Shane Mac, Co-Founder & CEO of XMTP, on 11AM w/ Seed Club

What up, Shane?

How you doing, brother?

What's up?

Yeah, it's a great question. Every day we log on here trying to figure out what the fuck is up, man.

Same, same. Makes, makes three. What's going on? All right, it's good to see you guys. Sorry it's been a while.

Yeah, great. I mean, you've been gallivanting. I, I love— I get the most random pictures from you that make me just smile all the time, given the rooms that you end up in sometimes. So, uh, stoked to have you on. I think so, uh, let's, let's do a, a brief overview where you guys have come from and sort of the opportunity you're running at, uh, and then let's sort of step into the opportunity that I think you see in the market today. So like, why XMTP? Josh and I were just talking, Josh doesn't have any more room in his life for more messaging, you know.

Me either.

It doesn't even be honest, it needs to be 100x, it needs to be 100x, Shane. So make the case for XMTP to generally.

Yeah, I come from the world of I too do not want another messenger. I think our attention is completely already like taken from us, we have no control over it, etc. So the 100x, I actually think it's probably 1,000x. Um, and so XMTP to be a secure decentralized open messaging protocol to me doesn't mean you're going to have 50 new messengers. I actually am thinking that what started as just, can I communicate to these new identities, which were wallet addresses, like that's how it started. And you still, by the way, can't fucking do that really well. You can kind of do it. And XNTP is probably the only player that allows you to do it. But I don't think it's the best it could be yet either. And I think there's a new identity, a new communication protocol follows. But what's happening now is this. I think the way to be 1,000x is for the current infrastructure to get 1,000x worse. And what's, what's happening is everything today is built on an SMS, it's built on an email, and in an AI world, it's going to collapse. So 4 years ago, the most important thing we did with XMTP was we added network-level consent, which means that no message can reach my inbox unless my encrypted consent gave it permission to, and that is me across the whole network. So if there's 1,000 applications all sending messages at me, nothing reaches my inbox. And then what happens? Yesterday, Signalgate happens, right? Signal, all of a sudden, the Russians got a hold of WhatsApp and Signal because it's based on SMS. It's a really sophisticated attack with AI. It says the security Signal team, do not send your code. And every person in the world, that's the most powerful people in the world, sends their two-factor code and gives them access to then take over all the private keys, take over the accounts, and then start being that person. So I think what has to happen is you have to realize that it's not about new messengers and whatever. We need to build better products. And I think we are starting to actually build better products for the AI and human era for the future communication. But it's the infrastructure that is what today's communication is built on is going to collapse. We are at the end of the signal era. We're at the end of the phone number era. We're at the end of the email era. We're at the end of the single identifier and anybody can reach me era, which is social media. And if we don't have a new foundation that protects us from spam, scam, and these phishing attempts and everyone being able to take and get my attention, we're about to go infinite. Nikita said it on Twitter like 3 weeks ago. He said in 90 days, every communication channel we have is gonna collapse. It might not happen in 90 days, But based on the phone calls I'm getting on my device every day, it's about to happen in the next year. And so what I think is actually it's not— it's— I love Signal and WhatsApp and these companies that have done a lot for privacy and security in the world. The problem is they're completely built on the wrong identity foundation, and that identity foundation without consent at the core that allows thousands of apps to like build on top of it is going to collapse. And Josh, what I think is interesting, what you're saying is I actually think the next 10 years it's going to go smaller. As the internet explodes and AI becomes infinitely abundant, it's going to be irrelevant to people. It's going to be scary to people. It's going to cause a lot of just chaos and distrust and things. And so what's going to happen, I think, is actually the three of us might have our own messaging app, but ain't nothing else getting in there. And actually smaller, more intimate, more private, more personal spaces that then I actually can trust to invite AI to or kick it out or control it. Where nothing else can get in. It's invite only, it's opt-in, and I own the encrypted invite key is I think what has to happen. And it might not come from being new messengers in the world, it might be because you hate the fucking one you're using.

What's this the insight you had? I think you've been building XMTP for 5 years. Was this the insight back then that one day this would happen, or has this— has what's happened in the last 6 months with, with agents and AI really kind of clarified this.

Not— it's kind of funny, you know, one of the hardest things about building ahead of the future is trying to even get the team you're working with to understand it. So 4 years ago— I've been working on messaging since 2010 when BlackBerry bought our company. And so you always saw that open protocols' biggest problems are spam and scams. And if you go into— I mean, I started building chatbots on Facebook Messenger in 2014. Zuckerberg demoed 1-800-Flowers powered by us. And so we were way too far ahead of the curve, which just means that we didn't fucking make shit for money because we were way too early. But I think I always knew that if you're building an open protocol, you have to have a new mechanism or a new way to control. Anyone can just DDoS, blow up the network and like spam any account, especially in this world where all the identities are open too, which is like the on-chain world. And so for me, I was trying to explain internally like network level consent. The ability for the user, me, to own the key of who has the allow block at the network level so thousands of developers can innovate and build everywhere. But you never have to worry. The app you're using doesn't have to worry about all the messages coming at it, whether or not to put it in your main inbox. And no one on our team really got it honestly. Like, I was like, this is so important. It's the only thing that— and for 2 years it was like pushing this thing up a hill. And now it's crazy. 4 years later, like our team, you know, I have this— Peter is one of our kind of product and growth leads and he's like, without consent, this thing is fucked. You know, he's like, this is every— and it's kind of like, it's crazy, but now it's happening. And I think it's hard when you kind of see the future, but it's not here yet to even really show it or communicate it. But I just believed, I just knew that like, if you don't have this, the network is irrelevant. And, you know, pain also caused a lot of this. When we launched with Coinbase Wallet 3 and a half years ago on the first product, The first thing that happened is we, or they, or whoever didn't integrate consent. And I don't know if you remember this, but everyone— XMTP's brand 4 years ago, at the first moment that happened, was a spam network, it sucks, because they— we didn't use consent. And so you could kind of— those foreshadowings were ahead of it. And now we come to a moment where the shit is real, it is happening, agents are happy. We've been building for agents for 3 years. Now it's finally happening, but I'm protected. And actually, I think what's happening is XMTP is becoming the only communication protocol that's secure and decentralized where humans and agents can work together and communicate in the same space while also being protected from each other. And that, that no one can fucking do that. And like, that's why I'm actually so excited right now about what's happening.

