The Internet Needs an Upgrade

Zac Smith00:00

For the past 20, 30 years, like point something, 0.1% of society has been able to write software and ship it to production. I mean, maybe it's, maybe it's, maybe I'm off, but like fractions of society. And it feels like we're going from like 0.1% of society to like a couple percentage points of society.

Jake Moshenko00:19

Yeah, which is massive, right? It's huge. Millions and millions of people, hundreds of millions of people. Welcome, everybody, to the Base Layer podcast. I'm Jake, one of the co-founders and the CEO at AuthZed. Today's guest is Zac Smith, one of the people who has been building the real internet underneath the internet for more than two decades. Zach is the co-founder and CEO of Datum, an open network cloud for the AI era, built around a big idea. As intelligence moves everywhere, the internet itself has to become more programmable, more distributed, and much smarter at the edge. Before Datum, Zach co-founded Packet, the bare metal cloud company that helped developers automate physical infrastructure at global scale. Packet was acquired by Equinix for $335 million. How's that sound? Where Zach went on to lead bare metal, edge infrastructure, and software connectivity efforts inside one of the most important digital infrastructure companies in the world. He started his career in the early Linux cloud days at Voxel, has helped shape the evolution of cloud networking and edge computing, and now sits at the center of one of the biggest questions in technology. What has to change when AI agents, data, users, and policy all need to meet in real time? So today, we're going deep on intelligence at the edge, why the internet needs a major upgrade for AI, and what it means for authorization, trust, and the next generation of digital infrastructure. And on a personal note, Zach and I have known each other and have been working together since our CoreOS days. And so Zach has been one of those voices that I'm just always turning to, always listening to, always working with, to try to figure out what's next for myself, our businesses, infrastructure. So this is a really special one for me. So Zach, thank you so much for joining today. I'm sure this is going to be an absolutely awesome conversation. So welcome to the show.

Zac Smith02:05

Oh, thank you, man. And that's a really, you did a good intro, man. I think I learned a couple things about myself just listening, but it is a pleasure.

Jake Moshenko02:13

I hope you're hyped. Yeah.

Zac Smith02:14

And, you know, it's just great to have known you and the team through two companies so far. And, you know, great part of the scene here in New York and just this is a great topic. So very excited to be here. Yeah. Awesome.

Jake Moshenko02:29

So let's kick it off. What is the coolest thing that you've done with AI?

Zac Smith02:34

Well, I do a lot of not cool things with AI.

Jake Moshenko02:38

Well, we don't want to hear about those. I'm just kidding.

Zac Smith02:41

My mission in life is to not be staring at my phone. And so I need to talk at things. And, you know, I guess I've totally turned into my dad, who was, you know, big on like next tell walkie talkies. He was in the construction industry. So everybody just like did that. I've been trying to get back to it ever since because, boy, that was a better way to communicate than having to stare at your phone and type all the time. But in all seriousness, beyond enjoying some improved voice transcription capabilities is some people may or may not know, but I'm involved in a soccer club out of Denmark. And it's been a little passion project of me.

Jake Moshenko03:21

It's like a part owner kind of thing.

Zac Smith03:23

Yeah, you could call it that. We're part owner. This is a club. It's called AB. It stands for the academics. It's been around since 1889 and about three years ago, maybe four years ago, roughly 200 people from primarily here in New York crowdfunded and bought this team out of basically destitute, you know, bankruptcy place. And we've been kind of slowly turning it around. And my job is like all around tech guy. You know, I'm like the guy who fixes the Wi-Fi. And we just promoted to the first division, which is the second tier there in data soccer. And that means we get on live broadcast and all kinds of great stuff. And anyway, so I have been building our tech stack. And it's so cool for me because I'm a very kind of, I would say, dangerous scripter. Like I can copy other people's code pretty good and I'm not bad with bash commands and I can do a little shell scripting here and there, but I was never really a software developer. And I always felt really limited by that. And what's been super fun for me is I've been able to, you know, take some real capabilities that I wanted to see happen. And this became my outlet to do it. So it built this media asset manager. So one of the things in a football club is you have like a lot of media. People are taking pictures. You have like, you know, video streams that are coming. You have, you know, all this stuff. And then you need to get it to your players and your sponsors. Oh, this is a great shot of this player trying to go with this cool sponsor logo in the background because like, you know, that's what they pay for. And, um, and so I built an open source asset manager, uh, for taking in all these media files, um, basically auto identifying facial recognition, but dealing with like, well, what if it's from the back? Well, you got to get their name and things like that. And then finding the sponsors and then kind of identifying the great shots, um, and then distributing those to our team members and players. So it's been so fun for me to be able to be like, you know what, I want it to do this. And, uh, I've learned a ton, uh, myself in terms of just pipelining, how to, how to kind of, uh, you know, I wanted to do the whole thing full end and to end myself instead of like, you know, calling in, you know, tapping in, uh, calling in the big guys and it's been great. Um, and I've definitely been humbled a couple of times and, uh, it's, it's been seriously fun. Um, and I have this thesis, you know, overall, and I'm sure we'll talk about this is that, you know, for the past 20, 30 years, like point something, 0.1% of society has been able to write software and ship it to production. I mean, maybe it's, maybe it's, maybe I'm off, but like it, it fractions of society and it feels like we're going from like 0.1% of society to like a couple percentage points. of society who can build, is huge. Um, and so millions of people, hundreds of millions of people, hundreds of millions of

Jake Moshenko06:29

people, hundreds of millions.

