cursor 11:00 and all these tools a year ago or two years
8:35
ago, everyone was like, maybe engineers are going to get
8:38
replaced by AI. And what I think the general
8:42
consensus right now is like engineers are still going to be
8:46
pretty much needed to 1st figure out what code to write, but most
8:50
importantly to review code. Like reviewing code is becoming
8:54
the bottleneck now. And like the highest performance
8:58
engineering teams, and I think it's probably one that maybe
9:02
what, what it means to review will change.
9:05
You will less need to review if the code is correct, but more
9:09
you know is the right setting the right abstractions or
9:12
solving the problem needs. But I think that part is going
9:14
to remain human for a long time, yeah.
9:18
I agree it's awesome and I want to start to bridge this towards
9:21
your story because I really want to learn more about Cross Mint.
9:24
There's a bunch there. Now I did bring it up and this
9:27
is a fact that you created the I am not a Robot captcha.
9:31
Now, is it true that GP TS Agent mode actually cracked that code
9:38
and can click that box and and prove that they are not a robot?
9:43
Let's see, first thing about this story.
9:46
So a fun fact is that that was my side project as an intern at
9:52
Google My this is, I'm not making this up, but my real
9:56
intern project was to create a time machine.
9:58
I couldn't crack that one down, unfortunately, or if I did, I
10:02
couldn't tell you all. And I just happened to find some
10:06
engineers that were idle working on the recaptcha team that were
10:09
like, we need the PM. We need someone to take all
10:11
these new technologies we have for detecting humans versus
10:14
agents and we need to snap AUI and turn this into a product.
10:17
And that's some of the part that they did.
10:19
And, and the interesting thing about this is like this was what
10:24
like 2013, like 12 years ago. We're already starting to see
10:27
what we're talking about now, which is robots or agents or AI,
10:32
however you want to call it, was already getting better than
10:35
humans at one particular task, which was detecting letters and
10:38
numbers. You guys you guys may remember
10:40
Captcha's used to be just letters and numbers back in the
10:42
day. Yeah, super annoying.
10:45
Yeah. And and then that was the like.
10:49
So I think that's one of the first problems that maybe like
10:53
AI cracked better than humans. But I think everything else,
10:58
it's just been a continuation of that.
10:59
And we're going to see like the same challenge that Google saw
11:02
12 years ago of I cannot no longer the tag by just reading
11:06
letters and numbers if someone is human or not.
11:08
Now it's going to be extending to a lot of other things from
11:12
content generation to browsing and things like this.
11:14
Hey. I I learned this the other day
11:17
too. What?
11:18
That CAPTCHA is an acronym. It stands for Completely
11:21
Automated Public Turing Test to Tell Computers and Humans Apart.
11:27
She's kind of hilarious. Funny that it actually.
11:33
I don't know what came first, if it was the they had a full
11:35
sentence or they tried to make it.
11:37
They tried to reverse engineer the the the acronym.
11:41
Yeah. So, OK, so let me back up really
11:44
quick here. Your your internship was to
11:47
create a time machine. Yeah, I'm not sure how many
11:50
details I can comment, but basically that was that was
11:54
actually the prompt. I fear that I will not get a
11:56
return offer. I have never done product work
11:59
before in my life. And so, yeah, I'd like how I was
12:03
just looking for side projects and it turned out one of them
12:06
turned out to be this thing. No.
12:08
And and eventually that's what I've been interested in in my
12:11
career, is this, this battle of identity between humans and
12:18
robots. Worked on that at Google on
12:21
other products as well on authentication, then their first
12:25
agent SDK 2017 at WhatsApp fighting fraud and abuse and
12:29
encryption and at Crossman. It started also based on the
12:34
seed and but, but focus more on how do we detect and how do we
12:38
reorganize a little bit like the economy around programmatic
12:42
rails in a way that basically works for for humans and agents
12:45
or AI or whatever we call this entities in the future in an
12:48
organized manner. But you know, I think lots of
12:51
impact there. We'll, we'll be unpacking that
12:53
throughout the rest of the conversation, but.
12:56
What can you tell me about The Time Machine?
13:01
Let's see. OK, so I'll tell you a story.
13:04
I think this one, I've seen it published somewhere before I get
13:08
introduced to actually what the challenge was, my manager told
13:12
me how Street View came to start.
13:14
I don't know how much of this is a legend versus a real story,
13:17
but the legend I was told is this was Sergey or Larry
13:20
grabbing beers with some people on the Google Maps team one
13:25
Friday. He was like, he just came out
13:27
with a what seemed like a rhetorical question, which is
13:30
could we build that teleport machine?
13:32
And then everyone was like, OK, this guy's already drunk, Like,
13:36
what's going on here? And then he was like, wait, you
13:39
know, imagine you're a bird and you, you have bird beer.
13:42
You would you see like the Earth from the top?
13:47
You're seeing the same that you can see today in Google Maps.
13:49
And so as a bird, you could theoretically have the feeling
13:52
of teleportation if you're looking at the Google Maps
13:54
screen. And then we switch you to
13:56
another view. And then he was like, now you're
13:59
a human. You're not a bird.