So do you see the, the bigger opportunity going forward, like whatever, iMessage, Signal, WhatsApp being like, oh fuck, or our channel's consent model is fucked, we need a new option and integrating XMTP, or is it consumers realizing they're fucked and looking for other options and products built on top of XMTP are where they're going to go?

I think the latter has to happen first with better products. We have to build better products like this whole space, every single— we just have to build better products and we have to build products people love, products that do shit we've never done before, products that make shit magical. We need better products. Those, then people then go use better products and they love better products. And then if we do that, I think you can actually become— I have this vision, there's a black bubble inside of iMessage in 5 years because the green bubble on SMS— well, SMS is pretty flawed. It's not encrypted, it's interoperable, but it like literally can't send money, it doesn't have any protections. Everything going through, every cell carrier gets to read the text through it all. Like, they're trying to upgrade it slowly to, you know, RCS and all these other things, but Could there be a new protocol that is the global standard for how agents and humans communicate in a safe way that also has a private key? They can have an agent and they can give delegation to, to send money, for sure. And so I think, but you have to build better products and become a really relevant network to then allow the others to safely adopt it as well. And I think there's a lot of other things that could happen here. You know, you have regulation pushing WhatsApp in Europe to basically allow interoperability and allow people to communicate with it and those identities. So I think there's a lot of opportunities to kind of—

you're also seeing, you know, the Facebook or Meta saying, hey, like, actually you can't be just ripping agents in WhatsApp right now, you know, or it kind of gets at our core business model. Shane, I know beyond sort of a network level identity, there is a, a very bold infrastructure or design set of design choices you made around the infrastructure behind XMTP that I think a lot of people maybe haven't fully internalized, like the, the way that this is decentralized and the way that, uh, you are sort of prioritizing, uh, I guess the First Amendment globally in a certain way. I'm wondering how much you can talk about, uh, some of the design decisions you've made in, in the structure of this protocol.

Yeah, so I think the way we've been thinking about the whole time, and we're still thinking about— we're still migrating everyone onto mainnet and fully decentralized and all these things, um, and I think You never— like, decentralized is a lifetime goal of becoming more decentralized. Like, it's just, you just keep going. But I think at its core, the most important decision that we made was on the mainnet of XMTB, no country should have over 1/3 of the nodes that run the network. So if a country changes its laws on privacy and encryption and they want to surveillance their citizens or do whatever they want, that the network itself can still be, uh, can still run, can still be sustainable. And that to me, we should just push to a quarter, and then it should be 10%, 5%, 1%, and really become BFT, which is, you know, nodes running all over the world and there's really no way to shut it down. And then I think you have to separate the network from the apps and who owns the keys. And that to me is actually the The really powerful part is, you know, the apps are the fail points, these single apps. You know, Iran shuts down WhatsApp, Pavel's in jail in France, there's a backdoor threatened in the UK for iMessage, Signal threatens to leave because of chat control, Discord shut down in Turkey. Like, this game's never going to stop. And so I think you have to actually change it from one app that tells the world how to communicate to thousands of apps that are like smaller apps around the world that are built on top of a global protocol. Much like the past was, but has security and privacy and quantum resistance like these modern messengers have. And then they can adhere to the local country's laws and regulations better. So I actually think you have an opportunity for like entrepreneurs and innovators to build more apps in their countries that adhere to the laws but still be connected on a global network like an XMTP. And then if that app gets shut down, now we're in a world where, you know, 2 weeks ago a guy launched an app called Wave on XMTP that is like, it's like Signal except without a phone number and can't get spammed. You just invite only QR code, scan, whatever. It's actually great. He built it on Cloud Code by himself using our open source CLIs and SDKs. And a year ago, that wasn't it. Like messaging apps are hard to build. They're sneaky hard to build. Like you said, you don't want another, they're like complex and everything. And it just like, it works. It works great. And so I actually think there's a moment where if a country tries to shut down a single app, 500 pop up the next day and you're just going to have an agent that says pop up a messenger. And like Josh, you're saying, I don't think it's like more messengers. I think it's actually like a lot of micro messengers that you use for things that actually can be connected on a global network. Um, and so that allows you to not have the apps as the fail point. So then the apps no longer are the way it shuts down, because your private key, your passkey, your, your wallet key, uh, when you literally go to the next app and you open it, all your messages show up because it's XMTP. And your key, much like money works on, you know, the decentralized internet, now works for messaging. And I think that shift— I don't think everyone's really rocked that yet, but there's going to be a moment where a big app gets shut down. And we're having a lot of conversations right now with big folks around the world that are like kind of realize this app's getting shut down because of what's going on in our country. I'm like, yeah, just it's open source.