Zac Smith06:30

It's just amazing. So for me, that's been awesome. So that's the, an open source, uh, uh, sports asset. I call it Sam's sports asset media system.

Jake Moshenko06:40

Now you say it's open source. Is this something people can go and just get on GitHub or where do you get on GitHub?

Zac Smith06:46

Um, the, our, our repo doesn't have a lot of followers, uh, for AB. Copenhagen, uh, is, is the handle on, on GitHub. And yeah, you can find our, uh, our open source website and our open source Sam's. That's cool.

Jake Moshenko07:00

I like the idea of like an open source soccer team too. Is that like, uh, is that a thing we're doing? Can we do that? Yeah. Yeah. Just all the decisions just right there.

Zac Smith07:08

And, um, should I tell you my Easter egg? Uh, okay.

Jake Moshenko07:12

Well, uh, are you sure you don't want people to discover it?

Zac Smith07:15

I mean, maybe. Okay. I'm going to give you a hint. All right. Okay. Um, our founding year was 1889. So just, you know, go look. Yeah.

Jake Moshenko07:29

My control F skills are pretty strong. Yeah. Um, no, I've asked quite a few people, um, senior leaders, especially the same question. And one of the things that always pops out is basically agency, right? Giving people agency to go and solve their own problems. Um, and I know the word agent is crazy overused and is used for like a separate thing, but like fundamentally what it all boils down to is this concept of agency, right? Like being able to do things for yourself. Um, another thing that's come up and this is a little different cause you're, you're, you intend this to be a wider audience, but it's this concept of like personalized software, like software that's just for me. Have you done anything or seen anything in that space? Yeah.

Zac Smith08:12

I've been helping my son with something. Um, and, uh, basically we have this problem. I mean, I have the fortunate, uh, situation where, you know, I've become an addict and investing in private companies. And so I have dozens of small investments in companies, which are great. Um, but not all of them are sitting in a nice little clean website like Carta. Um, and so they're all, it's like a big portion of what I have to do. And, um, I've been thinking more and more that if something were to happen to me, like my wife would have no idea. She doesn't know about any of this stuff, nor do I want her to have to deal with it. Um, and so my son is kind of interested in, uh, you know, finance stuff and he's been helping me build, um, some personal software that controls your own data instead of putting it out there and something else. So we can kill some spreadsheets and files around a lot of what I do in my day in, in things that I don't enjoy. I thought about, well, what do I hate doing? And I'm like, you know what I hate doing is taxes. I hate doing taxes. I hate like looking for that document and I get like K ones and weird forms. And I got to remember about this one. And I forgot about that one. And it just bothered me to have to pay like lawyers and account. Sorry to all those lawyers and accountants out there. I love y'all, but I just can't stand like paying people a thousand dollars an hour to like go look for a form. And so, yeah, we've been writing some personal software to do this and it's been great. It's awesome. Yeah. I'm just trying to take everything and like, Hey, I hate doing that. I don't want to do that anymore. I don't want to remember to do that. And it's been really fun. It's not totally there yet, but, um, that's what I've been working on.

Jake Moshenko09:54

I mean, I'm right there with you. I run my family's whole financial life out of a spreadsheet and my tax thing, like that, that same thing that you're saying, my tax thing is that I have a to-do list that every year when it's tax time, I just mark everything off. And then next year I just uncheck it all. Like I cross it all and start at the top. Uh, so yeah, I, I totally hear you. That's awesome. Um, cool. Well, one thing that I find really interesting is as a CEO, you're also a CEO, um, educating customers and investors and coworkers is a big part of what we do, right? Constantly having to tell people our mission, our vision, reiterate it, uh, et cetera. What is something you're constantly having to teach people that you wish kind of everybody already knew so that you could start the conversation at like step two?

Zac Smith10:40

Um, well, the one that, that comes to mind and I, I'm awful at like, people are like, give you an elevator pitch. I'm like, Oh, I don't, this better be a long elevator ride. Cause I love to give context and, um, where we work, um, where I've been spending most of my professional career is in internet infrastructure. And I basically have to start every conversation with a new employee or a partner or a customer with like a step-by-step day in the life of how a data center actually works. Um, you know, um, because people will say, Oh, I need a data center. And what they really need is they like, I want AWS or they'll be like, we're building our own hardware. And they're like, really? And then they're like, Oh, I'm just renting this from that guy. Who's built it from that guy. Who's doing it for that way. And, and so like, just like walking through the layer cake of buildings, physical infrastructure, connectivity, you know, hardware operations is like something I have to like preamble a lot. Um, and I wish that there was, you know, a better way that I could just, you know, plug into the matrix, the Simston that goes in the back. And now I can like, we have a shared common understanding of what digital infrastructure is actually composed of.

Jake Moshenko12:00

So when someone comes to you and they're like, I'm going to build a data center and you're like, you really don't want to think you mean that. And you would know, right? Like when you were doing packet, you were racking and stacking servers. That was, that was what you guys want to build the data centers, you know, it was just like

Zac Smith12:16

putting things in the data center that it was already built and full of creature comforts. It was like taking a, a not furnished apartment and putting my furniture in it, you know, and, uh, versus renting a hotel room.

Jake Moshenko12:28

And, uh, so they just dropped off networking, power, cooling, and the racks. And all I have to do is put my hardware in like, yeah, that's awesome.

