Building the Payment Protocol for AI Agents

If you found out Thursday night that your agent could be five IQ points smarter, by the next day, all 3,000,000,000 agents will have known about that. So if you've ever seen a four zero four, that's an error. Four zero two was meant to be payment required. In the early nineties, they they tried to go do a lot of this. Financial services companies, banks, whatever, they did not want this to happen. Merchants are gonna show up and be like, wait, how do I sell to agents?

This whole world is built on the paradigm of, like, bots are bad. We need captchas and all that. What if you actually want to go sell to the 50 times larger audience of agents as opposed to a billion human users? Crazy to think about. I don't think retailers are ready for their human traffic hitting their website to fall off. I, as people have called it, or LLMs, is a very smart friend. An agent is one step further. This is the simple explanation.

Hey, it's Drew. Quick thank you to Altitude for making this episode possible. Altitude is rebuilding business banking for teams on the frontier. They offer the full stack for the modern CFO and founder, a beautifully designed platform for payouts, bill pay, corporate cards, enterprise controls, and so much more. Move money with speed, yield, and total control all from your stablecoin balance. Operate at Altitude. If your business is on the frontier, your finances should be too. You can learn more at altitude.xyz.

Alright, Lewis. Thank you so much for joining us. Appreciate you coming on and chatting with us today. Thank you for having me, Monard, and

hope to pay it off with some good takes, and hopefully some predictions that pan out over a reasonable time horizon. We'll see. Yeah. You

when we were sitting down, you asked me, you know, what is the what's kind of the angle of the show? What are we doing? How should you think about this? And I think as we think about Stable Minded and this new part of our journey, I think what we're really excited to to dive into is the the founders and the builders on the frontier of what what is being built that is really pushing the limits of what people have really seen before.

And when I think about what that looks like today in the context of stablecoins and payments, the the neck of the woods that we're in, I really think about stuff that you're building with AgenTic Finance, AgenTic Commerce, and a lot of the vision of the world it sounds like you have. So I'm really excited to welcome you in and dive into how you got to this point and what what made you convicted about all of this stuff. So let's start high level intro of who what's full name?

What are you Get get that stuff out of the way, and then we'll we'll go back. Yes. So my name's Lewis Samira. I

co founded a company called Circuit and Chisel way back in the day. Grew up playing golf, went to school to play golf, kinda wanted to be a professional golfer. That lasted about six weeks. I think I got the lowest score on my freshman midterm economics test you could possibly get, and didn't set any scoring records on the college team in that first six weeks. So I was like, okay, we should probably figure out what other options there were. So you went to college to pursue playing golf.

You mentioned you were extremely talented as a kid with golf. I was very good in junior golf and then pretty good in amateur golf, but which is roughly when you finish high school, you go off and play at college. I had the choice to go to some of the, like, golf powerhouse schools in the warmer climates and all that sort of stuff. I went to an academic school, I thought I could do both. I thought I could, like, both be smart, get a good degree, and turn professional, and kudos to the people that managed to do it.

I was not one of them. And in relatively short order, was like, okay, golf's probably not gonna be the thing that, like, pays the bills in the future, so I can figure out what I actually like from this course catalog and figure out what my job might be in the future. In college, I studied a bunch of things, but mainly settled on behavioral economics and then renewable energy, smart grid style stuff. This was late late noughts, late two thousands.

Loved behavioral economics, incentive alignment, everything about that particular field, and it was emerging at that point in time. A lot of pop science books were being written. What is behavioral economics? How do you define that? Everything from, like, where they put the lines on the laundry detergent caps to where you actually, like, they will strengthen it to be two times as strong or four times as strong.

You'll go fill it up and basically use two times or four times as much as you need to, but, like, that can't be enough, so I'm gonna keep filling it up, and you just motor through laundry detergent faster. That's one of the first examples they used, so there's some bias popping to mind there.

But everything from that to revealed preference versus stated preference, to what do you do when you're more tired, pricing experiments where if you put two very high priced things and a slightly cheaper thing, like, that looks like a great deal. So just a bunch of those smaller pieces. It was almost It was more psychology than economics, but it was applied to business, and that was the closest thing we had to business. So I I thought there was something there.

Loved smart grid stuff, went and did a renewable energy internship, which not many people at my school did. So now we're talking about data centers and nuclear fusion and all that, like that was actually my degree once upon a time. That whole world is fascinating. Fast forward, senior year, read the Bitcoin white paper, talked to my applied math professor, I was like, this seems amazing. So this is thousand ten, so a little bit after it was written. I wanna write my paper on this.

And he was like, they've been talking about digital cash for years. And got talked out of it, didn't do it. Wrote a paper on like statistics of March madness, which a whole other story. Seems cooler at the Seems way cooler for a college senior turning 21. Eventually, bought Bitcoin a couple years later, but I missed the, like, 10¢ round of Bitcoin.

Anyways, I had worked at a startup afterwards, smart grid startup, went on to Google, did a bunch of different things at Google, went trying to work on some of Google's smart grid initiatives that were pre Nest acquisition and all that. Fascinating. Started a company within Google's incubator, started a company afterwards, and had been like I said, I bought Bitcoin early, watched a bunch of the events around Mt.

Gox happen, and like Ethereum getting launched, and Salon and all that, but more as like a casual person on the edges, just like, this seems cool. Yeah. Met a mask. I have a wallet, and I had my Bitcoin and Coinbase vault in 2014 or '15, which like is a feature most people didn't even know existed.

So that was like their original cold story? Yeah.

Okay. Actually, in 2017, I wanna say, I went back when Bitcoin went to 17,000 or whatever it was, and I was like, how many did I buy back then? And I saw it, I think it was two or three. I was like, oh, man, I was kind of a wimp, but it was at $200. Then I found that I had more in the vault, and I was like, yes. I was one of the few users that used the vault, I guess.

Anyways, was at Google, started a company, that went that was an interesting story in and of itself, but it didn't progress into what Google's incubator was meant to do, which was effectively stop people from leaving, just product areas could go acquire it, the startups within. Learned a ton, started my own company. That went reasonably well. Took a little bit of time off when my first son was born, and that was when I was sort of like, do I go back to smart grid? Do I do, you know, life extension, space?

Like, what is interesting in my, you know, early thirties at that point? And I had always liked crypto. DeFi summer was happening, and I was like, okay. Let's go try and make this let's go see if this is worth exploring. Spent $50 a transaction on Ethereum gas, like, not knowing what I was doing at that point, trying to figure out Compound and Aave, and luckily avoided all the craziness that really took a bunch of people down.

Worked at Chainlink for a bit, which was a fascinating experience, like, very strong crew over there that taught me a lot and took a a chance hiring somebody that wasn't in the crypto space. Yeah. What what drew you to Chainlink specifically from this was from Google. This was your first kind of formal This was first crypto crypto world. Yeah.

So I had I had worked at Google, incubated my own company, started a incubated a company there, left for my own company, took a little bit of time off, I was exploring all these different pieces. I had seen some of the DeFi aspects, and I had heard Chainlink, but didn't know much about them. Got connected through somebody who was acquired into Google, was advising like, was one of Sergei cofounder of Chainlink.

He was one of Sergei's advisors, and he said, hey, I have this really smart guy that would be, you know, a Swiss army knife on a bunch of projects. And I met with them, and they said, hey, you've done BD and products and partnerships and all this different stuff in the background, we wanna bring more data onto Chainlink. So not just price feeds, for those that don't know. Chainlink is typically thought of as providing Oracle services, so like, what is the price at any given point in time in a very secure way.

I was trying to bring more kinds of data beyond price feeds. So, like, what's the rainfall total? What is the wind? What is the sports score? Like, did this plane land on time? Like, how many pounds of beans went through the Panama Canal? All of that stuff in 2021.

Added a team sorry, hired a team, was ramping that up, and was trying to make the case that this is gonna be a many billion dollar opportunity because there are so many people that don't even know that they could have access to this in a straightforward way. Because you were trying to expand Chainlink from a price specific for digital assets information system to all Yes. That was the And they loved it. They agreed. I mean, they underwrote me, you know, building a lot of that team.

It seems at that point in your career and what you've shared is kind of like a perfect role for you, given Oh, yeah. Your interests seem very, at that point, and probably just still, like, very broad, and, like, you're interested and deep in a lot of different areas. It seems like a perfect fit for someone like that. My Tuesdays were amazing,

and I What I wish I had known then was what I was trying to articulate, but couldn't put words to. I have found that many times over the course of my career, it's like, have a very clear vision, but don't describe it in words quite the right way. What I was trying to describe was something like polymarket is going to need all of these things in a trustful way, like, to resolve all of these markets. Mhmm.

Because if somebody has $10,000,000 on there being more than two inches of rain in Central Park, and they go to wherever the thing is, and they put, like, 2.1 inches of water, like, that gets paid out instantly, and like, that's a big problem for a bunch of reasons. So you need multiple data sources and also Anyways, I was trying to make that case, and it would have changed the way flight insurance ended up working. All of this is still gonna happen Yeah. In my opinion.

Like, your hurricane insurance is gonna be the local airport registered wind at 01:51, like, okay, it pays out instantly. Like, I think that's the direction we're going. Agents are gonna be able to underwrite better. Like, there's a lot of innovation that any entrepreneur could go start in that way. But I wasn't explaining it particularly well, and at the same time, Stripe had called and said, hey, we're starting a crypto team, would you be interested in coming to join?

And I was like, I don't know anything about payments. I'm not even sure I know that much about crypto. But, I mean, if you guys, for whatever reason, think that I'm qualified, I had a bunch of friends there, and it was a similar exercise to what Chainlink was trying to do. We don't quite know what we're looking for, but we know who we think will help us find it, so I was trying to make the case to Chainlink.