14:01
You're not seeing things from the sky.
14:02
You're seeing them from the street.
14:04
What would it take for us to do it?
14:06
The team started getting excited.
14:07
They started estimating ways to crack the problem with cars with
14:11
like, I don't know, planes with 3D cameras, things like this.
14:15
The end of estimating, they could do it with cars for a few
14:18
100 millions a year, drive cars everywhere in the world.
14:21
And then on Monday, the project got started.
14:23
So I got told that story. And then my manager was like,
14:26
well, now we want to build a time machine, figure it out.
14:31
And basically that was it. There was no more prompt than
14:34
that. But presumably the solution
14:36
would entail something you know, related to Street View, but
14:41
going back in time and into the future as well.
14:44
Interesting. OK, so it reminds me of there.
14:48
Was that like, what was it? It was a podcast episode where
14:53
they're talking about Fallujah and they had drones basically
14:56
videotaping everything from bird's eye view.
14:58
So they could go back in time essentially.
15:00
So if there was a roadside bomb, they could go back in time in
15:03
the video and see who planted it and then see where they drove
15:06
afterward and go back before and see who they were meeting with
15:10
and then follow all those people.
15:12
So we think is it somewhere along those lines kind of?
15:16
You could say. So what I would say is like the
15:19
reality is we did not build 1. So unfortunately I don't have
15:22
the solution to the problem. I could tell you my approach,
15:24
but it probably not irrelevant now, but it's a combination of
15:29
collecting lots of data and then doing a lot of inference to
15:32
cover the gap with artificial intelligence.
15:35
Incredible, incredible. Which brings us to Cross Meant.
15:39
Give us a quick introduction to Cross Meant and really kind of
15:43
how you got started because I believe you're you're the
15:46
founding member that kind of right on the heels of your
15:48
career at Google, you really picked up some investment
15:51
opportunities with folks there and were able to start
15:54
Crossmint. So kind of walk us into this.
15:57
Yeah. So first, so Crossmint, we're a
16:00
developer platform, so we provide tools and APIs for
16:05
companies that want to bridge into what we call the
16:09
programmatic or programmable economy.
16:11
Is this new economy built entirely on primitives that are
16:17
programmable through API. So you have stable coins, for
16:20
example, to represent money, wallets to store this money and
16:25
other digital assets and fungible tokens to represent
16:27
asset ownership. And so they can be used for a
16:31
myriad of use cases. And I go through some of them
16:33
now, but what we do is in a nutshell, if a company wants to
16:37
like integrate some of these primitives inside their app, we
16:39
give them very easy to use hosted tools that abstract out
16:43
all the complexities of the blockchain and allow you to, for
16:45
example, deploy something that looks like a bank like
16:49
experience in 15 minutes inside your application.
16:53
I'll cover a couple of use cases, but we have a long list
16:57
of clients that can be checked on our website that have gone,
17:00
you know from like web three native applications to AI agents
17:05
now. But the biggest 2 verticals that
17:07
we work on today are one is embedded finance.
17:10
These are financial institutions, some older than
17:13
100 years, others are up and coming start-ups that want to
17:19
leverage stable coins for more efficient money movement like
17:24
cross-border transfers or payouts to kind of embedding new
17:28
kind of financial products very easily inside their inside their
17:31
products. And again, this this table, I'll
17:34
talk more later about this if if you wish, but are basically
17:37
making it super easy to deploy bank like infrastructure in
17:41
question of minutes where some which is something that took
17:44
months before. So that's one of our verticals
17:47
with this embedded finance use cases.
17:48
The other one this, this is more up and coming, but it's the one
17:52
we're talking about earlier, which is identic finance.
17:56
It's when you start having these new entities, call them agents,
18:00
AI, or however you'd like. At at now, we're using agents
18:05
mostly for retrieving information, doing research,
18:08
maybe coding. But in the future, we're going
18:11
to want to have these agents actually become financial
18:14
actors. They may need to buy things on
18:16
our behalf. Potentially one day they'll be
18:18
making their own money, they will be paying each other.
18:21
And we're trying to deploy the Rails that allow these agents to
18:25
do this. And we have APIs there, the same
18:28
APIs that are used for stable king use cases work for agents
18:31
as well. There's wallets for them to hold
18:32
money, there's on and off Rams so they can use their customers
18:37
Visa or MasterCard, This new thing called agentic tokens and
18:41
just making sure that everything works in a secure and compliant
18:44
way. That was a mindful, sorry.
18:47
If you would like, we can dig into some of these use cases
18:49
later, but that's what Cosmin does.
18:52
Now, you and I know startups are built in boardrooms.
18:55
They're built in the wild during 3:00 AM.
18:57
Bug fixes, botched launches, and that one meeting where
19:00
everything. Almost fall apart.
19:02
That's what Code Story is all about.
19:04
It's a podcast where tech founders breakdown the make or
19:07
break moments behind their products, What went wrong, what
19:10
went right, and what nearly wrecked everything.
19:13
Hosted by Noah Labhart, Code Story goes past the highlight
19:16
reel and into the honest stuff, late pivots, brutal bugs, and
19:20
somehow coming out alive with a product people actually love.