Zac Smith12:37

I mean, there's all that other stuff. And then obviously spending time at Equinix where we basically build data centers all day long. Um, and the connectivity for, to give that to people as a product, right. Um, you know, gave me a different perspective, but, uh, certainly that is an area where I wish, and it's actually become harder because although more people, um, in the kind of lingua franca today, understand data centers and power. It's like, you know, it's become the most hated thing in like, you know. Yeah.

Jake Moshenko13:05

I was just going to say, this is kind of like a hot button issue right now.

Zac Smith13:07

It totally is.

Jake Moshenko13:08

New York just passed, passed a law, I think, to ban data centers for a while.

Zac Smith13:12

New York state, yeah, passed a moratorium.

Jake Moshenko13:13

Um, uh, I think Michigan where I'm from, it's like a really big political hot button issue. It's like, where can we put data centers and are they going to scare away all the birds? Like what's the, you know, what is it like to live with the data center is a big, big topic right now. Do you have any thoughts on that before we move on? What's that? Do you have any thoughts on like, whether people should be afraid of these things or should be welcoming to them or.

Zac Smith13:36

I have a lot of thoughts on this.

Jake Moshenko13:38

Um, yeah, I think. Yeah.

Zac Smith13:40

I mean, basically, you know, this is the, the, the, the, the, the, the five G cell towers are going to, you know, are here for the aliens are going to give me brain cancer. Um, you know, but times a hundred, I think. Uh, and, and so I think people have very real and logical concerns about the large industrial scale infrastructure that, you know, is being built in some cases. And I think that we as an industry in the data center industry have done a really, really poor job. I don't think, uh, at communicating and in, in engaging in society. I mean, the typical data center, uh, tries to put up no signage. They basically want to hide behind, you know, security fences and walls, and they don't want anybody to know they're there. That would be the ideal. They just want to take resources like energy and do their business with like one car coming in and out per shift, you know? And I think that's not good enough. I think you have to engage and know your community. I think you have, you're part of it, right? Like, this is not in an Island. And so I think that's something that our industry hasn't done very well. Most data center companies don't do it. They're not like sponsoring the local T-ball team and telling people what they're doing for the community. And, you know, also, you know, yeah, you're not even allowed to go in, you know, when I was at Equinix, I got the first ones arranged where we could bring people under the age of 18 because you weren't allowed to go to one if you didn't have identification, um, you know, like that kind of thing. And so you couldn't even see one. Um, and so I think there's just a lot of work we have to do around transparency and being part of a community. And then second of all, there's a lot of people right now who are building irresponsibly and I think they should be, you know, um, community guidelines and there should be rules and requirements and there should be, you know, more engagement. You're, you're building real big stuff that use real resources. My personal viewpoint is that this is an incredible opportunity done correctly that we can invest significant amount of money in our energy, um, infrastructure, which we need. Uh, we need more, you know, resilient grid. We need better use of our resources. We need to kind of scale technologies like battery technologies and other kinds of forms of clean power generation and whatnot. We need to, you know, think about heat capture capabilities. Like there's all sorts of stuff we can do. Um, and a lot of that is happening, um, which is great. Uh, but I think in order to do that, we have to put in some kind of whether they're societal kind of norms, um, meaning like, yeah, if you try and build a data center and take all the things and just not ask anybody, it's going to get stopped and you should probably not do it that way. Uh, you know, and I think that's the, you know, to, to give the worst example would be, you know, Grok's data center, just basically like flouting all rules and, you know, putting out high emissions by burning, you know, tons of natural gas. Come on generation, right? Yeah. And that's an unfortunate way to do it. Is it, can you deal with the fines and the fines are super low and the output is super good for them? Okay. Yeah. But should you do it? I don't know, man. I think the answer to that is no. Yeah.

Jake Moshenko16:58

I mean, fines are a tricky one because they can be changed so easily. Yeah. Right. Oh, I'll put it into my business plan right now and just pay the fines and it's fine. Like, okay, what if the fines were 10 X or a hundred X tomorrow? Or what if each time you got a fine, it doubled. Right.

Zac Smith17:13

Like doing that every day. Yeah. Yeah.

Jake Moshenko17:16

So that's, uh, I mean, that's wild, but I mean, you touched on two sort of like power generation and data centers are both like very much. Yes. People want them as, as an idea, but they're very much nimby topics, right? Like I want them, but I want them somewhere else.

Zac Smith17:31

Like I don't know. I want the things they, they, they make possible. I don't actually want that, but I want the things that they enable. Right. Exactly.

Jake Moshenko17:39

So yeah, uh, super interesting take and, uh, totally awesome to have your expertise chiming in on it. Um, but let's talk about, uh, datum a little bit now. Um, so you say the internet needs a major upgrade. Yeah. What are you talking about? What's broken?

Zac Smith17:53

Um, well, a couple of things. So first and foremost, um, I'm gonna give you my perspective. I got into the internet in 2001. Um, it was a really interesting time. I just had finished up at music school. I didn't know what to do. I needed a job. Um, I ended up hosting a websites for some musician friends and needing to boot Linux servers. And in order to do that, I had to find a place with racks. And in order to do that, I had to get on the internet, you know, all the kind of like, you know, just one layer down every step. Bootstrapping. Yeah.

Jake Moshenko18:25

One layer at a time. One layer at a time. And, um, and so, and at that time, um, networks were, uh, how do I say? Very expensive.