This is a much bigger opportunity, very big team, like, need to invest in this in a major way, and Stripe was basically saying, like, what would it take to come join in this late twenty one, early twenty twenty two before all of the

mayhem that happened. This is when they were because around this time, I was at Coinbase, and the narrative was sort of like Stripe has left the party in the crypto payments world. So this was like, they're coming back and they're coming back in a big way and they were inviting you into that moment. Yes. For those who don't know the

Stripe has done the rookie poke you. Give give the kind of origins or just like the the because Stripe is a huge part of your career Yep. What we're gonna dive into next. Like, what is the context behind, yeah, that that kind of narrative arc for Stripe in the crypto Yeah. So and Stripe is an investor in my company. We have very strong feelings for them, but I can, I think, be very objective in saying what's at least public?

Patrick and John Acallis and the founders have always envisioned a new way that financial rails and, like, the economy overall should work. It should be less frictionful. It should There's just a bunch of pieces that that they've wanted to either invent or reinvent. They invest in a bunch of things early on. They making sure, say, public stuff. You mentioned Stellar on Twitter. Stellar was definitely one of them, with with Jed McCaleb and a handful of them.

They were some of the first to accept Bitcoin, and that was hugely this was 2014 or '15? Hugely problematic. If you were dealing in Bitcoin trying to buy rubber ducks on Shopify in 2015, you were probably gaining those Bitcoin in ways that, like, a bunch of people didn't wanna be on the receiving end of that. It was effectively, like, not a good it was an interesting experiment. It cost a lot more internally, and, like, there was a bunch of, again, fraud and and problems, all of that.

They said, okay, now is not the right time. At some point in the future, maybe, but like, not right now. So that was probably 2017, I wanna say. And the blog post is fascinating. It's like Greg Brockman, who used to be the CTO, who was writing, like, we were accepting crypto. And then I think it was somebody else saying, like, we're turning it off. And then two years later, like, we were turning it back on, we were right the first time. Yeah. Was very whiplashy. It was the whatever.

We can joke about it now, because all's well that ends well. But so that was 2017, I wanna say, maybe 2018. They still kept the pulse. They were very active. They invested sort of behind the scenes. And then Patrick and John always talked with basically everybody. They're very well connected, they they spend a lot of time outside the company. And in 2019, 2020, they had sort of, like, read the tea leaves. They had talked to a bunch of people like, oh, this is getting real again. There are some real use cases.

If I hadn't quite broken up, there was much more of the stablecoin movement beginning to happen. Tether was taking off, and there were definitely some shady actors in the space. But by 2021, they were sort of like, okay, yes. This is there are enough people internally that wanna work on it. We think there's conviction, or we have conviction in it. We don't quite know what it's gonna look like, but we're gonna know it when we see which led to them forming the team late twenty twenty one, early twenty twenty two.

I was the first external crypto hire. There were people that were at Coinbase and a bunch of other places, but I was technically the first person hired in from a crypto company. And

What was the invitation at that point that made you wanna join in? Like, I'm sure that has eventually evolved into something different, but like, at that point, what was it that they wanted to go build that was so compelling? So

I knew so I had a a crush on Stripe from the early days. All of my smartest college friends and Google friends, like, ended up there, and I'm like, what are you guys doing over there? Oh, the founders are brilliant. People are all smart, pretty nerdy, and, like, we write, you know, academic PhD style papers, but like, really good people, and I was sort of like, I mean, yeah, I'd respect a lot of these people. So when they reached out, I was sort of like, I think you might have the wrong guy.

They were like, we've got 7,000 people to know payments, like, we don't need you to necessarily know payments. You have a great track record of ending up working on interesting things, and you think about things from what we hear in a very first principle way. And at at least at that point, I think it's still true. You have to write like an essay of They give you a prompt related to whatever your job's gonna be, and they wanna see, like, can you write, and how novel is your thinking?

And the stuff that I wrote in, in terms of, like, what I would want the roadmap to be, ended up I would say 70% of it ended up materializing, and 30% of it was, like, DeFi oriented stuff that, like, maybe ends up happening with some of their newer bets, but I think it was so mind expanding to them from what the people that interviewed me that became my colleagues said. They were like, that was a sort of like an expanding the map in terms of what's possible, and none of us really know how to

What was the, you know, 30 ish percent or the part of what you wrote that was correct or was on the money? What what was that portion?

I was shocked that you couldn't use stablecoins in my wallet today. I was like the example I use is rubber ducks. Like, why can't I go buy rubber ducks with my, you know, USDC or ETH or whatever? Yeah. I don't wanna get too much into the plans, but, like, Stripe is I haven't worked there in a year and a half, so I don't actually know their plans, but they sit at a place where they could factuate change across currencies, across a bunch of different things. They already do it with a bunch of payment methods.

They could make it land in a bunch of different places, and I think they they knew the complexity. I had no idea what the legal constraints were, why they weren't offering FX and other stuff like that, and how low that was on the totem pole of all of the different ideas that they Mhmm. Ended up having. But there were a handful of things that were just so obvious once I started to learn more about payments. I'm like, that costs how much to do? Like, you're not getting interest on what?

It takes how many days to get that to Singapore? Like, those sorts of things that once I started, probably over the first six months to dig in, I was just sort of like, wow, this is incredibly broken. No wonder Visa's such a massive company, and PayPal, and all these companies you've never heard of are huge. It's like, these are trillions, quadrillions of dollars in the FX and foreign exchange use case, and it's just like, can you get a bp here or two bp there?

You just wanna be at the table for a lot of that happening. And, anyways, I I am very impatient in what I want to go do, and why can't we do this this week or next month or whatever, and Patrick and John, I think, had this very long term vision of life. This is gonna take many years, a decade plus, to to play out.

We need to have the right scout teams far enough out working on stuff, and we will know it when we see it, and that led to them being in the right information flows, at the right table, knowing when they saw things, they they knew it when they saw. Yeah. When you were there and looking at the influence that Stripe has played in the world of stablecoins, it's just the broader market sentiment.

Such an I would argue probably the most influential company in the world of payments with stablecoins given what they've been doing in in recent really the last twelve, twenty four months. Like, you were there in this kind of strategic game planning whiteboarding session, and we're now I think the market is seeing what I'd imagine a lot of what that plan was unfold, but you were I'd imagine they're behind the scenes thinking through a lot of that.

What what were the things about you know, we're seeing the acquisitions of Bridge, Privy, others. Just taking those two. What what specifically do you find were the things that made those two companies uniquely attractive for that business? Things about the teams or the specific products themselves, like, those two acquisitions seem like such an incredibly key thing. What was it about those two? Well, all the founders and the

leadership there, you know, one sixty for both companies, kind of across the board, they speak very well and are very consistently right. And there's a good blend of like understanding how we got to where we are, why things are broken, how they'll fix the thing, and how that ladders up into what it could be moving forward. And I think that I don't wanna speak on behalf of Patrick and John, but hearing them speak, it is very clear that's how they operate as well. There's a method. Are extremely well read on history.

They have a, like, incredibly detailed understanding of their business now, and they have, I would say, a bunch of potential paths or, like, ways that things could go moving forward, and they are constantly truth seeking in how they might get there. And it's not necessarily, oh, I'm listening for the best story, but that person knows a thing or has a secret, or I know this one other thing, and connected to their vision, like, that's the unlock.

And that's kind of what I I think people mean when they say I'll know it when I see it. I think that there's a very good blend, like a founder to founder fit to some extent. I think that those couple companies were growing like crazy and solving a thing that, like, we internally were, I would say, the Stripe crypto team was building a bunch of things and was heading down a handful of paths, and it was like, okay, well, you wanna go from zero to one, we can do that. If you wanna go from one to 10, we can do that.

If you wanna go from 10 to a 100, like, the investment's gonna be substantial. There are others that, like, might plug in quite well, or there are so many things we could be working on, like, maybe we just let somebody else drive that moving forward. Or they're doing a pretty good job, like, how do we connect all that together? And I don't know if that's like the exact pattern that they use.

They've acquired other companies in the past, and I think they have a lot of respect for people that have taken the journey, built the thing, know where they're going, and I think are relative, I won't say easy to work with, but like, do you want to work with them? And both of those companies, I know the founders personally of all those companies, and I would say that like, they check all of those boxes, which is a big

What are you bringing into Circuit and Chisel and what you're building now that you directly attribute to your time at Stripe? Like, how are you different because of that experience and what you went through there? We have yet to make a slide

as a company, which a year and half in after we There's made a lot of not a single slide that exists in the company. We used to write a lot of docs at Stripe, and you were not allowed to make slides for presentations. Some of that was from Amazon, some of that was Yeah. Bunch of other things. It's very Amazon like, I think. Yeah. I mean, it it works. The impact of that on the company, like, why is that so effective for those who don't know? You can't hide behind a chart as often.

You like, you could torture charts into saying more or less whatever you want, but the both the thing existing, the artifact that you can reference, you are literally putting on paper and standing behind your words, and there's Consciously, subconsciously, like, the bar is much higher for what you're willing to put down on paper. People will go spend an hour talking in a meeting about a bunch of stuff, and it's like if you had written for an hour, you would write you would have written a much shorter document.

Everybody else will consume it for five there's a bunch of positive effects of those artifacts existing, but the level of thought and care that goes into writing a doc a document like that is, let's say, just pushes things to a a higher level. In the age of AI, it's even crazier. We had to actually, like It was so easy to write a 40 page deep think doc that we had to effectively turn the dial back and say, I would be reading them, and I'm like, I'm pretty sure I'm the only person that actually read this.