19:23
If you're in the build stage and the scale stage or the please
19:26
Let this feature work stage, I can relate.
19:29
This show will speak to your soul.
19:31
Listen to Code Story at codestory.co or wherever you get
19:34
your podcast. Just search Code Story, it'll be
19:36
right. At the top.
19:37
Because great startups aren't born.
19:39
They're forged one hard moment at a time.
19:41
Code story.co This episode is brought to you by MO Burst, the
19:48
AI first growth engine trusted by iconic startups like Calm,
19:52
Robin Hood, Reddit, Uber and Google.
19:56
Moburst is a digital full service, mobile first marketing
19:59
agency that helps companies to scale and become category
20:02
leaders. Whether you're finding product
20:05
market fit or ready to accelerate growth, Moburst
20:08
delivers creative, data-driven solutions that help with scale
20:12
and efficiency and impact. At Moburst, we're not just
20:15
guessing what works. We're predicting it.
20:18
As an AI first agency, Moburst uses cutting edge tools to
20:22
analyze user behavior, adapt creative in real time, and.
20:26
Cover hidden growth opportunities.
20:28
The result? Marketing that's faster,
20:30
smarter, and built to convert from early stage startups to
20:34
global brands. We've helped hundreds of
20:36
companies overcome complex challenges, unlock new
20:39
audiences, and optimize every metric that matters.
20:43
At Moberst, we believe that creative thinking, advanced
20:46
technology, and data are the core drivers of success at
20:49
scale. If you're serious about scaling
20:51
your business with precision and purpose, it's time to talk to
20:55
Moberst. Visit moberst.com and let's
20:57
build something category defining together.
21:01
Now how do we get started? Do you want me to?
21:04
Yeah, please. I love the use cases.
21:07
Let's let's start with the top, the the first one you mentioned
21:09
and we'll get to the agentic. And I want to frame it like
21:13
this. I love crypto.
21:15
Crypto is great. It seems like there's a lot more
21:18
use cases for blockchain than crypto, largely because I'm
21:23
still paying 2.9% + 30 cents on merchant service fees across the
21:27
board. And I can't figure out why like
21:29
if we have the infrastructure on the chain, then why do I still
21:34
have to pay this 3% merchant service across the board?
21:38
Now is, is the application here for kind of the wallet
21:41
infrastructure and being able to have bank like transactions on
21:44
website, Is this paving the way to basically get rid of or a
21:48
dramatically reduce that kind of merchant service processing?
21:52
Yeah, that's a good question. So I'll take two things.
21:56
One fees are most certainly point down with crypto and
22:01
stable coins where we're seeing this already in cross-border
22:05
transfers. And I think in the future we
22:07
will see it on payments even not at like nationally let's say
22:11
when you go pay against the merchant.
22:13
At the same time though, sometimes you know, I think
22:18
people imagine, OK, the 3% is going to disappear completely
22:21
when you use crypto. The reality is that 3% has a lot
22:25
of different components and some of those components don't go
22:30
away with crypto. Like for example, when you pay
22:31
with a credit card, you get you have the protection from your
22:34
bank that you can do a chargeback, for instance.
22:36
No, yeah. Or in there's also protections
22:44
that the acquirer, the issuer needs to impose as well.
22:48
And like these things don't come inherently by stable coins if
22:53
you want to pay and still have the ability to charge back or
22:55
things like this. Potentially credit card
22:58
networks, potentially new protocols will emerge and they
23:01
will still take a fee for some kind of like buyer protection or
23:07
insurance. So that won't go away
23:09
necessarily. However, a lot of the
23:12
intermediaries and things like that do disappear.
23:14
So the fees will go down, but I want to say it's not going to go
23:17
down to 0 or if it goes does go down to 0, you lose some of
23:21
their protections and functionality points.
23:22
Is the other one that you're getting today?
23:24
No. So it's important to to set
23:27
expectations accordingly that fees will go down.
23:30
But the biggest use cases for these new technologies for
23:33
stable coins and such are not only fee reduction.
23:37
It's more the programmability aspect and the idea that you can
23:40
launch financial products, You can launch something like a bank
23:45
or you can even like create your own credit card issuance program
23:48
in a matter of minutes or days instead of months.
23:51
That is what I see as the true unblock from these technologies.
23:55
Incredible it's cool. So I I understand that like you
23:59
can't get rid of fees entirely, but three points seems bananas,
24:04
doesn't it? It's like how it how it got tied
24:08
into the gross processing is is pretty crazy.
24:13
I mean, I get it because it's a major revenue generator, but I,
24:17
I really would love to see, you know, a blockchain solution that
24:21
can compete with that in a major way, You know, largely just
24:25
because writing that check every week drops me absolutely crazy.
24:32
But I see your point. Of course, you're going to have
24:33
to have some kind of insurances and protections and also the
24:36
ability to stand up what, you know, bank, which would take
24:41
banking systems and you can just stand them up with Crossmit
24:43
seems completely amazing. The agentic solution.