Zac Smith18:34

So using the internet was expensive. Um, and what I mean is by pushing bits to reach anywhere in the world, um, had a high price point, you know, $2,000 a megabit, um, and, you know, $1,000 a megabit. And this is like real money, which obviously was constraining the amount of content that people would be able to, um, share. And a lot of us kind of worked in a community fashion to build internet exchanges. Um, some of them happened right here where I'm in lower Manhattan, you know, we're working on the New York internet exchange, which was just, you know, a bunch of hosting companies and some local ISPs and some other, you know, corporates and whatnot coming together and saying, Hey, what if we just connected our routers on the same switch? We would say, if you send me traffic, it's free. As long as you take my traffic, that's free. And then we will do this settlement free peering thing. And, you know,

Jake Moshenko19:30

These weren't the first peering agreements though, right?

Zac Smith19:32

No, there had been peering agreements, but you know, with the deregulation in the late nineties, um, and this just huge movement of data, um, from the internet, there wasn't very common. It wasn't very widespread. And you had tier one, you know, carriers, which were really, really monopolist. Um, and they didn't want to give away access to their networks, right? They're like, no, you have to pay to have access to my network. Um, cause what I sell is full IP transit. And so, you know, we, we just needed to work on how to make content more exchangeable. And, and so there were these, and many of the peering exchanges today are still nonprofits, um, like D kicks out of, out of, um, Europe, um, and open IX and whatnot. And then there's the private exchanges like Equinix's exchange, et cetera. But this has done a really good job. The point I'm getting at is that the internet was built, you know, primarily through community work between network operators and other users of content and consumers of content, um, to create a highly connected world. And I'm going to say, wow, that's gone pretty good, but things are changing right now. And I would say that this is, um, kind of one of the reasons why I think the internet needs an upgrade. Um, first and foremost, the internet's like really big. So you're in lots of places. It's not really easy. Like what, what we did back in the early two thousands by like showing up at a small data center and connecting a switch with the other guy would not be possible today. Meaning we'd have to go into 20 or 50 places around the world to start doing that just to start to make a dent in the, you know, global traffic patterns, but also we're going through a period of, I call it the great, the splinter net, you know, we're having a decoupling of societies, you know, in ways, you know, from a geopolitical standpoint, almost aligned with the whole kind of globalization was the major wave for a long time. And now we have more protectionist, you know, you know, trends going on right now, let's put it that way. And technology is at the center of that. Okay. And so, you know, whether it's, you know, chips and you can't have these ones or, you know, you, you know, these models or that models, you know, it's like, you would say this fight of open source and open technology is like raging right now. And I think that's, you know, networks are going to be at the heart of that. And so I think we're going to have more rules and more things to comply with around who I can exchange traffic with, where and why. Basic ones like, hey, you have to keep this data about French citizens in France. And you gotta use it only with French AI things or cloud things or whatever. That sounds like it should be simple, but I think you and I both know that that's incredibly complicated, especially with the way software is being built today, where you assemble a whole bunch of APIs and use this and that. And who knows if this guy is using this other thing and it's serverless and it when moved and live migrated and everything like that. So I think like, we're going to have to curate more who and how we connect with. So that's like point number one. And the internet today is just too complicated to, you would, you could try and add some of this policy at a software layer. And I'm sure you could tell me a lot more about that than, than I could tell you, but I think it's really hard to do that everywhere. And I think people are going to need to curate some of their connectivity in those regards, but the tools for doing that are really, really limited unless you own your own network. The second thing I think is happening is that data is moving into so many places. And the way I like to think about it is for the past 25 years I've been in this industry, we've had like smart servers and dumb clients, right? That's been the internet, I guess. Like we have a data center and it's got an application running on it. And we have like a dumb client called a web browser, you know, or something. But like, it's not doing much. Most of the data is over there in U.S. East or some data center in the corporate office or wherever.

Jake Moshenko23:31

Literally U.S. East. Yeah, literally U.S. East.

Zac Smith23:34

But now I think we have smart clients. And it's not just that AI agents are now doing stuff or people's laptops are doing a whole lot more. I, you know, sandboxes are running here or there, but we have local first movement. We have, you know, assembled applications. We have really, really intelligent, like JavaScript front ends that do a whole bunch of stuff. And like just the way that software is built is way more distribution friendly in terms of highly distributed. And I think that's putting data and the intelligence that goes with it in lots of places. So what used to be a pretty centralized thing, I think is getting moved really widely. So that's my second reason why I think the internet is a little broken. The traffic patterns are changing a lot. And the third is that I think we have so many new cloud providers that people are building and utilizing other infrastructures. They're using infrastructure services from dozens of providers. And that has created a fragility. And if you want to control security footprint, if you want to control who can access what content in terms of your network, your perimeter is wide. And I would say that the internet has to move into all zero trust. And so I believe that we need to have more of a private internet capability so that your application or your users or your things and your data can stay on your network no matter where you ship it. And so that's why I think the internet needs an upgrade. Because in order to do that today, you'd have to be a telco guru and you'd have to set up physical routers and start running MPLS and making a transport network and control your things. And I think that would be, sorry, very difficult. We don't have enough humans who know those things. And I don't even know the time. And I think the operating pattern just doesn't work. So that's why I think we've got to upgrade the internet then.

Jake Moshenko25:34

So when you say your network, you're not actually talking about my own physical fiber or copper. You're probably talking about an overlay of some sort? Exactly.

