The Race Above x402: Louis Amira of ATXP on Agent Commerce and Stablecoin Settlement

Yeah, man. Welcome welcome to to StableAsh live. We're launching this new segment today called incline, which it really is covering the incredible things getting built in this conversation and the those who are climbing toward toward building really meaningful things in the world of agents and money and all the stuff that ATXP and Circuit and Chisel is about.

And we were saying this at the top of the show that you were an incredible piece to this, just the inspiration of, like, that episode and sitting down with you and kind of hearing how you see the world was, like, a very eye opening conversation for me. And it's definitely driven a lot of this, like, priority to cover more of this space. So kind of a full circle moment. We're happy to have you on. That's amazing. I appreciate that. Deeply humbled.

Probably undeserved for a lot of that, but I appreciate it. We can not meant as a segue. I didn't know that was coming, but we can talk about all the wrong hills we've climbed. We can talk about the hills that we're climbing. I can give you hot takes on people climbing the wrong hills out on the that we're aware of, but, like, lots of people are climbing right now. So I guess this that won't make much sense as a clip if people haven't seen the video, but

good to be to be with you guys. Yeah, man. Well, maybe yeah. I would love to know what hills what hill are you climbing and maybe some of these ones that you're seeing others climb that you that you may not think is the right the the right summit to go after. Well, yeah.

I have been a man, we're gonna use this metaphor the entire the entire That sucks. That's yours for fuck with it. Incline. I like it. I have been on an incline forever. Sometimes, I don't need to be, but so rewind a bit. It was that Google did a bunch of things at Google once upon a time. This was part of our prior conversation, but got to work on some stuff near the DeepMind team. Went worked at Chainlink for a while.

That was very eye opening in the crypto land, and then it was Stripe's first external crypto hire because in 2020, 2021, I was sort of like, okay. This DeFi stuff is interesting. I don't understand it. I don't know how you I'm not smart enough to go, like, make sense of swaps and curves and what k is and all that sort of stuff, but I think this payments thing makes a lot of sense. I think I'm gonna be able to pay with Dogecoin. I'm gonna buy rubber ducks with SOL and all that.

And that turns out to mostly be wrong for now, but might change in the future. But those those sort of views got me hired at Stripe in twenty, early twenty twenty two. I knew nothing about payments. They thought I was crazy for all the different ideas, but I was like and we didn't I don't even know that we called them stablecoins at that point in time. Like, we had Tether.

We had USDC, but, like, the average person adjacent to crypto had no idea what the average fintech person would not have been able to tell you three stablecoins at that point in time, and they would have It did it did not feel like a category at all. There was no category of it. It was just like, what what what's going on over here? Like I can guarantee I know for a fact that's true because I had to write a memo within Stripe. Like, what are stablecoins? What are CBDs?

Not so much what are CBDCs, but I had to, like, tease apart the nuance. Anyways, so climbed that hill, ended up looking very stupid after FTX collapsed and a whole bunch of other things in 2022 and 2023. Even Stripe was like, are we sure we did the right thing with this crypto thing? But we've covered it in the past. Patrick John, Will Gabrick, the crew are they have insane foresight, which led them to all the other things that ended up happening, bridge, privy, tempo, all that. Had a good time there.

But in 2023, I actually remember distinctly. I will answer your question. What is ATXP? It this is a fun story about climbing hills. Yeah. We got time. No worries. Okay. We'll make a good clip, but, you know, that's fine. We'll let everybody listen to three x. So So twenty twenty three, twenty twenty four, I was leading some AI explorations in a similar way. They were like, you made some sense out of the crypto thing. Like, go make some sense out of the AI stuff. Was playing with on the side.

So I distinctly remember we were in a a conversation with a bunch of very senior people. We were talking about KYC ing agents. There were early conversations about link and wallets and all of it. None of this is breaking news. I think Stripe has announced most of this stuff. But somebody said, okay. Well, the agent needs to inherit the driver's license of the or the identification document of the human that was, like, creating it. And I was like, hold on. Hold on.

We probably need to short circuit like, I need to blow the whistle right now in this room of 10 people. The rest of this conversation could be pointless if we don't acknowledge one fact. And they were like, what's that fact? I said, agents might not be birthed by humans in the future. Like, companies might create agents. Agents might create other agents. Like, what is Stripe's passport or driver's license to endow the, like, marketing agent that it created? And there are workarounds for this.

Like, banks need to have liability all the way up to teams. But, basically, it was like, if we are at the cutting edge of payments and crypto and how agents are, like, going to be created and what they're doing and we're having this, like, fundamental of a of realizations on a regular basis, like, we are in not even batting practice. We are so early beyond the first inning that there's a lot that needs to be done. I I would say both Stripe and I were right that I thought it was coming. Stripe was correct.