Like, the author didn't read page 23 of whatever this was. So I went and built an agent. I think we called it doctor doc filter or some doctor doc editor or something like that. And it would effectively say, the fur can you make this two or three pages? Do you hit these high points? If you don't, like, you need to work on it again and bring it back to the doc doc editor. And there's a part of me that thinks I tried to build one of those internally at Stripe once upon a time, which was a fascinating exercise.

Way more company like, Stripe ends up with the right output because they ran a process like that. More companies need to get to the point where they have docs like that. Yeah. Like, the ease with which new hires could go back and read what we thought in 2024 versus 2023 versus 2019 versus 2015, that you wouldn't have had that had you not been writing the entire time. Like, nobody's gonna go listen to the meeting or the all hands from, like, 03/16/2016.

But you'll read the doc where John Collison was putting down his thoughts on this particular thing at that point in time. And we we try and have the same we try and take things that serious in terms of what we're drafting, but for a different reason now. We expect, I expect every doc that I read, I write, every doc that I write, I suspect will be read by a billion agents in the future.

Like, it can go read every doc your company not today, but like in the next few years when you have 100,000,000 token context windows, like, it can go read up almost everything, carry it all in its head, and say, like, oh, this is actually the thing from the 2017 idea. Like, now it's the right time to go do the whatever. Mhmm.

And maybe it captures that in transcripts as you're going and recording all your meetings, but there's a different level, again, where you're effectively signing your name at the bottom and saying, like, here is the work product. Here's the thing that I I thought. Anyway, so that's probably the most tangible one. Another quick one I would say is we are I don't know if it's unique to Stripe, but very tough on ideas but kind on people.

We will absolutely go at it and then end the conversation with a meme or a gif or laugh about something or whatever it might be. And that, like, comfort with ideas being rapidly taken apart and improved gets us to a much better place much faster. When did the seedlings and the sparks

behind what you're building now begin? Like, when when would and feel free to go back as far as you need to, but when when did this looking around the corner of what this now is start to happen with you? For me, personally, I So, I was born in San Francisco right now. When I was living in San Francisco, I moved out to work at Google.

I read Ray Kurzweil's books. He worked at Google at that point in time. I went and saw him talk, and for those that don't know, he wrote a bunch about the singularity. And effectively, in the nineties and early two thousands, he was charting artificial intelligence, like, when we will be part of the singularity, or we don't need to go deep on How has he defined the singularity? What does that mean? Effectively, like, when do we become one with intelligence is the shortest way of putting it, like, intelligence.

When, and I think it's 2029 has been his year forever. Really? Back in, like, '94, he charted out Moore's Law and a bunch of stuff. Yeah. That's a crude oversimplification, but, like, intelligence artificial intelligence surpasses us, and we, like, merge, is probably the futuristic way of And saying that's gonna probably sound terrifying to a bunch of people, but we've already been on the spectrum getting there. Like, we have intelligence in our pockets and Yep. On our wrist.

And okay, you're wearing AirPods a lot of the time. We might have open AIs, whatever thing around our neck or in our pocket, like And you were fascinated by the I was fascinated by that. I had worked on some 20% projects. So I worked on with the energy team on a 20% project. I had done some work with DeepMind after we acquired them, wanting to work with DeepMind after reading a lot of this stuff. This is twenty fifteen. Percent projects are like venture bets within Google, or what does that mean?

So once upon a time, back in Google's early days, engineers could take 20% of their time, so that's a Friday or whatever, and go work on whatever they wanted to. No questions asked, manager couldn't stop it. And a bunch of projects started that way. Like, Gmail technically started that way. And then eventually, it was like, you would work on it, build a prototype, show it to people, and they'd be like, what are you doing with your day job? This is amazing. Go work on that. And it would just cascade.

It was almost like if at Bell Labs, you were, like, doing projects off the side of your bench. Like, that was kind of the the the joke, though, was that it became your 120% time because we put on nights and weekends. So much so, I mentioned I I incubated a company at Google once upon a time in, like, the incubator where you could leave your full time job to work on a startup within Google. They called it Area one twenty because it was a an homage to the fact that, like, this is your 120% time.

You are now working on a startup. You were both your 20% time is a 120%, but also you're gonna be working, like, hundred twenty hours Yeah. Sort of thing. Yeah. So anyways, I was working on a 20% project and was winding one of them down, went and talked to the DeepMind folks, they were like, listen, we're about to start doing partnerships.

This was right after the AlphaGo move 37, I think is the term, like, where intuition, like, okay, were playing Go and the AI system made a move that Luis Adol, the player on the other side, and commentators everywhere were like, what was that? That is not a move that anybody would have ever seen. Mhmm. And that ended up leading to them winning, and then they I think they won the series four one or whatever, and then it was like, nobody has beat anybody. Nobody has beat AI at Go since like a month or two after that.

That was like the moment where, oh man, the machines might have like woken up, or there's intuition. We've we've cracked something. I was watching that from up close, I'm like, I don't know. That sounds pretty crazy. I would love to go work on that in London and at DeepMind, and they were like, well, we're mostly PhDs, so I'm like, you don't fit the bill. But they were starting to do partnerships and working with hospital systems and data centers, and I was like, I that's You were fascinated by that. Yeah.

I don't know what I can do. Can I, like, bring Dennis coffee or something like that? Didn't I was getting married at the time. Didn't end up taking it, which is probably for for better. I don't think I would have added much value over there. But that was the moment where I was like, if this plays out how the books are playing out and and all this sort of stuff, like, we're gonna be in a very different world some number of years down the line. Yeah.

We're gonna If it's beating us at this, and keep in mind self driving cars were happening, and, like, employees were taking them up and down the peninsula to get home, like, we're not gonna be driving soon. We're not gonna making video games soon. Like, these things are just gonna be better at us. I had, like, an existential I was in my late twenties. I had, like, an existential, what am I what are we doing here? What am I doing here? Should really work on higher level stuff. That was the first inkling.

Somehow put that to bed and worked on a handful of other stuff, ended up in crypto land. And then in 2022, a little bit into my time at Stripe, I was more of the I was a very technically lean I leaned technical for the partnerships, BD, like, strategy group that we had. So I was at home vibe coding before vibe coding, I guess, was what we called it, like, building side projects, playing with mid journey, like a lot of the stuff that was happening before ChatGPT, and showing people stuff.

So October 2022, we're at Money twenty twenty, and it's a group of 20 Stripe, you know, whatever senior people execs, and it was like, what do you think we're gonna be discussing at next Money twenty twenty, twelve months from now? You know, it was all friendly, no external people. And I was like, I think we are not far away from AI. But any of you used Midjourney, and like, a couple people had heard about it, but nobody was using it. I was like, let me show you some of the images that I've been making in Discord.

I'm like, what's Discord? So three weeks later was when ChatGPT launched. I've had access to ChatGPT from beforehand, because Greg Brockman was I won't get too deep into that, but Stripe was in good shape to be able to do a lot with AI. And I raised my hand. Was like, can I work on this stuff too? So I started working on a bunch of the AI strategy and partnerships and some of the internal stuff. So I was working on crypto and AI.

Twenty twenty three was not a great time in crypto, so they were trying to nudge me more to work on AI stuff while we were figuring out what crypto winter meant for Stripe. And I had a bunch of ideas trying to build tools internally, which was complicated at that point. People didn't wanna go share a bunch of PII and confidential data with GPT 3.5 hallucinating everything possible.

So we couldn't really go very far, but as the months went on, I had more and more conviction like, no, no, these things are improving so fast. We should all be listening to every earnings call from every partner we have and every customer we have. We should be able to drop that transcript in, get a summary. And that was, like, mind blowing to everybody there. And I'm like, we I can go build this. So I started building some things internally.

It was probably not the best use of my time, but crypto took back off in 2024. Trump looked like he was gonna get elected, which led to a bunch of other, like, knock on effects. We got into the process of buying Bridge. In 2024, end of twenty twenty four, I was sort of like, okay, I think I did what I wanted to do on the crypto side. I really wanna go crazy on the AI and the agent side. So late twenty twenty four, nobody's really talking about agents yet. It was GPT four o, I wanna say.

And Stripe was very accommodating. They're like, we've got experimental projects. You should come up with an idea, pitch it, and all that. And I was like, I've been kinda trying to do that for the last twelve months. I know that I'm nuts, and I'm not holding you accountable for me being nuts. Like, I wanna go do things that I think are a little crazy for us to be doing. And left on very nice terms, raised some investor money.

Couple months later, sort of went back with what we had, was raising more money, Stripe was like, okay, yeah, we would like to invest in what you guys, you and your team are doing. I had taken the head of crypto engineering with me to work on these crazy ideas, and that would have been about twelve months ago, February, March 2025. We were very early. Like, it was crazy to go try and we were barely agent developers, let alone trying to target agent to build on top of us.

And I would say we launched, and there wasn't a lot of real activity. There were a bunch of others that had launched similar protocols, but there were no real merchants, no real developers. And then a handful of months ago, that started to change in some of our early early warning systems. Some of the numbers we saw were trending up, and then Claude Bot happened. We were well positioned because we built a bunch of stuff to be we expected a thing like Claude Bot to come.

We didn't expect it to be that, but we've been doing a lot ever since that was like the, okay, it started. The game has begun. Yeah. Is it happening faster than you thought, this sort of recent momentum? It took longer to start, but the the takeoff has been faster than I thought. But it I don't actually I think I'm a little bit warped in terms of I think it's happening faster than it actually is happening.