Zac Smith25:43

I think there will be interconnects to the physical networks of companies and clouds and data centers and whatnot, ISPs and related. But I really believe that we need to introduce the idea of private backbones overlay built. We've invested in SRV6. It's a segment routed protocol on top of IPv6, which basically allows you to add policy into the header of the IP packet. And this gives you a really, really elegant route anywhere overlay where you can start to enforce policy, which is really what you need is like, hey, send this packet here versus the internet just decides where to send the packet. And so you giving a little bit of rules on your traffic and your packets before the internet takes over and hot potato routes it and sends it to some router in China and sniffs all your packets. Like, I don't know. Like, that would be maybe not what you want to have happen. Yeah. And so I want to introduce the idea and hopefully popularize the idea that every application can have its own virtual backbone and that maybe developers can build virtual backbones for whatever they need. And then they can tie other services into it, like maybe their own services, but also third-party services. And this is really where a backbone becomes relevant with peering or interconnect, where you can say, hey, I have this thing called Superbase and I have this thing called Vercel and I have this thing called Databricks and I have this own thing that I run. How can they all be on the same network that isn't the big public internet? How can I have a private internet just for my thing and choose who gets to connect in and how that happens? So, yeah, that's what we're working on.

Jake Moshenko27:23

So, I guess in the way you see it, like, who are you pitching these ideas to, right? Like, your average end consumer has no idea. They think that the internet is just magic and they put bits in on one end and bits show up at the other end. So, are you talking about application creators, content providers, clouds? Who needs to know this idea? Who needs to be involved in this? Yeah, well, so definitely have to educate developers and application creators or the systems and tools they're using on how they can embed or add or leverage this kind of capability.

Zac Smith28:00

So, part of that's just similar to what, you know, we did at Packet, which was saying, hey, you can take this physical server and give it an API and then it doesn't suck. You can actually do things programmatically to it, which was before, like, not really a thing because it was like, you do that with a virtual machine. They couldn't do that with a physical machine because they were so wonky and required a sysadmin and then they'd boot half the time and not the rest of the time. And you just didn't trust that that was an actual experience. I want to do that with networks. So, that's part of what we have to do is enable developers and creators to know that this is a possibility and that this is going to work in the ways and tools that they're used to. So, that's the first part. But our target customer is a cloud. And so, I believe that, you know, our mission would be to help 1,000 or 2,000 clouds, like, win, survive, and thrive. I think more diversity in cloud providers and who's running your data and where that happens is a good thing, not a bad thing. And as such, I want to help those to embed this kind of capability in their applications or in their platforms. And so, those are our target customers. We call them magnetic providers. Of course, they will have end-user enterprise customers, you know, other people who will leverage this stuff and use it. But our goal isn't to sell directly to an end-user enterprise. It's to help providers like AuthZed or, you know, MongoDB or whatnot have superpower, backbone, and networking capabilities embedded in their product experiences. Okay. Yeah, interesting.

Jake Moshenko29:37

Where do you think all these clouds are coming from? Is this a durable phenomenon? Are they just waiting to get gobbled up by the hyperscalers? Is this NVIDIA apportioning hardware? Like, why are there 2,000 clouds?

Zac Smith29:51

Why are there so many clouds? Well, I think one is just an evolution. And we probably need to redo the previously aforementioned IaaS, SaaS, and PaaS terminology because it's all mumbo mixed up now, right? I use SaaS apps as if they're infrastructure, you know? I mean, what's Resend? I guess it's a SaaS company, but it's, like, email infrastructure. I don't know. Like, if it goes down, kittens cry. Like, I don't want it to go down. My app will stop working, you know? But is it IaaS? Like, I get no. So I think, like, we basically have just a lot of infrastructure aligned or, you know, adjacent cloud services these days. So I think part of it is just an expansion of the definition as people have began to offer, you know, kind of infrastructure-related SaaS is the best I've got. And so that's, like, one. And then the other is I think that there's, like, this pointing back to my previous thing about data going everywhere. Like, we're going through a huge technology shift right now in terms of how things are built and how they're run and who's being able to do that. And I think that's just driving a lot of specialty clouds. And, frankly, I think this concept of I'm going to go to one cloud and only use that cloud and keep all my things simple is, like, we're done with that. Like, the dam is broken. And so people are, like, well, I might as well find the best of breed thing that does that thing. As long as it does, like, all can work via good APIs or my cloud interface can figure out how to deal with it, like, that's not a burden for me anymore. And so I think that's causing demand. I'm, like, yeah, I see a sandbox cloud or I see an authorization cloud, right? And I'll use it because I can just import and start working. I don't have to go and, like, make some massive, you know, investment in the future of tech debt just to, like, understand how to work with this one thing. So I think that's causing a proliferation and demand side. And then there are the kind of external factors. I think sovereign clouds are here to say, I think if you believe in that kind of splinter net idea, you're going to see a lot more rules around sovereign data. And that's going to form fragmented clouds in those markets. And obviously, you know, NVIDIA is doing this as well. They're building a distribution channel by helping to support new operators. They don't want to go through their biggest customers slash biggest competitors called the hyperscale clouds, right, who are so big. And I think if we were in any other time, and if you all are listening out there, you know, if we were at any other time, we would have trust busters. They would be broken up, right? They're too big. And that's just not where we sit. And so I'm glad that there are market forces that are pushing, for example, NVIDIA to try and have other clouds that, you know, can distribute their product. I think that's good. I'd love to see that continue.