24:48
One of the things that stood out is you said you're going to have
24:51
agents that are working for you and need to buy things from each
24:54
other is very interesting. I would love to hear your vision
24:57
of that. And maybe to lead into that,
24:59
what do you call a, what are you going to call a group of agents?
25:04
What is that collectively? If it's a murder of crows, what
25:07
do you call a group of AI agents?
25:10
Yeah, I'm not the best at naming, but I've seen people
25:14
call them swarms. Swarms, but but let's say for
25:20
the rest of this conversation, let's use the words harm and let
25:23
let the brother society figure out if that's the right name or
25:26
not. As for so use use cases or why
25:31
why are we working on enabling agents?
25:34
Why do they need these new technologies?
25:38
You first got to start with the use cases.
25:40
No, the idea, this is still abstract, but in the same way
25:46
that we're seeing, for example, in coding that developers are
25:52
changing how they write code to. Before writing code meant think
25:55
about the problem, then actually like think about how to
25:59
translate that to the coding language, type it down, and then
26:04
finally test and verify that everything works.
26:06
What we're doing is replacing that, those two middle steps.
26:09
So people still have to think about the problem, how to solve
26:11
it. They don't need to maybe
26:13
necessarily think about how to translate it into JavaScript and
26:18
they don't need to think as much into like actually typing it.
26:21
But they also need to verify we're moving like coding is
26:25
moving to basically what is called like an intent based the
26:30
humans focus on intended verification and the actual
26:33
execution is done by an agent. The same thing is done with
26:35
research. If you use Chachipati research
26:37
things like this, there's many activities that were if the
26:41
human starts operating at just the intent layer, they just
26:45
express, I want this thing done and then verification later.
26:49
Many activities get stopped with a paywall.
26:54
The most obvious one is if you want an agent to, let's say buy
26:56
your groceries or buy, I don't know the best shoes for hiking.
27:01
Let's talk about this use case. Let's say you, you want you're
27:06
going to a hiking excursion next weekend.
27:08
You need some shoes that are delivered to you in the next two
27:12
days. You have a budget of up to
27:15
$150.00 and the specific hike you're doing is in, I don't
27:19
know, Rocky Mountain and getting the right shoes here.
27:23
This is it requires, it requires a lot of steps.
27:26
You have to go browse a lot of different websites, understand
27:29
their shipping and things like this.
27:30
And then finally, you actually need to send the money and pay.
27:35
And we do think that all of this, and we're starting to see
27:39
the beginning of this, all these comers will probably be
27:42
intermediated by agents that help pick the right one for you.
27:45
And then ultimately they handle all the payment.
27:48
The problem is today, there is no way for there's a way for an
27:53
agent to research, There's no way for the agent to make the
27:56
payment itself. You could naively just give your
28:01
credit card to an agent, but then what happens?
28:04
There's a lot of like the risk model that credit cards have
28:07
today blows up. What if ChatGPT you tell if you
28:11
want shoes for hiking and understands shoes for dancing or
28:15
something. I'd advise you some ballet
28:16
shoes. Who's taking the blame here?
28:19
Is it the credit card holder? Is it the company that's going
28:22
to get the chargeback? Is it open AI that screwed it
28:25
up? The entire entire system doesn't
28:30
work and as a result, basically Visa, MasterCard and others are
28:33
not are are basically saying we're not authorizing this use
28:36
case. So it's one of the things that
28:37
needs to get rethought. This is on the identic commerce
28:40
side. There's even more futuristic use
28:43
cases where agents will need payments.
28:46
There's this thing coming up like it's growing, gathering a
28:50
lot of interest the last few months, which is X4 O2.
28:54
Today we saw an announcement from Cloudflare and Coinbase
28:56
together, which is payments for accessing any website or any API
29:05
in the Internet. So this one is a bit harder to
29:08
wrap your head around it, but it's going to be big in the next
29:10
few months, so I'm going to explain a little bit.
29:13
As we're talking about at the beginning, detecting whether
29:16
someone that's using the Internet is an agent or a human
29:20
is becoming much harder and this is very problematic.
29:23
Before, what would happen if someone control your post on
29:27
Reddit or things like this, But now what's going to start?
29:30
We're going to start seeing is you want to spam your
29:32
competitor. They'll get 300 inbound emails
29:35
that'll look real, but they're going to waste their time
29:38
because it's developed by another person.
29:39
Or you'll get your DMS blown up on Twitter.
29:43
Or you're going to get like anything that you see today is
29:46
free in the Internet. You're currently link completely
29:49
spam. All these things are going to
29:52
start happening. The moment there's an incentive
29:54
to abuse, it's going to get abused.
29:57
And the only way to compensate it, and I tell you guys, because
30:00
I've been working on this for 12 years, is not going to be to try
30:03
to detect these agents, is that we're going to have to start to
30:06
think about the economics of these things.
30:08
Like maybe setting a calendar, inviting my calendly is going to
30:14
have to cost $1.00 or replying sending me at the end is going
30:19
to have to cost 5 or things like this.
30:21
And how is that actually going to work?
30:24
This is going to require some kind of micro payment solution
30:27
that agents can use that an agent on your behalf, your
30:30
browser agent, maybe comment browser in perplexity or ChatGPT
30:34
is going to know Alfonso really values like reading this website
30:39
or they really want to send AVM to this other person.