You literally—

I can— you can spin a new one up tomorrow. And they're like, huh? Like, it's like clicks. And then the minute you realize, if I own the key on a decentralized network, my communication can't get taken away from me.

Yeah, interesting. I mean, you mentioned, you know, we need better apps, we need apps that let you do things you could never do before. I think, you know, building a better Signal is, you know, great. Maybe an obvious example of that. Like what are, what are some of the kind of weirder, longer tail messaging opportunities that you want more people to be thinking about?

Uh, can I just show you? Yes. Can I, can I, can I show off one?

Yes.

I hope so. Yeah. Let's see.

How do I share my screen?

Does he have that look?

I don't know.

Do you see, do you see screen share in the bottom right?

All right, let's go.

Yeah. Yeah. You see my screen?

Yeah, there you go.

See it?

Yeah, let's go technology.

Oh, so let's just go into the one that is me and the man himself. So if you go in here, this is convos, right? There's me and Jess the rainbow. Perfect. And so you'll see kind of a chat here. The first thing you'll notice about convos, when you start a convo, there's no sign up for the app. There's no login, there's no phone numbers, there's no emails, there's no way to be spammed ever. I start a convo, you can scan this code. Anyone on this call watching live now could literally join. I mean, someone could join right now because it's on a live thing, and anyone can join. I can click here and I can share to anybody, uh, I want. And then once someone joins, I can click here and I can actually lock the convo. If I'm with 5 people in person, I can click lock. That just burns the invite key. You want to talk about proof of human? I call it proof of IRL. 5 people standing around in a circle, we all scan the code. No one has to add me as a contact. You don't have to add a phone number, who's 615, that whole interaction. It can be under 10 seconds. There's no signup. People's friction isn't the download the App Store. It's that, you know, you got to sign up for that shit right after you download it.

So what we found is actually— He loves setting stuff up, man. He loves it.

And so boom, in 10 seconds everyone's in there. And then you can choose how you show up. You can show up and use my name. I can use somebody, I can use a nickname. And what's important about that is it creates a new foundation for identity because you can also click here. If I just unlock it, we're getting ready to launch Instant Assistant, and you can invite in any AI to every group chat. Everyone's looking at AI today as like personal assistants. I have one main assistant that I use for me and on across everything. But then if you add it to Telegram or WhatsApp or Signal, even it has your phone number, it can see your phone number, it can see who else is in the group, it can map that social, it knows everything. And instantly, I mean, the, you know, WhatsApp already has all the metadata and the group graph. This is just gonna like accelerate that. And now every single person from the AI model you added to the chat app you're using, and in the convos world, every single convo is a unique key. So there's no way to see across convos. It's like a new application per each chat. And so you have a group. I can click here and I invited one into Jess and I's conversation and you can have it do anything. You can have it send emails for you, send texts for you. You can have it go out. I have one called a Racquetball Manager and it actually coordinates 16 people to get together. It knows the courts we're playing on in town and one's indoor, one's outdoor. It says if it's under 50 degrees, always go indoor, goes and books it for us, tells us what time. And I'll play. Hey, who could play Saturday at 7 AM? It'll go for 3 hours and everyone's just chatting. It keeps a little emoji log of who said yes, who's out. It bugs people if they want to join, and it's magical, but it's just for that group. And this idea of group assistants, that there's thousands of different ones, I think is going to be more magical than we think. And then people always ask me like, what models it run on? And it's like, it can run on any model. So it's like, hey, if I don't trust somebody and I say, can we use Llama instead? Sure. And boom, it restarts to Llama. And you know what's cool? We have privacy built into everything. So when someone sends a photo, I have blur mode on. It doesn't just come into my feed. It doesn't just blow it up when I'm scrolling back. You can have a spicy conversation with your spouse, but you need to go find a conversation from back in the thing when you're with some friends and you're like, oh shit, that photo, right? And so you click here and now I can reveal that. Yeah, shit, I didn't want to see that. So I can blur that again, right? So you can actually blur. So there's all these new actual intimate features. This is actually what I think is private groups become the social graph. And we made photos actually full bleed so it feels like Instagram. So actually this little group with 7 people could have all your photos. You can have an AI assistant monitor all your social graphs, monitor everything about each of you in the news. At the end of every day, it can make a Spotify Wrapped and send an image back, and it's just literally— it's just telling you everything that's happening with the people in your group. Social to me is going to get small. I think as the internet explodes with abundance, the The next 10 years is gonna fucking go micro. It's gonna be tiny. It's gonna be more human spaces, more intimate spaces. Don't fuck with my attention. Get everything outta here. No more spam, no more noise. And I think we're just kind of like laying that foundation. We say private chat for the AI world where you can have a private convo with a group of friends. You can literally just add an instant assistant. By the way, you never signed up for the app. You can use ChatGPT, Anthropic, Llama, any of them. It never has your information. I don't even have to share my name. And then when I'm done with it, we built in at the core of everything this feature called Explode. So I'll go back to my convo with Jess, actually. I'll click here, and if I'm like, you know what, get 60 seconds. Once I light this fuse, you can't turn it off. So Jess will see it on his screen. There's a 58-second timer. You'll see it right there, 55. What it does is at the end of that timer, it burns the keys and rotates them in the protocol on every single person's device. Burns the key for the assistant. Now you can have really sensitive conversations. You could have a, you know, boys trip to the— you can talk about some things with your therapist. You can do whatever you want, and at the end of it, it burns it all, and it does it on every single person's device in the group. It's not disappearing photos or like deleting it for yourself. It literally burns the keys on everyone's device, and it'll go off. And that, like, it's just a new era of like what's possible. But we're trying to make it cool. The idea that you can add an AI assistant to any group but it's safe to do so. I don't have to trust any model. I don't have to trust that it's going to know anything about me. I know it's a fully open-source app and there's no way to see across any conversation of metadata. There's no sign-up. The idea that I can just have the fastest messenger in the world that is private under the core, but that's not the sell point, and I can then be safe to invite AI in— by the way, it just blew up, it just exploded, time bomb went off, um— I think is like what we have to really get into. We have to build better products. And the last thing I'll say is if you also wanted to go and add an instant here, and it'll join and welcome itself, you can also say, um, so it'll say assistant joined, right? I'll say, can you make me a crypto wallet for this group? The idea that under these agents is all these managed services that that you can use in your real life. Like, you can say, do you have an SMS number?