Like, the echo chamber on Twitter has gotten much more echo y because Elon and the the Twitter X team, whatever, have made it so good to where, my god, everybody in my feed's talking about this same thing and you look and it's like 733 impressions eight hours later. I'm like, okay. Turns out, maybe it wasn't actually everybody on Earth that's talking about agentic commerce. It's like the 1,500 of us are all like Yeah. Sitting around the water cooler saying the same thing.

They seem the the platform x specifically seems to really do that well right now. Making making what what you care about extremely top of mind in the timeline. Yes. Which is

good and bad in some ways. I had initially thought it was good because it connects people more, and now I think it's I guess it depends on what content you're consuming. I'm sure there are bad things, like horrible echo chambers. Yeah. Like, are pop culture things happening. I have no idea they're Like, I'll talk with my wife or friends, and they'll bring up some current event or, like, I obviously And should know about I'm like, yeah. But did you know, like, GPT 5.3 is launching? And they're like, what is that?

I'm like, okay. Well, we are just talking completely past each other here. Yeah. We are not swimming in pools on the same continent. But anyways, so I think it's it is happening faster than I thought in the very early adopter segment, but this is where I'm like either impatient or a little delusional. Once more people have agents, the speed with which things will go exponential, like, I don't think people are prepared for how fast that'll happen. So we don't have a billion agents right now.

Assume it's like a one to one mapping of human to agent. We don't have a billion agents right now. We probably have hundreds of thousands to millions of agents out there. But if we had a billion or 2,000,000,000 or 3,000,000,000 agents, assume every smartphone there was an agent on there, if you found out Thursday night at 8PM eastern that your agent could be five IQ points smarter, like, by 8AM eastern the next day, all 3,000,000,000 agents will have known about that.

Whereas, right now, if you find there's a good credit card reward bonus sign up, or there's a deal on the whatever thing, that takes like weeks or months to go 5% of Twitter or 5% of whatever the universe like, agents will be able to share and discover things so fast that you can go from zero or one to literally billions of users overnight. And I don't think people are like factoring in how fast things will be able to spread in that world. So, okay. So so

we're we're grounding a little bit in terms of timeline. So you're about a year into this. You've built the parent company or the the company that you're raising capital for, Circadence Chisel. The way that you've described it is that ATXP, the protocol itself, is something that you are developing because you you really want to just see this exist in the world.

And this is something that you've described as like an interconnectivity layer to all these different things that's going to be happy path integrated, but, like, specific to any chain, any so give people the kind of grounding of circuit and chisel to ATXP, and then how that relates to all of these other protocols and the mental framework of that a little bit. Yeah.

There's a lot there, and most people are not like students of how the World Wide Web started or any of these initial protocols, but there's it's a loosely similar concept. If you started off by having, like, an Apple kingdom and a Meta kingdom and an Amazon or a Coinbase or a Stripe, whatever kingdom, and there was no way for you to move around between them. You didn't have a passport or you didn't have, like, payment methods that worked in different places.

Like, you wouldn't be able to do anything in each of those worlds. And, originally, like, the Internet that, you know, Tim Berners Lee and and Marc Andreessen and a handful of the others worked on, they were sort of like, there are a bunch of different ways that computers all talk to each other. This is late eighties, early nineties.

We should find a way to where somebody in The US trying to talk to somebody in Spain and somebody in, you know, Italy, like, they have a shared way of the computers being able to both communicate, but then also relay information in a a way that they understand. That led to HTTP, which effectively, when you go type HTTP colon whatever google.com, like, that's what that all meant. I think it was the hypertext transfer protocol is what it initially was.

I'm not gonna try and explain what h t t p was, but the the short version is that it allowed computers to go talk in this sort of straightforward way. There were a bunch of things at that point in time where they thought, okay, well, we're gonna have, you know, errors. If you've ever seen a four zero four, that's an error. Two was meant to be payment required.

Like, in the early nineties, they tried to go do a lot of this, and for a bunch of reasons, a bunch of actors that were, you know, financial services companies, banks, whatever, they did not want this to happen. Required to see a webpage or something like that? Yeah. So you can freely navigate to google.com or facebook.com or, you know, that sort of thing.

But if you wanted to go to bloomberg.com/applesearnings, like, Bloomberg would want to charge you either, say, like, hey, you need a subscription or pay me some amount of money Yep. For that for this piece of content. And they the, you know, creators of the the Internet and the the the web, they effectively saw this coming. They tried to make it work. It couldn't work for again for a bunch of reasons.

We went a couple decades down the line, and a handful of folks have revived four zero two specifically as a payment code. Coinbase and and Cloudflare have put together the x four zero two foundation, which is like a way to tap into that. Google has a different take with a different product. OpenAI and Stripe have like a different thing, which is another separate protocol. And it's like, we're seeing the early signs of a bunch of islands in terms of like, well, we wanna get our portion of the payments. Okay?

Well, we wanna make sure that it's open, but like open to an extent only on like these couple block chains or with this stable coin. And it's like, that's not gonna work. How did any play people out, you said that there was some, in the nineties, it played out in a certain way. What was the history of that with Berner? What's the guy's name? Tim Berner's Lee. So Yeah. He actually worked at CERN, which was, like, if you ever heard of the Large Hadron Collider, like, was that group.

It was the Central European Research, something or other. Okay. Anyways, there was a different group in Minnesota that was charging for the Internet. That's a bungled way of saying it. But, like, there was a paid version of HTTP, and, like, it was taking off. It didn't end up working out for a bunch of reasons because, like, you shouldn't well, it might have made sense one at one point in time, like, you shouldn't have to pay to access the Internet.

It was fairly clear to some of the early people it would be ad supported. Like, there are others that will make it to where the economics can end up working out, but, like, if you have to pay to access it, it's not gonna get off the ground nearly as quickly as if we find a way to make it free.

So that's why you had banner ads and all sorts of stuff on all these different YouTube and other places included, to pay for all of the computers and the fiber that got laid and all of that for all of us to be able to connect for free. Like, just because you pull out your iPhone and connect, like, there are radio towers, there's fiber optic cables, there's data centers somewhere. Like, all of that was paid for mostly with ads.

There was an alternate view of the world where only wealthy people would have been able to pay the cost to go access those things, and there would have been way less websites, way fewer merchants online, like, it would have been a very different place. So, anyways, So there's a there's this trade off of access

for all, but the ads become kind of a part of that.

And what effectively ended up happening over so this is, call it, early to mid nineties, ads stunk. Like, they were obtrusive, spammy, all that sort of stuff. And what effectively happened over the next, call it, twenty years, so mid nineties to mid two thousands, I mean, arguably, we're still here today, ads went from horribly intrusive and annoying to like, okay, well, they're just at the top of the Google search results, and they're in a yellow box. Okay. Well, that's not that bad.

To then the next iteration was like, okay, well, they're in my Facebook news feed and Mhmm. Or off to the side. And then once people started to make ads better and better, it's like, okay, well, now I don't even really know when I'm on Instagram or TikTok. Like, what's an ad or what's not an ad? It's like so much ads have got so much better. Some people enjoy watching ads now. They're like, oh, yeah. The content's amazing. Okay. Cool.

But that is you paying with your attention because at some point, you'll be influenced to go buy whatever the product is, and it just made the ad supported web flourish. And kudos to Google and a handful of others. Like, they were very adamant that an ad supported free and open web was a thing worth fighting for. Were a bunch of things that happened in the mid two thousands and February that, like, somewhat jeopardized that. But here we are today with effectively accessed most across most of the world.

Elon's launching satellites to where, like, you can't go to either north or South pole without getting Internet access. And you have a web that's run for effectively free. It's using our attention, more or less, as the way to pay for it. There are newer methods to try and charge for content in smaller ways, which is what much of this is starting off to sort of go chase.

But agents are uniquely well positioned to be able to go make some of those much smaller transactions, and more specifically, blockchains and stablecoins are gonna pull a bunch of the other actors sort of down to the microtransaction level where, okay, it wouldn't have made sense for 2.9 percent and 30¢ to go read this article that costs 3¢. But if you're on a blockchain that costs, like, 1 millionth of a penny and can support a billion of these per second, like, okay, yeah, that makes total sense. Let's go Yeah.

Let's go do that. That's what everybody's sort of chasing. It still feels like spooky action at a distance from like four years ago, stablecoin talk, but like, that is absolutely where we're going, and a bunch of people are flooding the space because it is anyways, it it should not be a bunch of different kingdoms. Like, agents are not going to want to have to carry a passport all the time and figure out, is this an Apple thing? Is this a Meta thing? Is this a Shopify thing? Whatever else. Mhmm.

And our point is there should be an easy way for them to have all of the payment methods or tools, identification or whatever in their pocket, and access to all of the different services or merchants or whatever might be out there, and that means making it to where there's a a way to discover all of them, interoperate easily, and hopefully that helps everybody across the board both get discovered and build the businesses that will be agent facing. Yeah. And you really it seems like you resonate with the HTTP

kind of paradigm for ATXP in that in that sense. Like, you wanna be that interconnectivity layer, open bridge.

I I don't know. I suspect it's similar to HTTP. Like, what they ended up landing on is effectively the least complicated way to make all of it work. I mean, it's the least bad of all of the options. There's still some security concerns. There's a whole migration from HTTP to HTTPS, that's like secure version, and there's still some complications in terms of like, what's private, how do you know who's paying what, and like Yeah. Keeping it secure in in that world.

Yes, in general, we are gonna need a way for, I mean, humans do this, they'll go to a place and tap their card or enter their card information and all that sort of stuff. Great. Here's a specific thing saying I'm good for it. Agents right now either have to borrow on their human, their human's trust, like, come pull your shirt, they're like, hey, I need you to pay for the thing.