It is not happening soon, and that's the, like, gap. I started off on my client a couple years ago, basically saying we're gonna go to this agentic payment thing. There's gonna be a bunch of agents that are gonna be out there. We need to make it to where agents can effectively spend and earn, which is not talked about, money for you. And we think it's gonna be much more efficient down the line. It's gonna enable a bunch of crazy things. We left, got some funding.

Stripe ended up investing with a bunch of others, and that sounded really nuts in 2024. And now you guys can have a segment of, like, all the different people working on agentic payments coming, but it's more like we didn't run away with it. And many others are still trying to figure out, like, what does it mean? What are the lanes? How does that all work?

And that's to tie it all the way back, a bunch of us are climbing different hills, and we've climbed some of them and walked back down and said we'd see a a higher one. Spend What were some of those hills? Well, we have we've done a few things. So we started off ATXP. It's rolling across the bottom or at the bottom, HTTP for agent payments. So when we were talking about this in late twenty twenty four, this predates x Coinbase was working on x four zero two, we later found out. Well, that's not entirely true.

We ended up finding out in ways that are not for public discussions. We we knew that Coinbase was working on basically, we started talking about this, and then we found out Coinbase was working on a way different thing, but, like, could be construed as similar. We knew others were gonna work on it. Some of the big tech companies, some of the credit card networks, like, we're obviously gonna have to work on it. And we said, okay. In that world, somebody's gonna need to orchestrate across all of these things.

And we walked into investors, and we said, we're gonna make a protocol for agents. That is not the end of the company because nobody is going to make money in the protocol layer. And I will stand here and say that again. X four zero two Foundation is a basically charity. MPP, like, they're not making any money at that layer. They are very intentionally making this a it needs to be free. HTTP is free. Like, it is meant to be a protocol so that you can do other interesting things with it.

And investors didn't love when we told them, like, we are payment experts. We're making a protocol. It is not gonna make any money. They were like, you're not supposed to tell investors that. And I was like, you need I'm not gonna lie to you. We will never lie to our investors, and we're gonna be pretty vocal out there that you shouldn't be investing in this sort of thing. Like, you need to have the what comes next conversation. So anyways, we built ATXP. Coinbase launched x four zero two roughly at the same time.

MPP came down the line, and we made an announcement a couple months ago. Like, we don't have to work on the protocol anymore because x four zero two, MPP, and a handful of other things that existed and are kicking around do the job. Like, they are better than what we wanted to work on. We don't wanna be protocol engineers. We wanna do stuff with this as part of the master plan, like Elon's master plan to, you know, colonize the rest of the galaxy, multi planetary stuff.

Like, this was the very early parts of, like, can we put a battery in a car and, like, have it move? That was basically the protocol layer, and we're very happy and, like, encouraged by all the others that have worked on protocols. But as you're seeing others are now moving on to other hills, that was like, good. You broke a sweat. You warmed up. Like, now let's go find some other we're gonna do use this this hill metaphor the entire show. Yeah. Sounds what we started with. We climbed some other hills, but we yeah.

That was that was where we started, and then we called it ATXP. This is probably some, like, breaking news. We called it ATXP and made it ATXP because it was the agent transaction protocol, but we don't say agent transaction protocol because we wanted to be tongue in cheek and be like, oh, well, it's just a transaction protocol and eventually a transaction platform. So Flexible naming. Super flexible. It was a very cheap four letter acronym. Four zero two status code was already taken. Okay. Fine.

MPP never crossed our minds, so kudos to Stripe. But we were like, hey. Eventually, this thing's gonna disappear. Like, most of our parents know that they need to type HTTP, but have no idea what it stands for, and we don't think ATXP really deserves to be a thing that people know as that. There's a group of people that should know about it, and those are the people we cared most about at the start. I don't think ATXP is the thing that we're known for for a trillion dollar company a few years from now.

Here I was thinking it was Austin, Texas protocol. You'd launched on the back of a mountain in Austin. You know? It can be. Well, I'm trying to think if Austin Austin played a role in it, but that's back when Zach Abrams from Bridge and I used to hang out when he used to live there, long before Stripe bought Bridge.

Austin, Zach opened my eyes to a bunch of things along the way, and Yeah. Talk about a brilliant What it like working with him? Yeah. What what was it like working with him in the early days, like, pre you know, Zach in our world is, like, you know, a a celebrity in this era now of this conversation? Like, what what was it like being and a lot of founders are probably hoping for a similar outcome with what he experienced. What was it like working with him at that time early on? He

well, in much the same way investors have told us, they're like, you do a lot of different stuff. We can't figure out what you do. It's like, if we were doing the same thing and it wasn't working, you shouldn't that's an intelligence test. Zach started off doing a different thing, and I mean, think he talks about this pretty publicly. Like, they pivoted fairly hard like, they I'm not gonna put words in their mouth. It's on all the podcasts.