30:43
For them, it's worth up to $0.50.
30:45
So whenever they say I want to write at the end, they're going
30:49
to handle payment and pay $0.50 from me.
30:52
But we need tools to make that happen.
30:54
And again, you know, please take, take a look at the
30:55
announcements from Cloudflare and Coinbase, because it's
30:58
happening faster than we think. No, and there's even more and
31:01
more future like I think in the future we're going to see agents
31:04
that make their own money and things like this.
31:05
Once again, they don't have any tools to do this.
31:07
They don't have bank accounts, they don't have credit cards,
31:09
they don't have anything. And this is the part that we're,
31:12
we're very excited about. There's also more, more early, I
31:14
would say, I'm sorry, I get excited, carried away and I
31:17
talked about a lot there. But let me know if you you want
31:19
me to dig into any of those topics.
31:21
It's great. It's amazing.
31:22
So you know, it's like you, you're taking away the friction
31:26
so that you you don't have to do all the passwords and everything
31:29
so that when we're creating all these like micro payments,
31:32
CrossFit is going to enable it. So you could use a stable coin.
31:35
So you're not necessarily like speculating on crypto, rather
31:38
it's just straight up payments. You know what, it reminds me of
31:43
the mafia. And the reason is, I just
31:47
rewatched The Sopranos and one thing I noticed is they're
31:51
always handing each other envelopes.
31:53
You know, everyone's paying up to Tony, you know, because he's
31:56
the boss. But then if something happens
31:58
to, you know, Paulie's garage burned down, oh, here you go.
32:01
He's handing him an envelope. They're always like exchanging
32:04
money for everything. Where you look at Goodfellas,
32:07
when he's going in there, he's tip tipping everybody 100 bucks.
32:10
So what you're saying is basically what maybe the mafia
32:13
is always known that you pay tribute, literally, you pay
32:16
attention, you know, and in that case you're using cash.
32:19
But this is going to be very similar because if the, the
32:23
exposure for everything just being free and you can just
32:25
easily create agents to like spam everybody in across the
32:29
planet. You know, it doesn't scale if it
32:31
costs a buck to your point, you know, so maybe we're going back
32:34
to this old world kind of vibe where you like just throwing
32:37
$100 around everywhere. And what you, what you don't
32:40
want to have to do is go to the bank and get 100 bucks every
32:43
time. Rather, you just always want an
32:44
envelope sitting metaphorically in your in your browser and it
32:48
can just pay for all these little micro payments and get
32:51
paid for all these little micro payments.
32:53
Kind of like it's own little reserve, right?
32:55
Yeah, exactly. I never heard that analogy, but
32:57
I actually love it. I'm going to use it going
32:59
forward. Well, there the one part that
33:02
you mentioned that's important is OK, you don't want to go to
33:04
the bank or, you know, authorize with your credit card every
33:07
single time you're doing anything on the Internet.
33:09
The second one is you also don't want to think about whether you
33:11
authorize this or not, but you also don't want to go to a
33:14
website and lose $500.00. So I do think that these
33:19
browsers of the future, whatever tool we use to access all these
33:22
services, we'll also need to be identic and we'll need to have
33:24
your own. We'll need to understand your
33:26
own preferences and make those financial micro decisions for
33:30
you all the time. So you need to be thinking, is
33:32
this worth it or my human owner to pay for this or not, or maybe
33:37
some things are negotiable. Am I going to negotiate if this
33:40
the price of this website can go down and so on.
33:42
I don't know how all of this is exactly going to work, but what
33:46
I could say is like it's an inevitability.
33:49
Like we are going to very soon reach the point where nothing in
33:54
the Internet is going to work today.
33:55
Already you go to Twitter and Reddit and they're starting to
33:59
become more like there's more AI slop than not and it's a matter,
34:05
there's economic incentives at play.
34:06
It's a matter of like years, if not months, that this becomes
34:12
unsustainable and it needs to be completely inverted.
34:15
It needs to be completely changed.
34:16
So you think that's how you do it.
34:18
You box them out with some kind of payment structure.
34:21
Yeah, sorry I didn't get this. Is that the best way to kind of
34:24
get rid of the bots is to like create this kind of micro
34:27
payment? So, so if you want to comment on
34:29
Reddit, it's going to cost you, you know, 1/4 of time or
34:32
something like that. I think so like, and again, I've
34:38
seen this throughout my career and other of the hats I've worn
34:41
in the past is I was the first PM working on a fighting
34:44
automated abuse at WhatsApp, which is the platform that gets
34:48
more abuse on the planet. Like some project we rolled out
34:52
reduce like we made effects that reduce.
34:55
I think it was like 30 million people got less spam overnight
35:00
daily. And also the things that impact
35:02
a lot, like entire countries worth of of users and the way we
35:07
solve these problems. And I think the approach I
35:10
brought to the team and that I've learned in other anti fraud
35:12
teams in the past is everyone starts thinking detecting or
35:17
preventing automated abuse is, is a technical problem is like
35:21
how do we detect OK, this pattern or this other pattern
35:25
that tells if something is malicious or not?