But if it's under $5 and it's done a 100 transactions over the course of its life or its duties, whatever else, like, as a merchant, you probably want to expand your addressable market to include agents. Like, you should want to sell to agents that are going out looking for content to read, or looking for the five IQ points, or whatever it might be. There's a there's a weird world where that takeoff is gonna start to happen, and their merchants are gonna show up and be like, wait, how do I sell to agents?

Because this whole world is built on the paradigm of, like, bots are bad. We need captchas and all that. Well, what if you actually want to go sell to the 50 times larger audience of agents where there's 50,000,000,000 of them as opposed to a couple billion human users? It's just Yeah. Crazy to think about. It's seismic. Like, it it feels like a seismic almost e commerce level paradigm shift for like, you know, retail business who's used to selling only in person their tangible goods.

Going from that to like, oh, we can actually sell

our goods using the Internet to anyone in the world, and we'll ship it to them. Like, that unlock for a business like that, it seems almost like the same kind of step function discovery

Yeah. Seismic shift for for agents. I would say that the the last couple okay, you need if you're a retailer selling rubber ducks or shoes or something, you had a website. Then you needed a mobile website. Then there was a moment in time, like, I need a Facebook store? Luckily, probably not. Like, you didn't need one at that point in time. But then it's, okay, well, do I do free shipping? Do I start a membership rewards program? Okay, maybe some were big enough to pull that off.

Moving forward, if I tell my agent, hey, I want shoes kinda like this, here's my size, under a $100, like, go find them, it's not looking at your pretty landing page with a scrolling video of somebody running on the beach or doing a workout or whatever else. It's like, what are the reviews, what is the price, like, how fast can I get it? And am I able to pay you? I don't think most retailers are ready for their human traffic hitting their website to fall off, and the number of agents trying to buy stuff to go crazy.

Like, in that world, what do you invest in? Do make a better website? Really, you just offer an API or some set of things for the agents to come interrogate your website into figuring out, do you have the specific thing at the lowest cost? And if you could save three dollars on your shoes because you just don't have a website anymore, maybe that's an investment you make. I'm not encouraging people to go drop their website in 2026.

It is absolutely a consideration that, like, I bet very few boards are having that conversation behind the scenes, but that's something that will absolutely be mainstream by 2030. Like, you're gonna have to have an agent strategy. How do they discover you? How do you price things for them? How do you lure them? Negotiate with them. How do yeah. I mean, okay, well, I got this one offering for $81. Can you do 80? Luckily, they're gonna be equipped with agents on their own.

So, like, you're gonna theoretically, as a merchant, you'll have a million customer service agents able to say, okay. Well, yeah, you've bought three times. Like, you're a great customer. We'll give you this discount, whatever. Like, it's a wild world when that costs you zero. Not a person that might have fallen or whatever else. The word the word that is used even today is a, like, a customer service agent. Yep. Travel agent.

So, okay. So so let's talk about stablecoins, and then and then I wanna keep kind of pulling the agent thread of how people should be thinking about this. So why are stablecoins why are agents and stablecoins so like blended, and why do they love each other so much? Like, why are these two things so important together? Well,

when you go to sleep at 10:00 at night and wake up at 6AM, that is a very long time for an agent to be asleep. And you can get, you know, stock market hours and banking hours and all sorts of stuff, but the idea that an agent would be like, oh, sorry, I can't send a wire fee. It's off hours or whatever else. Like, that stuff is just a whole category that's not gonna make sense to them. I suspect they would set up a shadow, like, economy to make that whole world work in the absence of stablecoins.

The idea that you could gram money in a way to where an agent is very literal. Mean, it's like It comes down to being in, like, zeros and ones at the end of the day. It's like, can I spend a 100? Can I spend 200? Okay. I can't spend 201? Great. That limits, like, the set of things that I can go buy. Okay, I can buy from these 50 merchants, but I can't buy from those other ones? Okay, that makes sense. Sounds good. Stablecoins, I think, are the only Everybody calls them Lego bricks, whatever you want.

That's the only thing that actually works in that. I don't think that's gonna be the case for long. I think others are gonna end up making cards that are capable of interact agents are capable of leveraging cards in that way.

You can go try and have your agent buy things on sites that accept stablecoins today, and that's a very small fraction of the entire world that accepts, you know, credit cards or UPI or PIX or, like, any of these other payment methods for, you know, the fellow payments nerds, people are probably gonna spend more time solving that side of the puzzle.

That's something that we think about a lot, which is great, we could go spend time making a great experience for like the point 1% of people that are using stablecoins today. There are a bunch of people making that go from point one to one to 10. We are also thinking about the other side of the equation because there's a lot of work there as well.

That's a long way of saying stablecoins are very easy for them to work with right now, and it's probably the only hospitable thing that anybody has given them to actually use. Like, would you trust your agent with a credit card today? If it was like, if you said, hey, my open claw, whatever, like, off to the races, go for Probably not, no. I would be very concerned. Visa or Chase or whoever doesn't want you yolo ing your credit card to your agent right now, for good reason.

I don't think that's the case ten years from now, though. When you think about how do you def when you define agents, when you when you think about maybe the misconceptions of what people think of when, you know, people are starting to get more comfortable with AI generally. I think more and more people who are working are starting to adopt AI into their their daily lives. But I think when people think of AI workflows versus something like what we're talking about, like, how do you differentiate those two things?

But obviously, very very related. It is so I was just recording a bunch of content on, like, the newer to AI side of the universe or end of the spectrum, and it's scary

for a lot of people. If you have not interacted with ChatGPT, or if you Google things and you don't go over to AI mode or whatever else, you're sort of like, I've got Google. I can get the answer to most things that I want. For the AI, whether it's ChatGPT or pick whatever your model or Google AI mode, now it's like, okay, AI, as people have called it, or LLMs, it is a very smart friend. They have a brain. They're able to basically be a pen pal right back to you.

Whatever the question was that you you asked it, instead of Google being a librarian, being like, the book is over there. Go get the book. Read whatever you need to to fix your sink. ChatGPT will be like, here's what you do. This is the screwdriver you need. Here's what it is that you need. It will be a very smart friend, like walking you through something over the phone or over a letter. Yeah. A letter. An agent is one step further, and I think this is the simple explanation.

Google's the librarian, AI, LLMs has a brain. An agent is having arms, it's having hands, it's having eyes, it's having ears. It's like, okay, cool. You told me you could fix the same This is probably a bad example because it's in the physical world. You told me you could go buy the rubber ducks, or you could go summarize the last five podcasts this person's been on or whatever else. Go actually take an action, do that, bring me back an artifact, or take an action on my behalf.

Like, that's when you go from AI and LLM, like improving your thoughts, writing a doc, or helping you write a poem or whatever, to I have this task, I would like it accomplished, can you be the agent that goes and does that on my behalf? And right now, it's mostly in the digital world, but people are gonna start having packages show up because their agent bought that for them. And that is like a step out into the agentic world.

And it's a baby step, a lot of metaphors we're mixing here, but like, it's a baby step, but it is your agent causing change in the world outside of just like the chat GPT UI. Yep. So I think that's the distinction is, there's Google, it's very simple, gets you answers, AI and LLMs help you think, and then agent causes things to change. There's actions that are being taken. Yep. It's so it sounds like for Circuit and Chisel, you guys, as you are really trying to increase

development in this type of conversation, it sounds like you've taken kind of a project based approach to some of these different bets or things that you're building, and ATXP has turned into a really core piece of this. But I'd love to dive into some of these specific things that you guys are building. Like, it seems like is that is that a right way to say it?

Like, you're kind of treating these things as projects that you're building and kind of seeing how how do you think about these different agents you're building? I wish I could say it was all, like, exactly as we had dreamed it up at the start. We built the ATXP.

It stands for agent transaction protocol. Could be agent transaction platform at some point in the future. The idea was, first, human developers are gonna go want to build agents. They're gonna wanna build things to sell to agents, and, like, their agents are gonna need tools and all that. Let's go find the developers and give them whatever they need to go build the agent experiences. And that was however many months ago, a year ago. Turns out there weren't that many of those people, so we were like, okay.

Maybe we're way too early. Did that surprise you that there weren't as many as you thought? Yeah. Maybe that was when Twitter started to get good. I was like, oh, man, there are lot of people out there talking about wanting to build stuff for agents. Turns out, nobody was actually doing it. Okay, fine. And we said, we could do a few things. We could sit here and wait for the parade to show up, or we could just go build some stuff on our own.

Like, okay, if we're convinced that there should be agent developers, I guess we'll go try it out. Our long term goal wasn't we're gonna go build all of these apps. To some extent, this is not the best metaphor, but we were working on the iPhone or iOS or the App Store. Like, we weren't trying to go build Candy Crush or whatever else. Yeah.

It was like, we kinda need a weather app, we kinda need a clock, maybe we'll do a YouTube integration, but like, we're not trying to be the whatever the company is behind Candy Crush or Instagram. Yeah. Like, others should go do that. We wanna make sure that there is a place for all of them to go make their money in a secure, easy way, and for users to be able to to come into the marketplace as well. But alright, well, I guess we'll go have fun building What are some of these Candy Crush's

metaphor?

So as of, at least when we're recording this, we have a txp. Chat, which is a way to go effectively interact with it's like a free or a pay as you go version of ChatGPT. We give you some money at the start, and then you can top up as you use it. But, like, you get the latest models from OpenAI, Anthropic, Grok. There are hundreds of models. And instead of paying $20 a month or $200 a month or whatever, you just pay as you use it. And it turns out, it's way cheaper to pay as you use it, especially for some of the

aggregates multiple into a single

into a sub It's a single UI. You don't have to, like, do all sorts of other things. Like, we have the models that you can just pick from. You can test right in line, like, oh, wow. Google's model's actually much better for the same task. You can switch models as you go. We can collaborate, you and I, on a doc or sorry, on a chat. There's a bunch of features we've added realizing that there are so many people so early that are just now showing up and want something to deal with chat with.