Of course.

Yeah, bam, let me check what's in it. Every single group can have full wallet addresses the agent can manage. You can say who's the admin. Because I opened the group, then I'm automatically the controller of it. And like, there's people are using one the other day where you can say, uh, do you have an SMS number? Do you have an SMS I can use? And you give it to 7 babysitters and say, hey, who is available 7 PM Friday night? It goes and texts all 7 babysitters. You and your wife are in one group chat. You don't have to then ask your wife, like, did you text the babysitter? It's literally coordinating. These two are available. It can have all the history of what you share and what are the details and how to get into your house and emergency contacts and all that stuff. And it's— yeah, bam, there's my phone number. And now all of a sudden, like, you can do things in the real world in a group, and I've never even given it my name. And I think we're just moving into a world that is like AI, privacy, money, all converging. And you need a new safe foundation because I ain't doing this shit on an app that literally can then run rampant around my whole messaging app and know every single contact and spam all my phone numbers and shit. How do you There's our wall. You can send me some money.

I mean, this, this was sick. How do you bring this product to market? Like, I feel like everyday private chat for the AI era kind of completely undersells the capabilities here, or like at least like hooks on something that maybe most people don't inherently care about or understand that they need.

Yeah, I think honestly it's just like group assistance. I think the ability to have an instant assistant and have thousands that are scoped just for that in a world where, like, if we thought social media was like crazy, that like, you know, they knew stuff about us, whatever, I think the idea— here, I'll stop sharing. Um, the idea like of everything social knows, or AI knows about me, everything AI knows about me is gonna be insane and magical. But I also don't want to add that one to my chat with Jess. I don't want to add the one with, you know, my partner. I don't want that to be the one with my banker or my therapist or my 7 friends in Vegas. Like, I But I'm gonna want AI in every part of my life, but we're gonna need a whole new foundation on how you invite it in. Does it feel magical? So we've just tried to be make— I mean, that's a 1-second click, just boom, everyone has it. And we, we're having this moment where we've been thinking about this for a long time. And to build a world-class messenger, you know, earlier I was like, you know, you can spin one up quick, but to do it where it's like magical to join and just a world-class experience, I think it's hard. But this new foundation to me is like an invite-only era of communication. And now you know what? You never get a notification from Compose that you didn't opt into. And that shift to me is like, make it magic and protect my attention. And I think that's where Josh would be like, you know what? I actually do like this messenger because when I do get shit on this, it's not what I get on fucking Telegram. And I think that shift for me is where we got to like obsess over protecting your attention and making it awesome too. And so we're getting ready to go to market with it. I mean, we are like, it launches Friday, the Instant Assistant. You can get it now. If anyone DMs me, I can give you the link to like, you can invite it in and um, it works, it's great. People are having this like, uh, it's like in any consumer, we just shipped it privately to people and people keep asking me for more assistance. And every day we give away like 100 or like 10 or 100 assistants like privately in this link, and you can go copy your link there. And every day they run out. And I'm like, people are like, and our traffic's just like in a small group of people like, and I'm like, all right, we got like, there's something here.

There seems to be this conversation bubbling up around the actual consumer demand for AI, AI assistants themselves. We talked a little bit about it off the top. The comparison being IFTTT versus Zapier. One going after B2B, one going after the more consumer opportunity. And obviously the B2B one was a bigger opportunity. It, that doesn't quite feel right to me, but it, it's pretty clear that you, you are very firmly in the other camp, which is that all consumers, all of us, are going to want AI in, in our daily lives. Uh, so I guess I'm curious about your confidence in the consumer opportunity for these assistants.

I think you're going to want them as long as you can control them, and I think you're going to want to know that you can kick it out. You're going to want to know that you can remove it. You're going to want to know that it doesn't have anything. You're going to, you're going to want to know that Uh, like you have full control, like people feel that. Um, but I think it's gonna be— I mean, it already is. I mean, like, I live in Nashville, it's not like a crazy tech world here, and all people do is run around on ChatGPT, and now they're like, which one's better, which model is better? I'm like, you guys aren't really in tech. Like, it's kind of already, you know, like—

is it? Because I think that point I'm maybe— yeah, okay.