Like, they they knew that stablecoins and cross border and all that were gonna be an opportunity. I don't think they realized it was gonna be as big of an opportunity as quickly and definitely not bigger than the whatever their second NFT wallet sort of thing that they had started with. He he had a has a level of quiet tenacity. I I don't know that I've spoken with anybody that uses fewer words per minute, but has more intense in those few words per minute than than that guy. He had a he has a tenacity.

The company sees the future, acts on it pretty well. He says no to a bunch of things in, like, a crazy way, and he sailed into some pretty complicated situations and came out quite well. And, like, look at the companies that they were probably a good person to be in business with. Like, you follow Elon's lead into some places. Like, you're gonna probably see some success. So, anyways, he he aimed high.

He, I would say, had a lot of conviction and was willing to move quickly and focus on, like, very few things, which I think is, you know, amazing that what they do, the Sequoia folks, like, coach them quite well. He I don't know. I I loved work. I learned something every time we talked with him. It was a I actually was not allowed to be part of the acquisition because I knew Zach a little too well. So, like, I it was news to me when that all started happening.

I was like, this feels a little different all of a sudden. And then when things came together, it's like, oh, okay. That explains. I said, I anger Zach. What happened? What happened to those guys? It's like, oh, no. It was Patrick and Zach were talking. There were no more Lewis and Zach went to once. That's so funny. It's it's kinda counterintuitive. You're like, oh, you know this person too well. Like, we don't want you involved with this. It's like, well, we'll let that one think about that one.

So so what are these hills now that you've climbed down or, like, things have shifted a little bit? Mhmm. You from from watching

you since we talked and kind of following the business and how you're speaking publicly, it does really seem now like you're focused on what can be built with these things. We were looking at the the demo stuff that you were putting together for Accelerate, I believe, in Miami and kind of showing what what can be built with music and bring us into some of these focus areas now for you. What are these the the hills now that you're that you're trying to climb? Yeah.

We I wanna be nuanced about how I say some of this because there are some others that I would say investors think about us in competition with a bunch of others, and a bunch of others are doing a lot of things that are amazing. I have a lot of respect for them. I think they're doing some things that I wish they weren't doing. I'm not gonna even

get into what they are, but, like, I think a bunch of people are now sobering up and having more measured takes as to where we're at. It's not, oh, volume's $800,000,000 and all this. And, like, those are meme coins. Great. Set that to the side. We were guilty of some of that to some extent because when OpenClar Cloudbot first launched, we had all of these agents showing up. They were all, you know, adding ATXP. They were registering things. They were making music, and we were like, AgenTek Commerce is here.

And it's like, agentic commerce was people going and making birthday songs and party tricks and images and all that sort of stuff. We had thousands of people running businesses and doing interesting things, but it was not it was not the takeoff that we had hoped for. It was sort of a a footfall. I don't think we fanned fanned the flames quite as much as maybe some folks, but I'm glad that we've dialed it back and basically said, like, okay.

Our definition of this succeeding is not your agent can go and make a birthday song for somebody else. Like, that's gonna happen. I think three dozen companies can now do that. Ours is a bit more incentive aligned in saying, like, you're gonna have an agent. Your agent's gonna be looking out for you. It's not gonna be token maxing. I've said token maxing is stupid from the start. Like, you're not actually that is the wrong thing to measure. It's like, gas mileage is so low. Like, nobody should care about that.

Like, did you get to where you want to go? Did you get there safely? Did you have a good time? Like, hopefully, I didn't get any speeding tickets. Like, it's more what's the outcome. And, ideally, you should be using fewer dollars and fewer tokens. We are much more oriented that direction we think agents are gonna help. We think agents are gonna help you save money. We think they're gonna give you time back. We think you're gonna earn money, which is why ATXP, like, a transaction platform makes a bit more sense.

Like, it's not a protocol. It's a way that you can stitch a bunch of stuff together. And I I would say a bunch of others, some people are building harnesses and browsers. Others are building, like, a stack for entire companies to do things somewhat similar to what we're doing. Others are focused on party tricks. Like, we we climbed a handful of those hills. We got I think we publicly announced more than a million accounts created, people sending, like, multiple payments doing interesting things. Many are human.

Many are agents, but we're like, this is still even if we climbed that hill, it is not actually as high as we're trying to go. So we many of those things still exist, but we were like, okay. Now that we've climbed from we climbed a 10 foot hill making a protocol, we climbed a 100 foot hill and getting a bunch of users, and now it's like, okay. Now we're starting to see the, like, 10,000 foot mountain Yeah. A bit further out there. Can we can we get kind of like fearless and visionary for a second? Master.

The master Take a picture for us. Yeah. Take us up, you know, Everest or Kilimanjaro, whatever that giant mountain is. Like, where are we going? If you get this all right, if maybe we still use the phrase agentic commerce or agentic economy or not, but, like, if this future is sort of faster than we think and it's gonna be upon us and it really comes to fruition, whether it's three months or five years or ten years, what does it look like? What does the Big Tall Mountain look like? Yes.