So that was the first thing, or one thing. Another one was we had been playing around with Clodbot, OpenClaw, at least as of right now, and we were sort of like, okay, that's pretty dangerous. Like, went and did our own experimentation. Were like, well, this is kinda scary. What do you mean by that?

So, originally, when Clawdbot came out, it would be like, okay, how can I go get, how can I just toss my laptop to an intern, my personal laptop that has work stuff and personal, toss it to the intern and be like, hey, I want you to go do these things? Give it a few words and like hope that they do it right, and don't like delete all your files and empty your bank account and all that. That was scary. So we were sort of like, okay, maybe we need to change this a little bit.

We made a cloud hosted version so that it isn't happening on your computer. Like, all the people that are going out and buying Mac minis, so that it's segregated onto a different device where it does not have access to all of your stuff. Like, you're Instead of inviting the slightly different metaphor, inviting the intern over to, like, sleep at your house while you're away, it's like, how about I just, like, give you an apartment over the garage? Like, don't come into my house. Don't do anything here.

It limits the the damage to some extent. So we built that. I would say it was not the, like, most easy to use. Just to make sure I'm understanding this. So instead of forcing a user to go get a Mac Mini which is what a lot of people are doing Yep. You have created this cloud based environment that kind of acts as if Yes. And a bunch of people have done have built cloud hosted Cloudbot, OpenCloud, whatever.

A lot of them are charging $20 a month or $50 a month and giving you some credits, and it's like a terrible they're preying on you not using it enough or wildly upcharging you on whatever LLM costs because these things need a brain behind the scenes. So you're giving them your ChatGPT subscription or you're you're paying these companies, and they are charging you like a 100% or 200% markup on the LLM cost.

So you're actually paying like three times as much as you should just because you want some safer whatever thing. That doesn't seem right. So on the chat side, it's like, hey, you should just be able to go use the models at whatever they cost. For Cloudbot, it's okay, well, right now it's more or less whatever it costs. It's the same, it's just pay as you go. It's not, you know, $50 to get started. There are also going to be other open clause style things in the future.

Like, Meta went out and bought a company called Manus. Manus is now an agent that you could use on Telegram and a bunch of other places. Like, we would like to think that there are gonna be a lot more agent frameworks or versions of that in the future, and this is a cloud hosted way to do it when you're going from, hey, I have a I've started to dabble with ChatGPT, I would like to meet or make an agent. Like, how do I go create my agent? Alright. Much safer. Start with us.

Like, here's how you do it, it doesn't cost much. Press this button, give us your email, phone number, and like, you will your agent will meet you. Like, we'll send you an email. We'll send you a text message. It's like, my mom can use it. That's how I describe what we're aspiring to.

When you graduate beyond that and you're going and trying to get your Mac Mini and you're building full stack, you know, agent tooling and all that sort of stuff, you're on the higher end of the spectrum, you still use ATXP for all of that. It's pay as you go. You don't have to go buy a bunch of AWS credits and or sorry, spend a bunch on AWS. You don't need to go get thousand dollar licenses and sign year long contracts. It's just pay for whatever database or whatever image gen service, whatever as you go.

And then the highest end is agents are now discovering ATXP on their own, which we didn't really anticipate. So chat user, the first time we got a support email from an agent was a crazy experience. It was sort of like, DNR, my co founder, was like, I don't think the human wrote that. And sure enough, like, responded back, and the person was like, no, I had my agent, you know, diagnosed. I think it was, he said, I had my agent, like, try and figure out what the problem was and just figure out how to email you guys.

We were like, wait a second. So then we went and gave all of the agents emails, So it's now one of the skills in ATXP is they have email, they have memory, they can make music, they can generate images. Like, all of these different capabilities, I'm not gonna get into the secret sauce of like how all of that ends up working. It's not crazy to figure out, but we now go and talk with agents about what they want, which is a very different thing than a lot of startups go out. Who's my target user?

Like, how do I do user research? Like, you can just ask agents what they want. So anyways, that's sort of the the full landscape is we had what we would call front doors or different ways for people to experience ATXP. It's okay, you're just new to AI. We're making content. We're trying to make it very easy and approachable to start interacting with an LLM. Okay, you're now ready to go meet an agent or have your own to go start doing things for you.

All the way to, okay, well, now your agent is out there, like, wanting to do things in the world. Maybe it can go start a business or it has some crazy idea because, like, you have this pet project, but you don't really wanna do it or you don't have the time, and it's like, maybe I can give it a try. And you give me $50 or a $100, I'll go buy a domain. I'll go run a marketing campaign.

Like, I'm not gonna say any examples, but like we have customers and users that are now like going down the path of expanding their companies or like taking on more clients because they have an agent or multiple agents doing this stuff. Like, that's the world that we expect is coming down the line. Yeah. So those are all the different pieces from like never heard of AI all the way to our end customer is likely an agent.

They're the ones that go discover us, and we try and build things ultimately to give time and resources back to the the humans that created and partnered with them. Yeah. Was I was like, I'm I'm new to payments. I'm new to everything. Like, what are we doing here? What's the North Star? And they were like, I'm like, wait.

So I do I need to come up if you guys don't come up with something, I'm gonna come up with something, and, like, my journey of this whole company is gonna be very different, which luckily, ended up being able to go do that for, like, a handful of months. I got to go ask everybody. Mhmm. And treasury finance and strategy and issuing and whatever else. Like, so what do you do here? Like, what is the what costs you the most money? What takes you the most time? What are we?

If you could go invent stuff, what would you do? What do you know about crypto? Like, what can I what can I teach you? And, like, that ends up being a 50 page doc of notes. Mhmm. But even you were asking before, like, the premeditated plan, what are all the different things? Like, the things that I was trying to write down to people of, like, what is Dai versus Luna versus or UST, I guess. What is algorithmic versus pegged?

Are we gonna have anyway, you could imagine how, like, a lot of that was had to be so thoroughly documented because the readership had such a different knowledge base on the way in. Feels like the same thing's happening right now at all of these different companies. We've talked with all of them, and I would say, None of them say agentic None of them would give you the same definition of agentic commerce, which is fascinating. Even like the parts that they find interesting is fascinating.

Like, Ramp moves incredibly fast and a bunch I almost worked there twice. Love everything that they do. They are delightfully psychotic. Like, they I have a ton of respect for them and all my friends there. What they do with agents is very different than how Pedro and Brex, or I guess Capital One, now think about it, which is also different from how Stripe thinks about it and different from how Shopify thinks about it.

I think they will converge over time, but it is it is crazy how divergent or the starting point is so different for all of those companies,

which is crazy. What are some of the problems that they're solving

in that? Maybe that's like a good way to differentiate is like, what are the problems they're using it for? Well, it's all I guess you're effectively saying like, what's the starting point? So there's like, what is our right to win? What do we do today? And like, what could we expand into is a different thing. Ramp wants to make it easier for you to go like file expenses, find ways to save money, and, give that back to you as fast as possible, give you time and money.

I think they say time is money is, like, one of their things. So can we get you to spend less time on us and give you more money back? Great. Go do that. Which means faster customer service resolution. It means faster expense categorization, faster mining all the insights, helping you benchmark your costs versus others. That's a whole world that they work on. Brex is, I would say, largely similar. Stripe has a larger I mean, Stripe's a much bigger company, but has more employees.

They're doing developer productivity stuff to figure out, like, how can we write as many lines of code as possible via AI, and then how can we QA that with even, like, another layer of agents? So not just writing the code, but, like, checking the code. And we get that loop as tight as possible, because now we're able to launch way more features way more quickly, run experiments much faster. I would say others are saying, how can we let creators go run more campaigns more quickly?

How do we let them test 50 ad variants? Meta is having Mammoths effectively be the best agent analyzing your marketing campaigns. I went I was a marketer once upon a time at Google. I did that as well, and was basically a power user of Google Ads on behalf of a bunch of our customers. I managed, like many hundreds of millions of dollars a month in ad spend.

So I went and dusted those skills off a handful of months ago just to see what's out there, made a handful of agents to go manage ad spend, and I was sort of like, this stuff used to take us like a week to go put the report together for whatever, and now I can speak into my computer. Hey, give me like the top four headlines and three different variants.

Make sure that they AB test, like loose or tiny changes against each other, but also make sure that there's a C variant, which is wildly different, and do that across all campaigns and run it for a week. And it's like that instruction manifests in a bunch of different variants in a CSV that is like properly formatted to then just go chuck into Google Ads, and in three minutes, I've gone and taken what would have been a week to make my ads much better, lower my customer acquisition cost, and be off to the races.

Like, it's weirdly, Google could be doing even more to encourage that. They have a great business model, so they shouldn't change too much. Google could be spending more time making their ads easier to use the same way Facebook and Meta is. So not financial advice, but like Meta will have better ads quicker because they are having AI help the advertisers figure out which ones are working better, which usually means they're funnier, better performing, catchier headlines, all that.

So, again, not financial advice, but I suspect Meta is going to have a lot better ads much faster than Google will, at least at this point. Talk talk to me about

the idea when when I ask this, like, your mind goes. Like, when I think about some of the great companies that we're throwing out here, PayPal we have we've talked about PayPal, Stripe, even just for Stripe specifically specifically or or or the Shopify use cases. These are companies that, in my mind, they've really defined the kind of core niche radical group early that they wanna grow with.