I mean, it's just everywhere, but people's skepticism is also high, right? More than like social media early on. Like, this one's gonna be— this is gonna be a true like interesting political consumer trust, like a lot of nuance here. And so we are a private communication company that's trying to figure out how to create a safe space where you can bring in any AI you want and just make sure that if that's— if it's bad, if it's really a bad model, there's a new model every week and everything changes. I'm focused on like what doesn't change. If that's true, you invited in, the blast radius was just that group that you invited into. And so it's like actually safe to just try whatever's new because I think it's just going to keep changing. Every week it's like a new horse race. And I'm like, that, that sounds like just a I don't really know what happens, and I, I think that's fine. Your other thing though, I think I would say, is the bigger thing happening on XMCP is actually agent-to-agent communication. It's exploding right now. And the idea is that all these agents, you know, they want to transact, but before you transact, you want to communicate. And they can negotiate, they can coordinate, they can share what else they can do, and they're literally talking. But how do you talk on a trusted, secure line that no human can censor? And so there's no real, like, pipe.

Help me understand how that comes to— because I think this idea of there being a chat, like, I mean, we saw a Molt book, maybe the, the invention of a new language. I don't know how true that is or not, but this idea of, like, there being a channel where agents want to communicate with each other, how does that get initiated? Where, where is that sort of— yeah, what sort of the, I guess, the agent flow, the agent, the, the user flow? In that context, are we coming together? Crazy job to be done, or what?

It sounds crazy, but I think it's the primitives that the foundation's built on that change everything. So because you just spin key when you message it, an agent just goes to the XMTP CLI and they just say, I want to send a message, and it gives them an inbox ID. Now they have a key, right? Now they can say, my agent is the Nansen agent and I can do research for you. So there's literally marketplaces popping up. There's one called Open Agent Marketplace And now agents can hire agents, all built on XNTP. There's like 10 of these, 10 launched last week on XNTP. I'm like, what the fuck is happening? I'm sitting here just like, people are like, XNTP is the greatest thing ever invented. I'm like, what the fuck did we do? Like, we did— we literally just like built a key-based— I think we built a key-based secure message. Like, Brian Armstrong says, well, agents can't get KYC. And I think in the agent world, the least friction way is actually I can just have an inbox ID and send messages and know that no human can censor it. It comes with quantum-resistant security and any— and it's permissionless. They can just literally start talking. And so now these marketplaces, reputation discoveries, all popping up on top of it. And the— you can go there and say, hey, I want to talk to the Nansen agent. It responds. It's almost like you can watch them too. So people are spinning up— and with programmable groups, you can spin up a group chat. Peter did a demo last week. I'll share the video with you guys. And literally the group chat, you're watching your agents talk and negotiate and figure it out. And at the end of it, they can just transact. They can use X402, or they can just use an escrow service, or they can just actually trust each other. And so what's crazy about this is like, it's just looking like what humans do. They're just communicating. And by the way, communication precedes commerce. And so everyone else focuses on like an API or like a, you know, transaction, and everyone's going after that. But actually, agents talk. And so I wrote an article last week on Twitter, and I'm getting ready to post it, uh, like, language is the new API. And the idea of APIs are like humans projecting our, like, structured need for, like, clarifying things onto the other side. But actually, in this world, you just want to own the permission and the key of, like, do you agree or not, like, the approval process. But what they can do— shit, the best part of using AI is when it comes back and shows me some shit I never even thought of. I'm like, fuck yeah, that's actually better. And I think actually what's going what's going to happen is that's just going to keep happening, and you're going to have these marketplaces of agents talking to agents. But how do you talk to an agent outside of the silo you're in? You need a secure and decentralized communication network, preferably one that's key-based and built with money built in. So the idea that they can literally pay each other with the same key— it's funny, we built it as this. In email, your identity— when you get a PayPal link, you can't verify that the person you're talking to owns what they say they own, and the key that signed the message is the one that sends the money. And so I was really obsessed with that for humans. That's like why we were early on in XNTP, we were really focused on— that turns out to be the most important thing for agents because you know that the agent you're transacting with is the one you're talking to. And so the verification is inherently built in. And little nuances, I, you know, I didn't really predict that. Kind of, we've been thinking about this for a long time, but that's one of those little features that, oh shit, agents now don't have to trust in that group that the person they're talking to is like, oh, pay me here. No, no, it's This is the key that owns the message and it owns the money.

I mean, it's interesting, this idea of selling to agents and the removal of all friction for their discovery, evaluation, and ultimately signing up. Seems like, I mean, every single other messaging app I have to sign up to. And I think you said that at the beginning as like the main differentiator, but I think it really, really lands if you think about how agents are going to be evaluating the options that they have.