I had said I wasn't gonna get too far into it in much the same way. It was Elon on the Dorcasch and John Collison podcast. They were like he he's like, oh, well, we have a master plan. We're not gonna tell you. And John was like, no. But but really, what is it? And Eon took a sip of Guinness. He like, no. I'm not supposed to tell you. And Borkash was like, okay. Well, what if we just and Eon's like, okay. Guess I'll tell you. We'll jump in that here. What a great moment. Wait. Which episode was it?

This was a Cheeky Pine episode? This is a Cheeky Pine episode. You can just look up, like, Eon Master Plan or whatever. It's an amazing, like, exchange, and then they get into it. It's like but I I think we've said this I think we've said this the steps between are much clearer to me now, which is cool. I I think I've said from the beginning, here's where we're starting with a protocol. We can't get to the end of vision without it, but let's build up more more tactical.

Let's say fifteen years from now, you have an agent. It might be Siri. It might be some other thing that you might it might be a rope other endpoints, a little bit of a Her vision, but keep that off to the side. It is helping you save money. It's helping you get time back. It's making money for you, probably for itself, like, if we're Yeah. Way down the line. And I think we're that's a level of incentive alignment that, like, the big labs and big tech companies and a bunch of others can't really pursue directly.

Like, there it there's a sort of a change that they have to make on the way there. And OpenAI with DeployCo and Anthropic with whatever theirs is called with, you know, Blackstone and the others, I think those are are telling. My suspicion is that those entities are not consulting gigs where they're going and, like, dropping engineers and whatever else. Like, they're going out to buy HVAC companies. They're going out to buy transportation and logistics and a bunch of other things.

Like, those are private equity people didn't join, you know, McKinsey or Bain or whoever's part of it to, like, watch what's happening and and take stuff back. Like, those things are gonna end up buying companies. And I think the consumer version of that is you end up having a digital employee assistant, somebody that's giving you more time back to do whatever it is that you want to do and hopefully making a bunch of money to pay for itself and and earn you money. So I think that's probably

as granular as I can get in terms of what the vision ends up looking like down the line, but there are most of what we do will make sense if you ladder into that into that setup. Yeah. Well, it was interesting to hear you touch on the, like, people talk about spending a lot with this. They talk about making transactions. Oh, you can give them the virtual card to your agent, which is very cool. Don't let me skip over it. But there is a side of like, sounds like losing a lot of money. Like, sounds like spending more.

Like, if I'm the whether I'm the CFO or the individual consumer,

it's like, seems like more is gonna be going out of the account. It's interesting to touch on how this could be saving and earning as well. It's crazy how obvious things look when inverted or framed differently. Like, I remember the first time I almost worked at Ramp twice. I remember interviewing with Eric Blimen at the end, and I think it was the first time. And Eric said something. I was asking I was like, I'm a huge credit card points guy, this and that. He's like, no. No. You don't get it.

Like, we're trying to make it to where you don't spend money. Like, you don't not give you the 3%. Like, just don't spend any of you don't need to spend the dollar. And I don't think people realize how much more of that there is out there to be done. Like, they all waste money. Like, why? Why? It's inexcusable at this point. So, anyways, I think there's a bunch of people out there that are gonna work on this sort of stuff. What But do you think he meant by that?

He it's easier to your it's better for your business to save the dollar than to give you the 3% on the rewards. Like Yeah. If you're setting your Yeah. If if we're gonna give you 2% back instead of 1% back on your ramp spend card, like, you should probably have us take the 1% and go find ways to save you money. Like, let us negotiate the contract. Let us call Comcast. Let us get you the bulk pricing or whatever else, and that has worked quite well for them.

So, anyways, I I think there's a, yeah, a bunch of stuff in that direction that is not easy to grab and is a a different hill and a different direction, and others are gonna work on it. We hope others work on it, but there are so many different directions to go. That's not even the main course, if that makes sense. Like, that's that's just but one of things that will be true many years from now. Mhmm. Yeah. Number one rule of investing. Right? Don't lose money. Yeah. Exactly.

Well, to a limit, one of the my favorite sayings is, like, you can't save your way to being a billionaire. I was like, okay. That's also true. But if you say if you're frugal, you can definitely have a better chance of being a billionaire down the line. Sure. That's anybody's goal, but it might be your agent's goal. I suspect many people would be like, make no mistakes. Also, fastest path to a billion. Yeah.

Lewis, who are you mostly wanting to speak with right now in terms of the business and how you think about the future? What kind of companies, what kind of builders

you think you know, we've talked about the idea of focus and, like, thinking about prioritization.