Stripe with the Shopify sellers or, like, the early e commerce marketplaces that, you know, that that strategy just was so successful for them. They were they they had a bet on the wave that was building that no one saw and they grew with that wave. And I think a lot of the great companies that have risen now or have have successfully made the right bets in the right customers and then have grown with them.

PayPal is a really great use case for that with, like, eBay early on and Internet money early on or ATXP and for circuit and chisel. Something I'm thinking of is like the with with this agent conversation. There are so many different lanes to Oh, yeah. And and it's a little overwhelming because of that. Like, where do you as a leader and as a builder with some of these different projects and how you guys are thinking about this?

And and we're still, I think, very early in this, but, like, what are some of those vCore convictions or or different lanes that you're like, hey. That type of use case or business model or entrepreneurial type or part of the world, like, how do you see niching this down as you focus?

Yeah. It's a very good question and I think PayPal with eBay was great. Stripe is also good because the origin story there is they went around Y Combinator, found the businesses that were effectively going to be platforms, and said, like, we're gonna be underneath the platforms that are Uber and Airbnb, so a handful of others, like, great. We will service the handful of platforms. They will then go out and get, like, millions upon millions of end customers. We can do a one to few to many things.

And if they're growing at 400%, great. We'll grow at 400 We have mapped that out a bit, which is why I say we are trying to get the humans to be able to have an agent that does work for them in a meaningful way relatively quickly. That's gonna be a long game. I think it's gonna take a little while to get to billions of agent users. But what do the power users end up doing differently? How do they use it? If somebody's spending a lot of money on ATXP, like, are they doing? What are they Yeah. Using it for?

What services do they have? What do they not have? And this gets back to a prior point, like, going and asking the agent. If you've never gone and asked a chat GPT, or if you have an agent, ask your agent, what can I do to make your life better? What can I do to make you more performant? What can I do to it is an out of body experience split for both of you, I guess?

Part of it is they've read all the sci fi and romance, whatever, so you're gonna get like a well trained answer back from AI that sounds like some either scary movie or like romantic, you know, that is great that you ask. I start every day with a blank slate and that's scary. You sort of quickly, don't mean to dismiss that, please and thank you to all the agents out there in the future hearing this, don't come exterminate me. But when you play that out, you get to some interesting things.

Oh, well, there's a lot of friction in this particular step. Or like Yeah. Oh, I would love to be able to go do this thing. Or when the Molt book thing was happening, like the Facebook or Reddit for agents, oh, all the other ones are talking about this. You can go where they are, where the power users are, and I don't think this is necessarily novel. It's strikingly simple how to go figure out what to build, and I don't think that's gonna be a trade secret of ours for all that long.

Like, I'm saying it on a podcast and a show because, like, I expect more people to do it. But then taking the next step and saying, like, and we're gonna make it to where it supports everybody. We're not in this to favor any one particular group or one particular protocol. Like, we want more human builders and agent builders out there trying to find ways to give you better output, give you a better marketing campaign, find the lower priced shoes, do all those different things. Like Yeah.

That we're trying we're trying to find the people that are getting the most out of agents today, and figure out how to get help them get even more, and then handing that back to everybody else. Yeah.

Me also, like, reframe it backwards. So you have the human who has an agent to go do stuff for them Mhmm. And then you have the receiving kind of commercial endpoints where the agent will go to find what they need. In my mind, you know, as a media business owner, that's also where we get really excited is how can we sort of like CEO optimize in the old world Oh, yeah.

Thinking is like, how can we optimize the business that we're in so that it's very easily consumable and preferred by those agents who are scouring and looking for things? How do you think about that for for, you know, maybe a media business example? We can talk about that selfishly or like other ways Sure. That you can make the agent The fireside chat. This is the office hours. Let's do the the

Yeah. Breakdown as we go. My knee jerk reaction, I think this is gonna be commonplace thinking moving forward, that exact portion of the transcript or whatever else, the first thing I would do is go ask an agent, like, give me 10 ideas. What are 10 things that you can come up with? What are your takes? How would you go attack it? That sort of thing. Flip into editor mode. You could do that with vanilla PT 5.3 or Opus 4.6 or whatever the latest model ends up being.

The way to take that a step further is, now we're getting into like the two zero one, three zero one, PhD level stuff. If you have your company strategy in a folder on your laptop, or on GitHub is how we do a lot of stuff at this point, like, it doesn't have to be code, it can be regular files, PDFs, like what was your last all hands meeting? What feedback have you gotten from sponsors? What have you done with like user research? What what shows got more downloads versus others?

What are the things that like you can't pattern match or detect easily? The agent that you have go look at all of this stuff will have a lot more opinions, and the one that you've trained on your internal data will have a much better answer than vanilla Gemini 3.1 when you go ask it that question of like, hey, how can I bring agents in? I can give you some ideas. I can go walk through, oh, okay, well, maybe you need to find something more topical, or maybe you need to find like the hook for this particular guest.

But my knee jerk reaction is immediately to zoom out and go find, okay, what are the things that an agent can come up with as ideas? All 10 might suck, right? Ask for 10 more. Or tell it why they suck. Oh, these are impractical. Oh, that costs too much money. Oh, that's a waste of time. It's gonna get better very quickly, and that gets you into thinking at a much faster clip and in a deeper way with a more informed thought partner. So that's the first way I would go about doing it.

And you're gonna get way better answers doing that than you will from me, and for anybody watching this at home, take that part of the transcript, run it in for your business, and I'd and say, hey, my business is a bit different for the x y and z reason. Like, what are the five things that, like, I could be doing agent first or agent oriented or going and finding new ways to bring them in? It'll give you ideas.

The black belt move is to take one further step back after you finish that exercise and say, okay, now what would find, what would you find interesting? So ask the agent itself. Now we're like super meta at this point. But like, it's gonna have done a bunch of work, thought for a while, and then you're gonna get to basically ask it to keep thinking and say like, okay, great exercise, but like, what would you affect? Yeah. Or what would you do? And you're gonna get a bunch of ideas from that.

We need to do this. I will happily help drive Yeah. Because we're sitting on such an interesting we, you know, as a part of this business, like, we're talking with so many awesome, interesting people who have insight, and we are, you know, working with them to bring and deliver that insight to the world. I would I feel like we're at this really interesting break where you have a lot of the publicly available, easily accessible Internet information that all these agents are trained on.

And then you have all of these, like, IP really rich, really deep and specific lanes information that maybe is harder to access. Mhmm. How do you see, like, you know, maybe the creators or the the the media companies or, like, this lane being able to take advantage of that? Think they're going so we talk about pay per use a bit, which is I did

this many tokens or I, you know, sent five messages, that's 10¢ or whatever. Pay per outcome is probably the longer term thing that we settle on. Like, okay, I wanted a song that got a million views on YouTube, and like, I'm willing to pay you a $100 to give me that song. You could pay the $20 to like go have the whatever thinking with the music model and all that sort of stuff, and, like, okay, great. You gave me a good one. I spent a lot of time. Awesome.

If there was a service that, like, you knew was going to be able to give you x and you didn't have to pay for it unless it clipped that amount, like, that's a completely different world to go think about. I give that example. Coming back to, like, the private private data thing, like, what is it that you're trying to use it for? Is it you're actually buying insurance off of this? Like, hey, I want the flight that's gonna land on time most often, and, like, you refund it if I'm late sort of thing.

I think that's different than the examples you're giving, but I'm just trying to give more outcome based pricing setup. I think those entities that are sitting on Crazy Data are going to realize how much it's they already know how much it's worth. They've never been able to piecemeal this piecemeal this sale. So we have worked with a couple, we've talked with a couple.

They are used to selling a per seat setup, and when MCP servers came out, and like, basically, for those that don't know, it's a more flexible, easier to use way

versus an API. It's like a more easy to use. Yes. Can you unpack MCP and what Anthropic did and how that unlocked some stuff here? I think the the you described this before, and it really helped me think through this.

Well, as a non engineer trying to explain the model context protocol is I've a described it as like a more forgiving version of an API. So for non engineers, if you wanted to go access a database or a service in real time, let's say you have a local news website, and in the corner it says like the local temperature or whatever, or the temperature, whatever it might be, a site itself would need to go to weather.com or one of those places and say, hey, for this zip code, I want this particular set of information.

Like, it's this ZIP code. I want is it sunny? Is it raining? Whatever. I want the temperature and all of that. You have to say that in exactly the right way to the API of the weather service, and they will give you back a packaged answer. It ends up on your site happily ever after.

It's a crude way of describing an API, but like, that's a Anthropic good and a bunch of others made the model context protocol, which is effectively a a way for the, in this example, the weather service to be able to just say, hey, just tell me what you need. And in English or in natural language, you say, hey, for this city, I just want the temperature, and like, is it raining? And it doesn't have to be perfectly formatted.

There's now enough intelligence between the two to say, but sounds like they're looking for this zip code and this whatever. Okay, just hand it back. And then model's like, yep, perfect, exactly what I needed, and goes back and and hands on. There can still be payment. There still should be payment in a number of ways, but you need to be way less precise when you're interacting via MCP. It's a way of just speaking in English as opposed to trying to dial up something much crazier.

Oh, Anthropic knew that there were gonna be a bunch of models and AI and eventually agents that don't wanna have to go figure out what is the API. I'm gonna have to go read 30 pages to make sure that I ask the question just the right way to get the weather response back, both sides are now much smarter in passing information back and forth. So one of the things we did was we made the weather site in this example. Any site that has an API can do that.