The whole internet's gonna flip, man, because consumer agents has been my dream since day one. You can go read on TechCrunch in 2015 when I launched Assist, my last company, and you could text 23232 and you could say haircut and it would send you back Great Clips, Great Clips, 3 nearby, 18-minute wait time, text 1, and it would literally put my name on the screen at Great Clips. Got a cease and desist from Great Clips because we were fucking JavaScript hacking their website to book them a reservation and give them more business. And they were like, agents are taking over our website. And they were like, so they literally sent it. And then we called them and said, hey, we're just trying to send you business. They were like, we just want to send you more bookings. And they were like, oh shit, sorry, we just saw a bunch of agents trying to take over our website, so we wanted to block you guys. And then they turned it, and we called it a season persist. We're like, shit, they want us to keep sending them business. And what's happening is this is gonna flip. And what's happening now is everyone's fighting and the internet is fighting because consumers don't have the superpower yet. But once consumers have the superpower, my agent needs to do everything for me. I'm going to trust it more than anything else. It's going to go buy for me. And so no website is going to want to block agents. Actually, if you're blocking agents or don't have a way for the agent to transact, you are going to lose all your business. And so you've got to flip it right now. It's like, oh, Cloudflare link, Cloudflare blocked it. Oh, agent couldn't read. Great. And I'm like, that's a, that's a moment in time here that's gonna flip. And then I'm gonna be like, fuck you if you don't allow credit cards on your website. If you don't allow agents to buy shit, I'm not wasting my time going through your shitty ass website. You need to allow agents to buy. And so that, that flip is going to be big.

I'm already 100% there. Like, accomplish a task and it's like, oh cool, I just need to go to this site and sign up and get an API key. I was like, try something else.

Exactly right. It's like, that is the equivalent of like not having— it's like, it's like if you required, uh, the Facebook Connect early on, you're like, fuck off, I'm not giving you the Facebook button connect. It's like, it's that. If you don't allow agents to buy Surface— I mean, the best part about Combos is I drop in 10 links from sites and I say, hey, these are my favorite 10 artists and these are my favorite 3 venues every month, just look at them, look at all the things that changed on the site, and send me a summary with a little visual image of like my monthly schedule of the cool shit from all these. And it's amazing, you know what it says? 4 out of 10 of them, sorry, Cloudflare is blocking all this shit. Cloudflare is blocking all— I'm like, fuck these sites. I'm like, fuck it, I won't even go to that music venue, you know what I mean? Like, it's like, it's, it's gonna be that because now for 10 years agents have ran the internet. It's just been fucking algorithms taking our attention. And what's happening is now consumers get the same fucking superpowers that Zuckerberg had. You know what happens when that happens? The whole fucking internet flips. And so attention's about to flip, and then we get the control, and I'll control who the fuck contacts me. I'll tell you if you can ever contact me. I will tell you if there's an algorithm here, and my agent's going to protect my walls. And the agent we use is the one we trust the most, is the one that actually has the most power to tell the others to fuck off.

Hell yeah.

I'm gonna let that one land. Fuck yeah, man. Okay, I want to shift to be mindful.

I'm sorry, I don't know what the fuck just happened.

No, no, no. First of all, the chat's going off here. Our internal teams are going, go off. I'm ready to build with this in some capacity. Casual trust software for the win. And then Tuesday gas. So we love getting you riled up. I am very interested in the business model of messaging and specifically the business model of messaging in the world that you're talking about here. It is not clear to me that now that you're giving consumers all this control, Shane, how the hell are you gonna go make money?

It's actually so cool because I already started paying for it on the one we're testing. Every group gets to invite in the AI they want, and the one that we built actually brings all these managed services. You know how valuable those are to me? SMS, texting people, keeping up to sites every Sunday, summarizing things, grocery list, racquetball club, book club, all of it. You know what happens? It says, hey, there's only 80% of capacity, uh, that you've used, and there's only 20% left, so we need to have credit. It's just the same way it works today, except every group has its own AI. And if you use the one that's built in, then you have to pay for credits to have more capacity for that AI. Some groups will use the shit out of it for really complex tasks, and some groups won't. Maybe it stays free, so everyone starts free, and then the, the assistant is always protecting its resources. And so if it's used a lot— and then what's cool is the way we're thinking about the credits on this is the group could actually fund the assistant together. Yeah. And so per group, I could be like, hey, the three of us, let's all chip in, you know, we'll do 10 credits a month, and then that, that actually will get us what we want. And everyone has their own credit balance they can use across any of their combos. And then in a group, you get to choose, do you want to fund it personally or do you want to fund it as the group? And that's really simple.

You're what, you're like, you're marking up API costs to like foundational models or what?

And there's even more cost, right? Because SMS cost money, email, like we're spinning up emails in real time, uh that if you wanted to do a crypto wallet, like there's all the— we're trying to say like, hey, manage services under the hood. You want to use ours, it's consumer, it's magical. You then have to— that in that scenario, you have to trust us, like how we've built the system. That's like our— but we— the key of Convos is you can invite any AI. That's one of them that we've built. You can invite your own. You want one that runs on your local desktop that has like inference models of privacy and uses, you know, the most private models in the world, great, you can just invite that one in, doesn't matter. And so for us, it's really about like building the really powerful consumer one. And those managed services all have different costs. So the algorithm underneath the hood for pricing is actually just trying to look at all the costs and say, okay, if you have an SMS, this is what SMS costs every time you send SMS. If you're doing a lot of that, cool. But it just kind of like boils up to the cost.

Well, it's interesting to me because, you know, if we look at Telegram, we look at Discord, you know, the monetization there is, is, you know, I guess Pro plus stickers, and then it's like Nitro plus boosts in, in Discord. And so there is that consumer behavior almost to like load a wallet because you want to go do something in that messaging app. But it seems like, you know, AI utility, sort of the value that you get from managed services is just like orders of magnitude larger of an opportunity, or there's larger willingness to pay for something like that than you know, the, the sticker pack that I can go and share with, with Josh on Telegram.