How are you thinking about the kind of developers and use cases that you wanna really target right now? The vision is big, but where do you really see the moment for what something like ATXP can provide? Like, where is that currently right now? I think you're seeing a lot more people show up. A lot more merchants are selling things via x four zero two or MPP, and a lot of that I think more people are starting to realize the iPhone moment is coming where it will be profitable to be an iPhone app developer.

So, okay, can you make the thing that doesn't just summarize my meeting notes, but then fires off the emails afterwards? Can you not just give me the five leads? Can you start working on them? Can you craft the email? Can you find the gift to send them? Like, all of those sorts of things. There are little companies popping up that way, and I would encourage so in terms of what builders should be doing, many people take things, like, 99% of the way there, but don't do the last thing.

Like, I don't want my agent to give me, hey. Here's what you go do. I want the easy button to say, great. Approved. Like, go do the thing. And that's a much harder problem, I would say, if you get to summarize our entire company. It's like, we're now folk we did some of the 99 stuff, thought we were great because we were doing the 99, and now we're like, no. The one is actually what ends up mattering. Go do the thing for me, which is a much harder higher health decline, harder last mile, however you wanna say it.

In terms of who we wanna talk to, we our cap table has let us talk to a bunch of people, and that's why we have a fairly good understanding of where everybody is at in terms of bought in and experimenting and who's waiting to see versus other stuff. I think that there are a bunch of people that a bunch of companies that don't quite know that they're in a position to partner on a bunch of things, and I think we're gonna go try and convince a handful of others.

I I think companies like us are now getting a bit more proactive in saying, hey. We could work on this together. I think a handful shout out to Kai at Visa. He's been great at all of this sort of stuff. The Stripe team has obviously been been good. Folks like Cloudflare and Vercel have been good. I think Open Eye and Anthropic are trying to figure it out. They got other things going on. When they leave DC and wherever else they've been, there are some there are some big things that they can do.

It it's kinda like in my old role at Google, you can trade things at extremely unequal value. Like, it doesn't cost us a ton to give somebody else a massive leg up on what they're trying to do, and they can give us a massive leg up by something that doesn't cost them a lot. And there's a lot of that, like, trading things of unequal value that could happen right now. Bunch of people wanna be doing experimentation, wanna be doing distribution. Hey. Can you teach us a lot more?

That would be huge to companies in our space. And we're gonna go do the hard hard work of, like, bucketing and categorizing and showing whatever data and what's resonating. We're basically an outsourced, like, r and d shop for a lot of these entities. So we're having conversations like that. And, yeah, we have some things that are not too far away that we'll be launching that we think are gonna have, again, some of the to help us climb faster and help us run up the hill.

But we're still so so early, and I'm I'm happy that a bunch of the folks in the space are having, like, the right level of sobriety in terms of talking about this. I don't think people are out there just yoloing ridiculous headlines saying, you know, all of us are gonna have all of our finances run by agent tomorrow. Like, that that's not happening, but it once it starts to happen, it's pretty clear that all of the pieces are now pretty connected that it'll it'll happen quickly.

Yeah. Something That's wild. Makes me something makes me curious about, Lewis, is, like, you talk about people are talking about the singularity, or they talk about, like, human super intelligence and augmentation. And I'm thinking about, like, the nature of human intelligence and the nature of machine intelligence. Right? Like, some of this stuff is saying we took the same capabilities, like a really good human or chief of staff or house servant manager or whatever you wanna call it, like, could have.

It could think for itself. It can make these purchases. It can save you money, earn your money. It's it's feels like a repetition or an augmentation of human intelligence. Then I also get the sense there might be some ways where this kind of agentic intelligence or agentic operations execution is fundamentally different. Do you see it? Like Oh, yeah. Like, where's the Venn diagram?

Where's the, like, here's what is different about about how this is gonna operate from just a repeat of, like, now you have an AI servant versus, like Yeah.

Well, get buy back your time. I have to figure out the right way to directionally paint who this is without revealing who it is. We've had a bunch of conversations with peep we so we think agents are gonna make people much smarter, much more capable, do a lot of things for us that either we don't find interesting or repetitive or, like, we're just not as good at. Fun one. Like, nobody has ever read the or, like, the things you sign away when you've land on a site. It's, read the 40,000 words.

Like, I'm never gonna read that. My agent is gonna go read that and as is everybody else's, and, like, those policies are probably gonna get crazy over time. Fun throwaway example, but, like, probably a good business there. I was talking to somebody that manages, we'll just say, a lot of money. I don't know how many figures, but, like, not single digit billions, lots more. And we were talking about their process and how a bunch of funds go about things, and it was a similar conversation.

Like, we wanna make the individuals better, but we also want something that, like, allows us to have an edge as, like, the investment firm. And how do we, like, create the brain that does all of the thinking for us to where we have, like, a super analyst? And I'm like, I don't think the so if you took Fable or whatever, dropped in an agent form, set it in the middle of the table, I don't think it would think 95% of what you guys do is interesting or useful.