Jake Moshenko33:03

So, yeah, super interesting. I want to double click back on to your second point, which was how we were going to sort of smarter clients and wider distribution of data. Yeah. I think as an industry, we've made this journey like back and forth a few times now, right? Like the very first computers were the size of a warehouse and people would rent time on them. And then we had, you know, the big mainframes. And then we went to personal computers. Yep. And then we went to cloud, right? Like we went back to the data center and then everybody's laptops got really, like really powerful. And I was like, wait, my iPhone has more power than the servers that I'm running my company on. Like what's going on here? So then we got like really powerful hardware again. Do you see another reversal of this trend with AI? Because like right now, if I wanted to run a frontier class model, it's going to cost me a million dollars in hardware. So it's going to be a while, I think, before I have that level of power on my desktop. Do you think that that's reversing the trend?

Zac Smith34:07

I mean, I think you might have pointed out a very, very particular and narrow use case of Frontier model. I think the big thing is actually distributed systems. And when we first met around the kind of Kubernetes, you know, world of, I don't know, portable software and distributed systems, that's gotten really pretty good. You know, and that combined with some better technologies, like what you guys are doing with your database, with people like MongoDB did, you know, what Cockroach did, you know, I think also what other SaaS companies have done and provided like magical backends, has really just caused this like new way of distributed, you know, apps. And so that plus, like, I think in our little entry point, we were talking about, you know, 0.1% of society going to a couple points. This is like more developers, more software, more software in more places. I think that is a new shift, right, which is going to cause a lot of workload to end up in places. Hey, if it's already easy to put it in all the places, like kind of wherever it needs to run, like, well, what's the difference between that and like running on 100 servers or 1,000 laptops at my company? Like, I don't know. Is it really different? And so I think we're going to see more, quote unquote, on device, on premise distribution doesn't necessarily mean people are like getting back to building their own data centers in their closet. But like, what's the difference between having it on your 10,000 laptops and having it in 10,000 little VMs? I just don't know that there's a big difference, right? At least like, we still have to think about all the same problems. Like, oh, man, I don't really trust where that is, or they can do anything to it. Or, you know, like all those things are the same problem, because you've taken it out of your little central, you know, castle, and which you still have those problems to think about. Anyways, they were just like, you could put a big, you know, eggshell around it or something.

Jake Moshenko36:04

I've noticed a new trend, which is people walking around with the lids of their laptops open, because they don't want their like agent to stop coding or stop serving traffic or whatever. So, I mean, from that perspective, you're absolutely right. Like 10,000 laptops is becoming the infrastructure for a lot of these things. Yeah.

Zac Smith36:21

And so, you know, I think we'll do better with devices over time. I'm pretty sure that those guys over in Cupertino are going to make really cool laptops that do a lot of things, like really cool phones that do a lot of things. And maybe they're not going to get in the server market, but I think people are going to have servers on their desk or on their laptop again, right? But, you know, I think that's kind of like the side effect for a very narrow portion of society. Like people are doing certain types of work. But, like, the idea that software is going to ship into a ton of places, I think, is new. It's just becoming easier. And whether it's good or bad, like, I don't know, there's going to be a lot of crappy software put in a lot of places, too. And so, I think we're in a brave new world, hence why, you know, I think both connectivity and I'd actually love to kind of turn the question back around, authorization, like, who is this? What is this? Like, is it allowed to do that? Like, that's going to be a hard thing to ask. Like, we need to ask that a lot. And, you know, I think you and I are probably on the same page there, which is, like, there's some structural shifts that regardless of I'm putting it in 20 or 30 cloud data centers, some of which might be in countries that I don't have a lot of access to or don't know, or operators that I've never met and don't understand or trust, or it's four layers down in the supply chain. I don't even know that, like, it's running on Supabase, which is running on something, which is running on Amazon, which is running on somebody, you know, like, I don't know. And so I think we're going to have to kind of prepare for those questions anyways.

Jake Moshenko37:52

Yeah, that's actually a great segue to this next question, which is, what's an assumption about cloud or internet architecture that you think that AI is about to invalidate?

Zac Smith38:06

I think the one that I think about the most is that the computer that your thing is running on is safe. Which we kind of being like, oh, like, Amazon's, like, really good and taking care of, which they are, they're great, you know, they have a lot of thought about all the things. I think you just can't trust pretty much any of your infrastructure. And so I think being able to...

Jake Moshenko38:30

I already can't sleep at night. Like, what do you mean I can't trust my infrastructure? What do I do about that?

Zac Smith38:36

I think you're going to have to design your software to understand that you don't trust where this is running, or you might not be able to. Okay. And I think part of that means, you know, kind of end-to-end encryption. I think part of that means, you know, knowing what you're connecting to, like, where you're doing it. That's what I'm focused on. I'm sure part of it means, like, fine-grained authorization, meaning, like, are you still allowed to do this? What happened if I can't talk to you? What happens if, you know, I can't talk to you again in an hour?

Jake Moshenko39:04

What's going on?

Zac Smith39:05

And so I think, like, these are just design principles we have to take and say, like, make the assumption that your app, or whatever you want to call it, and the related data is going to end up in way more places, some of which you will not be able to control, and then start working backwards from that.

Jake Moshenko39:22

And what's the, what role is AI playing in making that trust, like, decreasing that trust, just making things easier to build in general, or how does, help me connect the dots? Yeah.