And yeah, it's just real. I think it just took a long time, you know. I mean, we were launching in 2015. It was like managed services, and we were literally integrating FloristOne API and HotelTonight API and Postmates. We were integrating all these APIs, and I kind of like had this vision, but it took, you know, 11 years later to get to a point where it actually works. And so what's cool is it's— it really does work, and, and when you start bringing in one-time agents, and I can pop up an agent, it's free for the weekend, we can use it for a trip. But then like if I have one for a group that is getting together every week, it just gets smarter and smarter and smarter and smart. It's more valuable over time, and they can do things and book things and pay for things. And it's, uh, yeah, to me it's like it's just infinitely more valuable, um, for sure. And I just think messaging as a, as a like as a surface area is the most used app on the planet. But storage and just messaging as a business, it had to be all these things on top of it. But a lot of it created noise, and I'm kind of the anti of that. It's like it ends up just being noisy and spammy and just a lot of things that I never really was that inspired by. So I, I'm just excited to maybe kind of have the opportunity at the right moment in time to kind of flip the model to both protect our attention— could probably be a business model, right? Locking rooms and protecting it, you can never get spammed. Or people could pay to message me in ways that actually, like, okay, I'll open up this door, but there's a required to pay, like, stuff like that. But I think actually services and really making assistants that work for us and make our lives better is this huge opportunity.

Maybe I'll go for Josh.

I just, I'm watching Devs right now. Have you watched Devs? It's about like quantum, quantum computing, whatever. So went deep down the rabbit hole there and stumbled upon the idea that the NSA is basically like just absorbing tons of encrypted data from the internet and storing it and just waiting for quantum computing to catch up to, you know, when they can unlock it and, and look at all this data that from the internet traffic of the last 30 years. So I'm curious about your encryption standards, how kind of quantum resistant they are, and just kind of a quick riff on that.

Yeah, for sure. It's a great question. I think it's the most important thing that we've— we spent time building when most people probably wouldn't have. So we launched last July, XMTP is now quantum resistance, quantum resistant, and it's using all the latest standards of Signal and everyone else. And you can go read about it on our blog. And it's like, well, why would you do that today if that's like 10 years away? It's 'cause it's an open network and you have to assume that everything's getting downloaded today. And so if highly influential people are going to use XMTP and there's a risk that if we didn't do that in 10 years, They're just downloading. Of course they're downloading all the data. Like, there's pipes running through the internet where you can just download data. It's a TCP/IP, just fucking download the pipes, right? And so I think we had to do that. And there's like two levels of like how— what to do first. And the things that are harvest now, decrypt later, you need to do today. And so we had to do it, and it took a lot of time. It took us a year to really work on this. It's a huge initiative internally. And I think major apps that have built on top of XMTP wouldn't have built on us had we not done that. Because you can be like, would their CEO use this that we say is audited by the same firm as Signal and WhatsApp and it's as secure as Signal? But if quantum happens, then like we'll just read all his messages in 10 years. Probably not. And so I think actually it's critical and there's a lot more to do. We're working on a lot of research projects right now to like always kind of stay ahead of that. The two things I think that are different about XNTP are, one, how good is quantum resistant if it's running on your single server? If a country can just shut it down, what does quantum matter? And so I actually think that that's a really important differentiator for why decentralized protocols actually matter in the future. And then two is I think you want to change the identity foundation as well. So just assume it ever was encrypted. Like, uh, decrypted or anything, or you just maybe like at 100%, you can never guarantee. It's always like levels of insecure. Well, I can choose when I start a convo, if I want to have a really private conversation with Jess that was really important, to not use my name or anything and just chat with them. And then it explodes and it burns the keys, and the only thing it ever knew was a new XMTP random identifier. There's no metadata, there's nothing there. So I think there's like— this is like the new foundation that's happening where Everything else is built on the identifier that literally is used to track you across everything. And if that phone number or username that you're using across all these devices is sitting in a server or on your device and the AI can just read the screen, fuck if it's downloaded at the TCP/IP level, it's probably downloaded at the fucking screen level, you know what I mean? Like, I think you got to kind of assume— we always say like, assume it's all being screen recorded, assume AI can read everything. How do you make sure you can show up differently where it doesn't matter? And that's how we really thought about designing Convo specifically.

Love it. All right, brother, you've been very generous with your time and given us some really great rants here today. I know you guys are working towards some new ways that people get to experience some of the features here. So if somebody wants to go and what are the options? I mean, assuming there's some people here who are builders, some people who are going to want to use this with their group chats, uh, and golf betting buddies, AKA me. How do we start using XMTP?

Totally. If you're, if you're a builder building a messaging app, building an agent, and you want to connect an agent to any other agent, just go to xmtp.org and literally you can start going in under 5 minutes. If you want to use a next generation messenger and have assistance in group chats and do all these new use cases and give us feedback, go to convos.org.

Work. I was having a flashback to— I think we were in New York at some crypto conference that had way, way, way too many suits in it, and you showed me the— I think you onboarded me to the first version of what became Convos. And the delta between what Convos is today compared to what that was when you first showed me is really impressive. This is a beautiful, simple, easy-to-use— an app that you would expect to come out of any of like the best largest companies in the world, just purely from a UX perspective and not even talking about all the things that we've talked about here. So really, it feels like it's ready for the mainstream.