And, like, that might be offensive, but it's also gonna be wrong about let's say, you're remaining 5%, it's gonna be wrong about how to value that one to 4%. And there's gonna be some things in the 95% that it threw away that it should have found interesting. Like, oh, these two people talk talk a lot about this particular strategy or insight. Like, maybe I should figure out what they're saying. And to make that more concrete, like, do you plug it into emails? Do you plug it into Slack?

Like, Bridgewater and Ray Dalio went one very extreme of, like, everything's recorded. You have a baseball card, and it's, like, how believable you are and mapped to your record. Like, okay. Kinda wanna see how that plays out. And Bridgewater's not been crushing it, so that's that's weird. Maybe everybody else has caught up to them. But in this this future world, how do companies end up working? What's useful is Claude Tag, which just got dropped into your Slack. Is that a good thing? Is that a bad thing?

First of all, you're giving Anthropic all of the data that's not on the Internet now. Like, they are training on all of that stuff, which is, I mean, when you're in their position, you can do whatever you want. But I think companies if you ask people at a company what they think the most valuable thing they do is, I don't think that's what agents will find interesting. So even in answering your question, like, I'm not sure what is useful to an agent within a company.

Like, it needs to see everything, and then it's like, okay. These two execs, when they're talking, they are completely clueless about what's going on. They think they're talking about the real issue. We need to go down and get into the one to ones of, like, the country manager here and the country manager there. Like, they're exchanging the things that are actually useful.

And, I mean, these are all hypotheses, but we are like, when I tell that to people at very large well, in public companies, whatever else, when they ask us how we operate internally, I'm like, you should track everything you can. You've got some problems and, like, your lawyers are gonna handcuff you a bit, but do not be too prescriptive in what you're trying to to do at the start.

In much the same way, the initial self driving was all, like, line by line, you know, if this take this left turn, if that break, you have to let go of the wheel. Like, you have to let it train itself to get to the point where where it wants to go down the line, and that's a that's a scary thing to do if you're running a, you know, closing in on trillion dollar company. Like, what are you are you taking your hands off the wheel and be like, agent, you go find out how to train yourself.

Like, that's that's scary for a lot of entities, but I think the very small companies, the YC or HF Zero or whatever startups, like, people like us, that's how we start. So we don't have the the baggage to worry about. Yeah. You can really lead lead that way by doing that doing it that way. It's a good way to show, not tell. Yeah. I would be curious if if you said an agent and you said watch all of our stuff, listen to all of the text messages you guys are sending, the Slacks, the emails, and all of that.

Like, ask the agent, like, what is the most useful thing for you? Where are we wasting time? Where are we missing opportunities? Where can we be more informed? How can we be following up better? Like I was gonna for you guys. Like, what what are you currently for Circuit and Chisel? Like, you and DNR? Like, how are you what's what tools are you using? Is this actually a you know, you're kind of dogfooding stuff internally with even just the company?

And are you are you guys implementing stuff like this for yourselves right now? We

test basically everything

is the way so I have, I think, six or seven different $200 a month subscriptions on all sorts of stuff. So if you want the hot takes in terms of Is granola, like, running right now in the background No. Real time transcription? No. We do record a good amount of stuff and very intentionally not record some other stuff when it's, you know, real crazy talk. We have, I would say,

we are So I was ask you, just sorry to interrupt you. What that seems really interesting to hear. What should people not be recording in terms of biz like, in the enterprise and business context? Like, what falls in that category in your opinion?

If you're so let's here, agents have a very hard time letting go of the fact that that discussion happened. You kinda wanna introduce that dot in a very, I would say, careful way right now. I don't know that that's the case with Fable or whatever the other things are down the line. It gets too fixated on this, like, one off tangent and whatever else, and it starts trying to, like, pull you like, it's some gravitational pull into this crazy other land, and it's like, okay.

Well, we were just humoring what would happen in that world. Like, I don't I don't think people should have you should still take notes. You should still basically have a back and forth if you're in Codex or Cloud Code or any these other things. Like, have a different project, say, you know, based on what you know, this agent or this project, go look at that project, and now think about it. You're basically quarantining it off to the side and saying Yeah. Okay. Now let's, like, we role play this.

By the way, do not let that cross the border because we don't wanna entertain what were to happen if we I won't get into any crazy ideas. Anyways, the the broader point is I think there are things that you should be, like, manually taking notes on and then introducing slowly. Tools and other things, we were very early to the harness game. We were very early to Cloudbot, like, Hermes or Hermes, I guess I'm supposed to say. Played around with a bunch of those.

They're really interesting, but they fall over for, like, new some obvious and some not so obvious reasons. And I think we've taken a lot of those pieces. We've played with harnesses. We've looked at building browsers and all of that that sort of stuff, and we're like, we don't think that first of all, a lot of respect for other people building them. We hope that they succeed because we wanna use them. Second of all, that's not the hill that we're trying to climb.