35:27
But it that's not the right approach.
35:30
The right approach to all these problems to solve them is always
35:33
to do an AOI analysis to understand who is causing abuse,
35:38
How are they making money and how can you change the equation,
35:42
The cost part of the of whatever they're doing, like let's say
35:46
generating posts or things like this, so that they do not lose,
35:50
they do not make money on every single time they spam.
35:53
And so imagine the case of WhatsApp, right?
35:55
There's someone making a campaign of 10 million spam
35:59
messages a day and maybe they're making $10,000 because one in
36:03
1000 click and I don't lose a dollar.
36:06
And there's a very easy way to make the problem go away, which
36:10
is how do you see it? But imagine it's from a abstract
36:13
point of view is if every message you send costs one cent,
36:18
obviously what's that going to do?
36:19
But imagine you roll this out, then suddenly you have or 10
36:24
million users is 100,000 on the cost side and then you make
36:28
10,000, you go bankrupt. And that that's ultimately the
36:32
playbook at the top. Anti fraud teams always play is
36:35
look at problems from an economic point of view.
36:38
And the problem is we haven't had the tool of, OK, now if I
36:41
start charging for these things, how are real legit customers
36:46
going to pay? But that's the part that's
36:48
getting unblocked in the last few months is on one hand, the
36:52
micro payment side that's probably going to use stable
36:56
coins and blockchain and #2 is the AI agent that actually makes
37:02
the financial decisions on your behalf of when to pay money or
37:05
not, your financial advisor or your micro payment manager that
37:10
again, LMS are enabling. It's really cool, you know,
37:13
because if web 2 point O was really that the the audience is
37:18
the product, You know, your your metas and everybody else.
37:21
Google's are basically paying for the data that they have on
37:24
the user. Well, if it goes to this micro
37:26
payment, then actually the the the person me, I could get paid
37:30
for my own attention essentially, right.
37:33
So I have my agent swarms that knows what I like, knows what I
37:37
don't, is out there bartering with brands so that like they're
37:41
basically getting the best, yes, bang for my buck.
37:43
So if somebody's selling me black sweatpants, they're like,
37:46
oh, we know Ryan wants that, but like he's willing to charge
37:49
$0.15 for his attention right now.
37:51
We're coming into the fall, we're going to make sure he
37:53
needs that. The best advertisement for black
37:55
sweatpants, it comes in and now I'm being advertised to, but
37:58
they're paying for that privilege, right?
38:00
Something like that. Exactly.
38:01
Yeah, I think that's that's probably what we're going to see
38:05
maybe starting with other simple use cases, but that's still
38:07
worth where things are moving. Incredible.
38:10
That seems. Well, go ahead.
38:11
Sorry. Oh, I was going to say there's
38:12
an interesting project in crypto and it's very maybe on the web
38:15
3D Gen. side. But some part, the revolution of
38:19
it may come, may become mainstream, which is I think
38:22
it's called time dot fun. And it's a project where you can
38:25
buy other people's tokenized time.
38:28
So let's say Ryan, you have a token associated with you and
38:31
people can be able to buy minutes of your time.
38:34
Yeah, today this is used maybe more for speculation and so on,
38:37
but it's ultimately what he's doing is creating like capital
38:40
markets around the time in your calendar.
38:43
And I think it's a it's a good project to watch.
38:46
And in general, I think the analogy of like creating on the
38:50
fly markets for, for all of these scarce resources, it, it's
38:56
going to exist. And I, I think the key is how do
38:58
we make everything transparent and happen in the, in the
39:01
background as a vacant technology so that people just
39:04
have a regular browsing experience, But all these things
39:07
are happening in the, in the background negotiated by agents
39:10
running blockchain rails or whatever it is.
39:14
I love it. I love it because, you know, I
39:16
probably am rare, but I like advertising like when it's
39:20
something that I really want. Like if there was a way I could
39:23
filter everything out. But when like some kind of
39:25
Pharaoh Sanders vinyl record from Japan is available and it
39:30
would let me know immediately, I would love that.
39:32
I would stop everything that I had and would would buy it right
39:34
then, you know, so it creates a lot of opportunity.
39:37
Now this seems like something that is destined to happen
39:40
whether it's like you said months or years away.
39:42
What puts Cross met as as kind of the best position to bring
39:46
this forward into the market? Thank you for asking.
39:50
The first thing I would say is should be our customers who
39:53
judge if we are the best or not for them.
39:56
We're the positioning we have and where we're focused is we do
40:02
have a strong view that like. Block chains, table coins,
40:06
wallets, these technologies are going to be the back end
40:09
infrastructure for this new hyper financialized world that's
40:13
going to be needed for all this to happen.
40:16
But we think these things are back end technologies.
40:19
This is not something that a human needs to like you know,
40:23
see or observe. And 2nd, these are also things
40:26
that are very complicated for the businesses that will need to
40:28
deploy it. It's also something that as a
40:30
business owner, you kind of you will need, you will need someone
40:34
to take care of for you. Like suddenly someone that's
40:38
running a medical office and gets appointments doesn't need
40:43
to learn how blockchain works in order for them to not get spam,
40:46
right? And so our approach is we have
40:49
tools and APIs that are one, make all of the blockchain parts
40:55
completely invisible to users. So there's no, we're like
40:59
passphrases that you may have seen with crypto wallets or gas
41:02
fees or even mention of blockchains or things like this
41:04
is just, you may see dollar balance and things like this.