Zac Smith39:34

I mean, for me, like, let's go back to my sports, you know, management system that I've, you know, effectively have not written a line of code on, but I've definitely seen a lot of scrolls on my terminal. Yep. The abstraction layer is very powerful. It's very rapid. And it's so interesting, even as a guy who's like, oh, I'm aware of all the things that could go wrong here. I can, I have learned to trust it very quickly, because the benefits are so good, you know? And so I'm not code reviewing this stuff. I'm just hitting yes, right? And so I think, you know, the AI thing is just, this is an incredible abstraction tool that makes, you know, making software and managing these things so much faster, and so much more accessible to people who are not actually system architects and software engineers, like me, I'm not, right? I can think about it, but I can think about it, and I do kind of a part of this industry for a living, but like, honestly, I couldn't tell you if this code is written well or not, right? And I think that therein lies a big risk, right? I don't think you can be a Luddite and basically say, we shouldn't use this, it's not going to, this has problems. I think you're just going to have to prepare and say, yeah, we're going to actually ship a whole lot of, you know, wrangly-ass vibe code. And that, we need to just acknowledge, we don't think is going to be very good. Right.

Jake Moshenko40:57

So the AI contributions that, like, the infrastructure as code layer and potentially running some of these distributed systems, those are just, in general, going to add a level of abstraction that we may not have full visibility into or fully understand. And you're saying that, in the end, what we're going to have, our takeaway is going to be, like, I don't necessarily trust these things as much as I did in the past. Yeah, I think you're right.

Zac Smith41:20

I mean, because I basically went to my one team who handcrafted the infrastructure and pipeline around it and then made sure, you know, we've just opened that up. And even if it is, you know, done with care, I think the velocity is increasing a ton. And that's just going to mean that we have to, you know, think differently about some of these problem sets. So, yeah, I would say there is another one, which is, I'm not like a security wonk. But I can certainly see how, you know, this mythos stuff or whatever. I mean, like, AI is going to do bad things, too. Like, it's just going to hack your stuff. I mean, like, it's pretty crazy. Like, the open AI, you know, hugging face. It's just like, what? Insane. These are probably some of the smartest people on the planet. And this, like, this is just going to go run amok. Like, what if you're not, like, spending a billion dollars on payroll for, like, engineers? Like, what are you going to do then? And so, I think, like, we're going to have to prepare for a very untrusted world, even from our own systems. Yeah.

Jake Moshenko42:28

And, like, as society, we're in kind of a weird spot right now. Because we don't have access to, like, if you try to use Fable right now to do security research, it'll be like, no, can't do it. So, we, you and I, who run companies, don't have access to this level of defense, right? We can't ask it to find flaws in our software. But an attacker who is potentially using some open weights model or something is able to go and just get access to those things and do research against our software that we can't defend against. So, I think we're in a pretty bad spot right now. And I hope we get that fixed up pretty soon as a society. Likewise. Yeah. Earlier, we talked about thousands of clouds. And I think I understand a little bit better now, right? Like, even though AuthZed mostly runs on top of the hyperscalers, what we're building as infrastructure could conceptually be seen as a cloud. I understand that. Do you think, though, that at the end of the day, like, are the hyperscalers the ones who are most well-positioned to win this AI race? Or do you think we will see some sort of democratization in the long term? Yeah.

Zac Smith43:37

Well, I mean, currently, the hyperscale companies are extremely well-positioned. You know, they got a lot of the raw materials. They got a lot of the capital. They have...

Jake Moshenko43:50

One in particular, right? We're talking about Alphabet here? Yeah.

Zac Smith43:54

And they've got just, you know, incredible reach, right, from a customer standpoint. And so, like, I think there's a lot that says these existing kind of dominant infrastructure clouds are going to kind of increase their dominance as things get more complicated and more important to society. People are going to spend more on technology. They're going to run more software. They're going to have it do more things that are going to require big investments. And I think that's going to kind of that cycle will repeat itself. Now, maybe there'll be a change, at least in the U.S., in terms of kind of, like, approach on monopolistic situations. But, like, for now, I think that's just going to, like, market-wise, if there's no, you know, hand on the scale, like, is going to keep going in that direction. My counterpoint to it is that although we have kind of, like, extremely dominant, you know, infrastructure operator clouds, you know, the technology behind that is more of a race, right? So, NVIDIA being, like, the dominant chip supplier from a valuation standpoint and certainly at the AI race, you know, they're not a cloud. And so, they're going to definitely try and continue to find a way that their technology, you know, reaches the end user. And I think Jensen is, you know, very aware and maybe even more so of what happened to Intel, you know, where they were the dominant brand. Everybody, you know, talk about an amazing brand campaign, right, in terms of Intel inside. And no, they didn't even make a computer. And yet, everybody was like, I need x86. I need Intel. And then look how quickly that changed because they lost their distribution channel because everybody went to the cloud and these hypersteel guys turned them into a supply chain partner, not a technology brand, and said, hey, that's actually my customers, says Google. Not your customer, Intel. That's my customer. And we'll run them on whatever we need to run them on. And they don't even need to know about it. And I think that Jensen's super aware of that. So, he's going to kind of continue to push his technology. Doesn't mean it will always be better. He's got a big fight coming against it, like with Google, for example. I mean, these guys have the full integrated stack, right, the whole thing. And I think you can see that. I think Claude primarily runs on TPU, right? And, like, Anthropic runs on TPU. And Google, like, okay. And then I think that there's, like, a third vector that might create some competitiveness, which is, like, I have not met a developer recently who, like, I don't think in the past few years who has, like, signed up for AWS. You know, they sign up for, like, a serverless cloud, or they sign up, they use Cloudflare, or they use, like, Bursell for this, or they've got, like, this railway thing. Like, they use this state, you know. Nobody likes the big, complex. It is so freaking complicated. Like, talk about high speed.