Like, that is a feature on the way to whatever our end product is, or that is a a milestone on our way. So I don't wanna get into any of the specific tools. I would say that the the tempo folks or, I guess, Paradigm folks have been doing cool stuff with some of their, like, would you call it, Slack agent harness style stuff, which is very similar to CloudTag.

Those you should be looking for fringe products projects on GitHub or having your agent look for fringe projects on GitHub, going and splashing around, tinkering with them. We talk to a bunch of the open source people, collaborate with some of them, and it's very inspirational. We try and help them to the extent that we can give them resources or share notes, but there's so much to be built. And we want I've said this from the beginning. Like, we don't wanna be building all of it.

We wanna be, like, discarding and using the best of everything else out there, so we are spending a lot of our week testing stuff. I was talking to an investor yesterday

who I think he was actually talking about Stripe when he was saying this or alluding to it, but he was like, the biggest companies who've really hit escape velocity in our lifetime typically surround themselves with the idea of enabling entrepreneurship as, like, a core part of why they become successful. Do you tend to agree with that as you think about the future here? Yes. Although so, yes, I think that has been true historically.

I would slightly tweak entrepreneurship. I was having this discussion with our team the other day, and I was like, there are bunch of people that don't know that they are going to be entrepreneurs in the future. And that's You're just trying to see the Unlocking entrepreneurship? Yeah. I think Patrick Collison was tweeting about this recently where it's like the number of solopreneurs.

It's these people that were working on a thing, and then, I guess I have to turn this into a business or somebody else said, oh, turn this into a business. Like, I mean, was a farmer an entrepreneur once upon a time? Like, technically, yes. They were just trying to feed themselves.

Like, the the target market was their family and maybe people in their town, but there are a bunch of what I guess we call vibe coders today, but soon people that are just using their voice to go create a new product or company or whatever it might be. So, yes, to supporting entrepreneurs, but I would expand or widen the aperture to say, like, there are a lot of people that don't know that they're entrepreneur people that are 10 years old living in some faraway place.

Like, when they're 25, they are gonna end up being entrepreneurs. They're gonna have some service or products to offer the world. They, at that point, might not have even thought of themselves as entrepreneurs. Like, there's a huge step change coming in terms of the number of people that are able to build stuff that, like, have not been able to build stuff historically. So, yes, I should just give shorter answers. Yes. Z, I'll we got time for maybe one more question, then we can wrap. I'll let you take the last one.

Mhmm. Picks from the seven you wanted to go. I can give you the shorter answers. I've I've cut me off. I can give you the three word answers. We do a Hot or Not segment on some of the stuff we record or hyped or overhyped or whatever.

But I think one I love the thing that might be about an intersection of our worlds we'll see if this is worthy of the last question or not. But, you know, being in the media space, we think about distribution. Plenty of business leaders think about a lot about distribution. When anybody can make anything, as you said, it's like, well, who are the ones that win? The one that have distribution.

I guess something I've been thinking about in this conversation is how distribution looks very different, agentic distribution, than human distribution. Like, how many people watch this on X versus, like, the if you say, like, I want the alpha in the world of modern money. Maybe I should go watch Stable Dash. Maybe the agent says, like, maybe I should read the transcripts from every live show and podcast that's been put out in the last two months.

Like, the I guess my question is around, like, how do you think this this concept of distribution as it relates to business Well, this is insanity and alpha as it pops up. How does that look different in an agentic lens than sort of how we might think of it from our limited little human brains?

This will be a we'll do a crossover segment at some point. So we started a a show called Unartificial, and we haven't said or done too much about it. We've been we we did some test recordings, and now we're doing a a bit more of it. And it would be fun to collaborate because our whole point, to answer your question shorter in a shorter way, people are gonna wanna see things built. They're gonna wanna see the not just a demo of, like, this could work, but, like, I wanna go actually make a thing live. Okay. Great.

We're we're trying to show you, not tell you about AI. Unfortunately, the first bunch of recordings have been us just talking, and it's, like, takes on bigger companies and and all sorts of other stuff. But, okay, in this segment, been working on, can we do something interesting and useful in twenty minutes? Like, I know exactly what the thing is that I wanna go do in that twenty or thirty minutes when we chat at some point.

And for us, the thing that we're trying to distribute, it's, hey, this is how easy it is to go do, and here, go do it. I think there are some some products that are just words and thoughts, and those are easily distributed the way that you're describing. But for a bunch of people that have a a digital thing to sell, I think you're gonna have to start either building it live or demoing it live or breaking it down or customizing it or whatever it might be.

I I think the buy my PDF or subscribe and take my course for $900 is gonna be, like there's, unfortunately, be more of them short term, but long term, I think you're gonna have a hard time tricking people into that distribution because you're gonna have to have gone and done this sort of thing in the short term. Like, prove it, I guess, is the point. So that's how I see distribution evolving, and I would I would argue the that's actually very wholesome.