41:10
And 2nd and most importantly for B to B or B to B to C, but we,
41:15
we rely on companies integrating our products.
41:17
We don't build the experiences ourselves.
41:20
And what we do is like all our tools just assume 0 blockchain
41:23
knowledge by our developers. If you're a blockchain
41:26
developer, you can use Crossmint.
41:28
However, if you're not the one, that's where we are like
41:32
especially suited. You don't need to understand
41:34
what chain you want to use. You don't need to understand gas
41:36
fees, transaction, things like this.
41:38
You just need to know how to call a JavaScript or REST based
41:42
API to say create a wallet, fund it, move money around.
41:46
We have started out the complexity, the legal
41:48
regulations that keep changing and every every like every month
41:53
and every country we abstract out all the security aspects.
41:56
You need to worry about having a chief security officer to handle
42:00
all that stuff and so on. You just call AP is and you can
42:02
get up and running in 15 minutes.
42:05
Amazing. You have to call me once it
42:07
comes down to I can just be like I want to give an envelope to
42:10
everybody with a stack of Benjamins and then I can go in
42:13
there and develop vibe coding on cross MIT.
42:18
You can make sure you got an example of that.
42:19
The Mafia Mafia board with its hand name.
42:23
They will come in sleeves to every other Asian they interact
42:26
with. I saw I was reading, I don't
42:28
know if this is true or not, but it looks like your revenue was
42:30
up 1000% in 2024, Is that right? I'm not sure if we're authorized
42:37
to talk if that's a public number, but things have been
42:40
growing definitely quite a lot recently.
42:43
I think some of these crazy things that I'm seeing here now
42:47
are not as crazy to some of the people in the audience.
42:49
Stablecoins are becoming hot. Every fintech company has a
42:54
stablecoin strategy. And the stuff I'm talking about,
42:57
the agents is there early. But I think next year every
43:01
every company will be thinking about that as well.
43:04
And yeah, we just happen to be positioned at the right time
43:07
with the right product. But things are still early, you
43:11
know, like very small percentage of the world is running on
43:14
stable coins today, although it seems to be doubling every few
43:17
months. I don't want to get partisan at
43:20
all, but the part of the conversation for for crypto and
43:23
blockchain right now is the Trump administration and David
43:26
Sacks and what they've done to basically raise awareness and
43:30
really cosign blockchain and crypto in general.
43:33
Maybe you could give us, you know, something good that's come
43:37
out of that and maybe also something that you would if you
43:39
hadn't asked for them of something that they could do to
43:41
improve kind of of the position for crypto and blockchain.
43:45
Yeah, No, I think one thing they've done that's good is
43:48
they've given a lot of regulatory clarity.
43:51
So Genius Act, for example, legalizes stable coins.
43:54
And you know, I think the imposition of the SEC and things
43:57
like that are, you know, are just like giving a lot of like
44:01
clarity to companies that were looking into integrating these
44:03
technologies because they wanted to.
44:04
They just didn't know they're going to, you know, be working
44:07
more than worse on things that I would ask.
44:11
I think one thing that probably the world the world needs is the
44:18
ability to make make it easier to make investment on things
44:23
without that part being overly regulated.
44:26
So a company being able to issue some of their equity on chain
44:30
and things like that, all of that part being easier, that
44:32
couldn't block a lot of value. And because that's not regulated
44:35
and it's also Muddy Waters, people are creating all these
44:37
like super scummy, like meme coins and things like that, that
44:41
sometimes are, you know, they're much worse than actually being
44:46
able to invest on a legitimate company or enterprise.
44:50
And instead people are just, I don't know, like investing on
44:53
memes. No.
44:53
So there's definitely a desire to invest on things.
44:56
Sometimes people invest on memes that look related to a company
45:00
because they cannot invest in the company itself.
45:02
I think I don't think they should allow anyone to just
45:05
launch a token crazily and let investors take the head, but I
45:08
do think we need to make a much easier framework for that to
45:11
happen. I think they're working on it
45:13
and just exciting to see where that where that goes.
45:16
Amazing. So we're going to enter the
45:18
Lightning round. Are you ready for this?
45:21
Yeah, go for it. OK, so the the audience of this
45:25
show is almost exclusively founders.
45:27
Many of them are have raised or venture backed.
45:32
Perhaps they're considering, you know, introducing crypto and and
45:35
blockchain into their their product stack.
45:36
Maybe, if not for anything else, but for payment, when should
45:40
they reach out to Crossmet to see if how Crossmet can benefit
45:44
their or their Your services and products can benefit their
45:48
company. If it's someone Fintech that you
45:52
reach out today, we're not only we're not a company that just
45:55
gives you AP is we also have a customer success in your team,
45:58
solutions, architects that can help you walk through the
46:01
problem. If you're not in fintech, the
46:05
moment you have a clear thesis on why you want to use wallets,
46:09
table coins or things like this, if you're in agents maybe in
46:12
like 3 months, reach out as well OK?