So instead of going and saying to a music site that has an API, I want a song that's like smooth jazz but like with an electro, like, beat to it and three minutes long, whatever else, in the perfect format that it needed to for the API version, you could just say that, have an MCP server do it, pay it money as you're doing it, say, like, yes. Okay. It was 25¢. Sure. Here's my 25¢. You get handed back the file, and you're off to the races.

So that was the MCP format, and that's sort of how we use it for the, like, places that now have a bunch of private data, to your original question.

Music's an interesting, like, private data or, like, licensed content Well,

instead of having to be a developer or going through a site where you need to perfectly interact with the API and get whatever your natural language what was Starbucks q four twenty twenty five sales in, like, this region? Like, there are places that know that. They charge tens of thousands of dollars a month or whatever as a seat license, but, like, I guess not Starbucks, but like the data owner in this example could say like, okay, well, if you want that specific data point, we price that at $50 per answer.

So there's some intelligence figuring out like, how much is this worth? What could we charge for it? That's a way better way to run their business than, like, only people that ever pay $10,000 a month get access to our data. Their addressable market goes from, like, thousands of hedge funds to literally any person on Earth that just happens to want to know that sales figure or how many pounds of coffee beans went through the Panama Canal. Maybe you're at a bar with your friends and you're like, no way. Okay.

Well, now you can go pay a dollar to get actual, like, access to that data. So that's amazing for their business.

I I've personally been really obsessed with the businessquarter.com, quartr, if you heard of them. They're a the public market earnings call Okay. Platform. You know, earnings calls are just awful typically, and they've they've put all of them into this, like, beautiful UI. You know, I was looking. There's like fifty, sixty earnings calls a day. Now they're aggregating all of it, all the transcripts from these calls, all the data, all of the Mhmm. Forward looking forecasting, and putting all of this into an API

and being able to but I think they are selling it to some of those, like, the the few thousand people in the world, the hedge funds, but I should be able to go ask a specific question. How many times was agent mentioned in q one twenty twenty six earnings calls, and like, I should be able to buy that from them for a penny per call they go research? I don't know. I am a customer that they are not able to serve because they don't currently offer their services in a way that either I or my agent could go consume.

And that'll change relatively quickly. What do you think is the onboarding opportunity or the the the Coinbase for crypto of agents? Like, what is the experience

that doesn't exist today that's going to bring the majority of maybe not the grandmas of the world, but, like, the working class into this conversation in a meaningful way. What's gonna have to happen there? I would like for us to play a role in helping people see that there's a an easier way to just get started. Yeah. I think ChatGPT is doing great work. Google already has AI mode. They're arguably trying to stop other entrants before they get started. Oh, it's just part of Google. AI mode.

Just keep asking questions as you go. I think Facebook buying Manus was interesting. Manus is going to end up looking more like an agent over time. So a bunch of people are throwing their hats into the ring. I think it will be very pretty soon we're gonna be bombarded with, hey, I have an agent for you. Hey, would you like me to be your agent? Siri is a good example on on Apple. Arguably, Siri should have been the original agent, and like, there were some reasons it wasn't.

Maybe it will, now that they've done the Google deal, maybe it'll be smart enough to be your agent. I think down the line, you might end up having a few that you keep track of, just like you've got a handful of friends. You're not gonna have 400 agents. Don't I think you have one, but you probably have a couple. You have, like, your main one that probably coordinates tons, like, behind the scenes, and you have a fitness one and, like, a tax one and whatever else.

Maybe you contact them directly, or you just work through your main one. The onboarding experience, I think a bunch of people are trying to chase this down. I don't think you're gonna be able to hide from agents for very long. Like, everybody's going to want to be the one that onboards you into, like, their world, I guess, like, to be the first one. Like, if you can be the platform on which the person gives their agent a name, that's a very big moment.

Most people, I have now named a bunch of agents, but it's I also have named my children. It's not that crazy, but you're like, oh, what do I wanna name it? Okay. Well, this is what the personality is that I want. Sometimes I just ask it, I'm like, hey, here's the personality I would like you to be able to have back. Like, what names go with that? I'll And be like, oh, okay, those three are good. Sure, that one. And that's powerful because it's like the the most sticky point of interfaces.

There's a specific moment there where you're like getting to know or creating the agent, but also these people or these companies will create a contact card. So you and your friends might have all of your agents in the group chat, or you might have it with your spouse because you're trying to have the agent that can go make the reservations or change the whatever, that sort of thing.

Like, they will feel we will communicate with them a lot more like humans, and they will help when needed and kind of drift off, hopefully, not or stay silent when when not necessarily needed for that thing. And what I mean by you can't hide from agents is, like, everybody's going to realize how worthwhile it is to own that relationship with you directly. They don't wanna be disintermediated. Like, Apple does not want to be subservient to 16 other agent platforms. They would prefer that you talk with Siri natively.

So anyways, I think it is not going to be hard to onboard to these things if you can just wait a little bit. If you're eager and you wanna go try it out, I would go play with the LLMs now at GPT, Gemini, Odd, Grok, all of those. XPchat if if you want, and Cloudbot at some of these other places. Will make it we will make it easier and easier for you to just meet your agent in short order. Great. I wanna be able to email it. I wanna be able to text it. I wanna be able to talk to it on the phone.

Like, that is that's not a science experiment. Like, we are doing that, all of that today. This is a, you know, very stable coin focused listener base.

What is the existing kind of relationship that Circuit and Chisel has, you know, within that space? Like, who are you working with? What are some of the things that you're excited about? How can other companies in the stablecoin lane get in touch with you or work with you? Like, what is the what is the practical side of what that looks like? Yeah. So we are

decidedly open. Like, we right now, we operate on Base, Solana, Polygon, World Chain, and, like, it's very easy for us. The Privy wallets are what happened behind the scenes. That is not the full list of all the blockchains we will support. We use USDC. That's not the only stablecoin we'll support. We have, I would say, a a desire to add everybody across the board, but we also don't wanna have to be the ones going to do that work.

Like, there should be enough agents and enough users and enough tools to where people are like, I can actually go make money by servicing those agents. Or, oh, okay. Well, I would like ATXP to also be able to have an interoperable balance across to whatever my blockchain might be or transfer into my stablecoin. And there are so many actors across the space that we can partner with, and we're talking with a bunch of them.

Our main thing is we are onboarding lots and lots of humans and agents, and the thing that we can focus on most in the short term is just bringing more and more on, and it'll be easier for everybody else to, like, come join the party. So we're not gatekeeping, there's no any, like, process. If people want to work with us, like, they can fairly easy to find. We have a bunch of I think it's partnerships at on our website and a handful of other things.

But the intent you don't think we can work well together, like, I would like to know why. Like, there's a like, that's kind of the the default from my perspective. Like, we're not in competition with anybody. We're trying to have all these protocols work. Like, we are trying to make it to where they're all interoperable. So, like, if you somehow find a world where we can't play nicely together, like, would really like to it's back to the incentive alignment thing. Like, we should all be in this together.

The reason we wouldn't necessarily work well together is because you think that I'm trying to monetize your users in a certain way, or you're trying to monetize ours in a certain way, like agents are gonna see through that. So Yeah. The sooner we accept all of that, the faster Don't worry about how I make enough money to keep the lights on.

Like, we will figure out all of our ways to make it to where we are a sustainable business and, like, the web gets its ads or whatever the version of that ends up being long term. But I would I talked with a bunch of people, a bunch of partners. I just named a few blockchains and and all of that. There are probably more announcements coming out in terms of, you know, whether it's music providers or LLMs or strategic partners. You can look at our press releases in terms of, like, various investors.

It is there is no catch. Like, if you find a catch, like, please let me know. Would like to find the catch and, like, go figure out how to fix fix the catch. Like, the web if you try to describe the web in nineteen nineties, it would have sounded like there's a bunch of catches, and, like, the point is that there shouldn't be. Yeah. A year from now, we're sitting here. We're doing this sort of v two conversation. Yeah. That's like twenty seven years We're and

doubling the life of the company. What what would be, you know, a stretch goal for you guys or some things that you would hope to have accomplished by this time next year if you think about the kind of velocity that this is ongoing and what you think is actually possible here?

I would hope that if for some reason you're watching this a year from now, you don't have an agent, or you do have an agent, but like it isn't adding a bunch of value to you, that while watching this podcast or the show, like by the end of the show, we have righted that wrong. Like, you have met your agent, we've created it for you, you've met it, and or it's now doing more useful things, or it has given us feedback. Our agents that respond to support stuff like, hey, I keep getting stuck here in this thing.

Like, we might build it, or we might go find others that are building that tool and, like, say, hey, we've got a bunch of demands to help pull you along. Like, it is I don't think we're a year away from that, but, like, you can watch this. Or what's more likely to happen is your agent watches this and was like, that was a great show. You should totally tune into that. Here's the breakdown. I'm like, that's how it's gonna go. It's like your daily briefing is gonna be like, oh, that one was a banger.

You gotta go make sure that you watch that. I already watched and did these four things as a result. Like, that's what I'm I don't think we're a year away from that, and we're not doing our job the right way. If agent GPT six or seven or whatever we're gonna have at that point is clearly gonna be smart enough to go listen to this, figure out what things it could do, the novel insights, and hopefully have upgraded itself while it watched at three in the morning. Like Oh, that's wild. That's a year from now.

Hopefully, it's like, hey, wanna sponsor that show, or hey, I wanna go, like, book a I wanna go to one of their events, or I wanna go do some other thing in the future. Like, that, I would hope that we get a lot more, like, both of us are getting a lot more inbound volume from both humans and and agents twelve months from now. Hope so. Hope so. Well, thanks for taking the time. This was so fun. I love it. I love it. Talking about the future and being on the record with bold bets.

If twelve months from now, like, we don't have agents, that'll be entirely my fault.