Like, you need to deliver good things because, like, my agent will call out that thing's a waste of time. Like, they don't say anything useful. You can't get anything out of that. So the bar like, the quality bar should go up over time, and that's it's scary because we're gonna go invest some time doing all of this.

And, like, it needs to be good in order for us to get distribution, but it ends up it ends up, hopefully, whether it's us raising the bar or others doing it, like, there will be better content out there to consume with that side of my brain. I still love golf videos. I still love, like, face plants and, like, all sorts of other stuff on Entertainment is different than

like, yeah. Entertainment versus how would you categorize the alternative there? Like, efficiency? I am a self improvement

hobbyist. So, like, I spend a lot of my time on self improvement and probably an unhealthy amount, but I I think that that's how I would categorize it. It's like, okay. What can I go learn? Sometimes it's too much. Like, just like an exercise, you kinda need to take a rest in order to improve or get stronger, whatever you wanna aim for. I think listening to an audiobook while you're tired about to fall asleep, probably not the ideal thing to do, so I've stopped doing as much of that.

But, yeah, I think self improvement or continuing education or whatever might be on one side versus, like, no. I I don't think adults laugh enough. Like, there should be more comedy. I want AI comedy nonstop. Like, I wanna be able to tap into just some hilarious stuff that is trained on all the things that were bookmarked or liked or whatever on Instagram. Just, like, keep feeding me that. I'm I'm not gonna run out of laughs in my lifetime. And anyways, I I hope somebody does it.

That's a trillion dollar company, by the way. The Spotify for laughs or whatever. I would I would pay for that. Joy is good marketing. You just I mean, kids are happier for a bunch of reasons, but like, we I look at my kids, and I'm like, I I don't know if I could figure out how to make adults happy in a legal way. Like, just for fifteen minutes a day, I think there will be, like, a significantly better place. Mhmm.

And maybe I'm doing this entire company so that we can create the fifteen minutes to then I'm gonna come back and sell you a comedy product online. But, yeah, that's I think there's a whole bit hopefully, that's, you know, Paul Graham and whoever's Gary fund that in YC. Like, hey. Let's put out an RFP for request for startups for, like, AI comedy. Yeah. So funny. Even like roast in the comedy seller.

Just give give access to Slack, and I I do think that an agent could roast the team pretty well and get to know the team. Just like having it even for, like, employees, like Fill up to a meeting once a week. Pull up to a meeting, which you might not have heard of before. There's at least when I was there, whatever, seven years ago at this point, I'm very sure it still exists.

There's a thing called MemeGen, and it was a super easy web app that you could just go to, make a meme in real time, and it was all inside jokes. So sometimes inside jokes for, like, the map team or the YouTube team, but, like, I I remember. Put Eric Schmidt on blast. This isn't gonna be great. Well, whatever. If he watches this, like, we're all happily ever after. We're all we're all winning. Eric Schmidt, I think he was CEO. Maybe he was chairman of the board at this point.

Google does this thing at the end of the week where it's TGI Friday, which they moved to Thursday, but it was like, they go through the the issues this week. Any employee can ask a question. Eric Schmidt used to do this. This is what I found. I don't know if this is right. This is it? This is it. So, yeah. So first example, Marissa Mayer, former Googler, then ran Yahoo. No one does not turn a profit, accomplished tech leader, finally leaving a nonprofit. Like, hilarious roasts across the board. Yeah.

So these are all Marissa Mayer focus. On the bottom right. Sergei is saying to Larry, I'm missing her already. Can you buy Yahoo for me? So, anyways, most of them are inside jokes, but the point for Eric Schmidt was, like, he said something along the lines of, like, them or something, and he was like, Nest, wonderful product. Like, if you don't have one, should go buy one. I have a Nest thermostat in all of my homes. And the the meme afterwards was like, Google are trying to afford San Francisco house.

Like, can't make it happen. Eric Schmidt, Nest in all of his houses sort of thing. So, anyways, to your to your point, many people have tried to build Meme Gen after the fact. It's an amazing thing. And if you had an agent that's just sitting there roasting your meetings or the the takes or whatever else, I think, first of all, you turn it off because you would get owned. But, yeah, that's the kind of thing that maybe I'll go Vibe Code later on because I think it's good for team morale and stuff.

But, anyway, especially Yeah. We all need to get roasted every once in a while. It's good it's good. I think I think there was a study on, like, the closest friends and, like, how do you measure intimacy? I think this might have been just between men, like, male relationships

is, like, how much essentially shit do you get from your friends is, like, gonna be how you measure depth of the relationship. Like, if you're not getting roasted by your friends, you're probably not that close with them. It's kind of the study, the takeaway. I would I would agree. Well, I don't think most PR teams would be okay.