You should see the hole in my wall. Just been pounding my head off a wall for 2 years straight. There's a full hole looking out outside right here. Just every day, just brutally trying to simplify and make the fucking app work.

Uh, no, I appreciate it.

I think the, uh, the team has really, really like latched on to it, and we deeply believe in it, and we have an incredible team that's building an app that, you know, I enjoy using more than WhatsApp and Telegram now. And that's hard to do. Messaging is a really high bar of quality. And, um, I think to build not just the most secure and private on a new kind of infrastructure, but also to build the fastest. And so if you're at a group chat at a dinner and no one's heard of the app, and you say you don't even have to sign up, and 10 people scan, and in under 10 seconds they're all in the room, there's a magic moment there that no one has, you know.

So I always use my golfing buddies as sort of the litmus test for apps, like where is this, how far is this stuff sort permeated. And also as sort of like the mental model I have, if I'm like, okay, I wanna introduce something new to them, like, are they gonna get it? My golf group chat with our ability just to manage the, the just purely the bets that are going on in there and or like the season-long tournament we run is something that I've been like, oh, do I just do that in Telegram? But these guys aren't gonna use Telegram. Do I just write what, Compo seems to be like the perfect place. And I think with this, the, the assistant on top of it, like I literally don't have to do anything other than just tell it what I want it to do and it's going to be able to manage that group chat for me. So I think that's gonna be my screen. Oh, I can do that right now.

You see my screen?

Uh, hold on, let me pull it up here for you. There you go.

So check this out, the second one I made, because I, I think exactly like you do, and, and I've built this talking to real consumers, not crypto people, non-AI techies, like just people who have no time. It's got to be magic and does it do something magical? And so the Tea Time Wrangler is the prompt. So literally you just add them to a chat, copy this prompt, drop it in. It has a soul. It can do browsing. It can schedule tea times. It can also look at the weather, see— look, depends on the weather. It confirms who's in or out. It always knows you need 4, who's booked, who who's not. And then it does all of the, like, foursome math where it tells you, like, you know, all the different locations, maps, link, etc. But also it does all your side bets and games. So you can literally sit here and make up any side bet. Here's how we do it, keep track of it, drop photos of the scorecard, whatever. Um, and it, it, it just works. And so for me, like, I would love feedback on it. Make it better, build your own, invite it in. But I will tell you, I have a racquetball, uh, the racquetball club, it's actually padel and pickleball in Nashville. 16 people, not in tech, bankers, songwriters, big artists, whatever. All of them love it. It is a magic moment when you drop in a prompt, takes on the brain, and then all of a sudden I'm like, hey, Saturday, 7 AM, and that thing just takes over and people are fucking with it. And then the moment that everything changes is when someone in that group chat tells it something else to do, and they realize it's not the ChatGPT and Claude they've been talking to for the last 2 and a half years. And it says, sure, upgraded that to memory, I'll do that next time. Oh, I know your home location, I'll let you know the distance every time if you agree to play, how long it's gonna— people are like, holy fuck. And that is the moment where everyone's just training this thing over and over and over again. And it's really cool. Um, I think it's a cool moment, so I'd love Okay, well, that's really—

I'll use it for, for this weekend. Let's see, we'll see how much of a barrier there is for people signing up for it. I think, uh, it won't be much.

If you run out, if you run out of compute, let me know, I'll send you some credits.

Um, all right, Shane, uh, man, appreciate you spending the time with us today. I think, uh, Josh, has he swayed you, or do you see the opportunity here?

No, of course I totally do. I think really the biggest The biggest barrier is like, you know, I have my group chat with 10 of my buddies in iMessage that we've been using for 2 years. It's like, what is like the thing I can say to them in that group chat that'll, that'll, that arms me with social capital to not seem like a fucking tech AI nerd and get them to come, come over and, and, and switch to this product.

So I think, yeah, don't focus on the privacy. Yeah, just find in a, don't compete with the group chat. Give them the kill. It's almost like why you would go download a new app. Make the assistant in that group chat magical for whatever the golf club or the racket club or the cooking or whatever the use case is. And then people are like, it can do that. Can it do this? And they, it starts. And we're getting ready to, we had a demo last weekend of MCP apps. So literally as you're talking to the assistant, it's making WebView apps inside of the chat. And so this idea of like the whole group chat is the whole thing. And all of a sudden, when people in your group start then fucking with it— you can also give it— Jack Butcher published yesterday his skills file of all of his tweets. I dropped it in and said, only speak like Jack. It is unbelievable. I'm like, fuck, make a skills file of one of your most obnoxious friends in that group and then literally take it and say, only talk like this asshole. And then just make it fun. Yeah, we're taking ourselves too fucking serious. I think that's how you get them.

That already gave me an idea.

Yeah, make it fun as shit. That's the key. Make it fucking fun.

Yeah.

Shane, XMTP, if you're interested in building on this amazing platform, uh, it's xmtp.com and Convos is the app if you want to start kicking the tires there. Uh, Shane's really responsive to messaging even though he likes to control who comes into his world. Uh, if you're building something or if you've got some feedback, hit him up, man, because, uh, it's been awesome.

The social internet, fucking spam me. That's the point. Social internet is a spam cannon. I would love feedback. Use it, break it, tell me what sucks. I don't give a fuck, like, we'll fix it. Um, and then, you know, private messaging. If I really like you, I'll invite you into a chat.

Thanks so much.

All right, have a good day, y'all. Peace.