46:16
Amazing. Now another thing that I think
46:18
kind of threads the audience to this podcast together as they
46:21
all love ideas. So you're an ideas guy.
46:25
Do you have, do you have an idea in your back pocket?
46:27
If someone who wants to go as fast as they can to $1,000,000
46:31
in revenue, what products should they start building today?
46:34
I. Have one.
46:36
It's a bit long to explain, but maybe I'll I'll put it in very
46:40
common terms. I find it very hard to keep
46:46
habits. I'm a busy man.
46:47
Maybe many of you are as well, but one thing I found like for
46:52
example going to the gym everyday.
46:53
One thing I found is with my friends and other older
46:56
founders, we have a support group order for three months.
46:58
For example, we do a challenge of post a tech talk everyday and
47:02
we all do it. And the reason we keep it is the
47:05
moment one doesn't do it, they lose $1000.
47:08
They need to pay to the rest of the group OK and it's
47:10
tremendously effective. No one wants to lose $1000 for
47:14
skipping a day. I believe that this can be very
47:18
easily extended using, again, stable coins, wallets, trusted
47:23
execution and claims or smarter contracts to automatically check
47:26
your Fitbit and be like, did you go to the gym today?
47:29
No, OK, I'm taking $1000 from your wallet and burning it or
47:33
donating it. This can be programmed today.
47:35
There's a lot of very valuable use cases.
47:37
There's also a Black Mirror dystopian use cases of it as
47:40
well. But I think something that would
47:42
be very powerful and at least myself, I would definitely use
47:45
so programmable habits by checking AP is and automatically
47:50
making you lose your money if you don't meet them.
47:53
I think that would be useful. It's amazing.
47:55
Do you think you're more motivated by the stick or the
47:58
carrot? For me, the carrot.
48:03
Yeah. Oh, Oh, you mean in the habits?
48:06
No, no, definitely the stick. Yeah, yeah.
48:12
Yeah, especially if you're burning it right or if it's your
48:15
buddy and you got, you have to see him taking a picture with
48:18
his stable coin wallet that that's up $1000.
48:21
You'd be pretty pissed. I love it.
48:27
Amazing. OK, Everything's moving very
48:31
quickly. You're at the tip of the spear
48:33
of agentic AI blockchain. We have robots right around the
48:38
corner, if not here already, and we have maybe the promise of
48:42
quantum computing. If if you're looking 10 years
48:46
into the future a little bit, how does it look in your mind's
48:49
eye? I think 10 years in the future
48:53
will definitely pass the singularity.
48:54
I think so. It's hard to say, but I think I
48:58
mean the optimistic view on me is we've we didn't talk much
49:03
about this, but we solve all these problems about
49:04
distinguishing organizing agents and humans.
49:08
I think we'll need not only crypto, but cryptography in
49:11
general to be able to understand, you know, what
49:13
content is real and many things like this authorization and so
49:17
on. And we're living a synergetic
49:19
society where humans are still ruling.
49:21
They express intense and their agents do most of the busy work
49:26
for them. But humans are tastemakers, as
49:28
you were saying, they're kind of by working everything that
49:33
you're just just conjuring things into existence and agents
49:37
doing the work. I think we can make this work.
49:40
And obviously the the negative or the pessimistic outcome is a
49:44
much more, you know, I think sadder than that one.
49:46
But I think a society, we're going to figure.
49:48
It out. I agree Fonzo, this has been
49:51
absolutely great. I really appreciate your time.
49:53
Where can people find you online?
49:56
Well, I'm always online. I think Twitter is probably the
49:58
best one. If you search for Alfonso
50:00
Crossman you'll find it or Alfon GJ.
50:04
Alfon G as in Gomez J as Jordana.
50:06
My last two names and I have my DMS open.
50:10
Feel free to DM me and ask more questions and always have a good
50:13
chat. Incredibleandthatscrossmint.com
50:17
where you can find Crossmint, of course.
50:19
Congratulations to your success and thanks for coming on the
50:22
show. Thank you.
50:23
For your time. Hey founders, let me ask you
50:27
this, are you building with AI or getting left behind by it?
50:32
If you're serious about scaling smarter, faster and bolder, you
50:36
need to check out AI for founders.co.
50:40
It's your go to resource for cutting through the noise and
50:43
actually using AI to grow your startup.
50:46
Each week, AI for Founders drops a sharp, no fluff newsletter,
50:51
dives deep with expert guests on the podcast, and hosts monthly
50:55
workshops where you can get your hands dirty with the real tools
50:58
and strategies. Whether you're precede or Series
51:02
B and beyond, launching something new or just staying
51:05
sharp, AI for Founders gives you the edge you need to win in 2025
51:10
and beyond. Hit AI for founders.co to
51:13
subscribe to the newsletter, tune into the podcast, and grab
51:17
your spot in the next workshop, and it's all free.
51:20
Build smarter, move faster, think bolder with AI for
51:24
Founders. Now back to the show.