Jake Moshenko46:59

They made it so powerful, but so hard to use. So hard. I remember with EC2, you used to be able to spin up a machine, and it was like, boom, I'm on the internet. Let's go, baby. Let's go. Now you've got layers of VPC, and, yeah, all kinds of things. So totally in agreement. Okay. Yeah.

Zac Smith47:17

So I think there could be, you know, some alternative clouds that really take the next wave of users because it's, frankly, too complicated to use your dad's cloud. I don't know. I haven't met a customer in a while that says, we're an Azure shop, and that's it. And I'm like, they're like, no. They're using all kinds of stuff. So I think that's good for startups and other independents. You still have to be awesome. It's a big, the high bar. The bar is high. But I do believe that there's more room.

Jake Moshenko47:46

I think NVIDIA specifically has an interesting fight ahead of them for two reasons. One, there's just way too much money flowing into that company to not attract all kinds of attention, like unwanted attention. And then the other one that I think people aren't talking about enough is that their stock has basically only one way to go, in my opinion. I mean, we've been saying that for a long time, but, like, stock-based compensation for future NVIDIA employees is going to be a difficult thing. And on top of that, all of the existing NVIDIA employees who built what they have today are very wealthy, and they're all flight risks, right? So you're looking at this challenge of trying to backfill for these very critical roles, and you can't properly compensate them because your stock price is so high. Like, there's just not, I don't think people will be pricing in a lot of headroom. So I don't think anyone's talking about that enough, but I think that's a huge challenge for NVIDIA coming up. Yep. There's a question that I just have to ask you, because I think you're probably one of the best suited people in the world to answer this question. You know, not to fluff you up too much, but-

Zac Smith48:54

I hope it's about the double base. I hope it's about the double base. I can do that.

Jake Moshenko48:57

Yeah, it's not. I'm sorry. At Packet, you took the bare metal layer, and you made it available via API, right? AIs are fantastically good at calling APIs, right? Have you connected those dots together yet and just let an AI go wild and start building out your infrastructure and your core network and everything for your company? Or do you still have, like, approval layer there?

Zac Smith49:22

We still do have, and it's mainly just if I had unlimited money, then I would probably be okay with it. Actually, like, I feel a little bit sad about this because there's not enough. You know, I think we moved the needle a little bit at Packet. The business doesn't live on. The product doesn't live on anymore, but it did create a, you know, movement. There are a lot more bare metal providers that have robust APIs and infrastructure, you know, clouds that can offer. But actually, it's kind of crappy. Like, there's a lot of bad ones out there. Like, there's the hyperscale stuff that worked great. And then there's, like, it falls off real quick. So, unfortunately, I don't have all the access I would like, you know, that I could unleash my agents on. But, you know, I think a lot of this just has to do with, you know, spend, right? And just being, like, cognizant of that because AI is also, like, super good at spending money. Like, really good at being, like, yeah, how about we add that? You know, I'm like, wait. And I have found, at least being in the infrastructure business myself, is, like, you also, like, cost is a part of the product, right? Like, we need to be cognizant of our cost structures. And so, I've been a little hesitant there. But the area that has been, I mean, we basically, at Datum, have invested super heavily in Kubernetes and the, you know, kind of CRD operator experience. And it's, like, great platform, build platform. So, at that point, we unleash it heavily against that layer of infrastructure, but not against the raw physical inputs.

Jake Moshenko51:07

Yeah, I think we're basically in the same, we have the same philosophy, right? We're, like, let's let the AI write the infrastructure, like, write deterministic infrastructure that we run. Right. Yeah, I got a chance to go and meet with Matt Garman a few years ago, the CEO of AWS. And I asked him, like, how long until an AI is, like, the operator? And it just has raw access to, like, start and stop VMs and put software around them and things like that. Just do whatever it needs to do. And he looked at me kind of thoughtfully, and he was, like, long time. So, I think maybe he's thinking the same way as you and I are as well.

Zac Smith51:46

Did you enjoy reading Gastown?

Jake Moshenko51:48

I don't feel like I was on the proper substance to have enjoyed that as much as I could have. But it was, yeah, Steve really likes to write, and he's a really great writer. By the end, I was, like, I'm in. Let's go. And then he was, like, he posted some stats about how much money he spent or whatever.

Zac Smith52:12

I'm, like, you know what?

Jake Moshenko52:13

Maybe let's not go yet. Yeah.

Zac Smith52:15

That was the exact part I was referring to, I think, in the openings, like, few paragraphs of his, you know, entrant to this crazy, you know, psychedelic, weird, you know, experiment was, like, and then you will need unlimited money. Yeah.

Jake Moshenko52:29

He's, like, hey, maybe don't do this if you're paying for it yourself. Yeah, exactly. Yeah. All right. What are you most excited about right now?

Zac Smith52:42

So, I guess, like, in the early 2000s when I got into here, like, new internet, new possibilities, new providers, new ISPs, new CDNs, new hosting companies, new software languages, 64-bit, Linux, like, wow. Like, lots of future. I sense the same thing right now. And I really love that we have new clouds in the market. I think it's awesome. I think it's great to see new operators and new services make it. And I think there's so much to do right now that we can make a better version of technology, a better version for society, a better version of the internet. And I feel that kind of, like, possibility happening again right now. So, I'm really excited about this new wave of builders and building entire services and companies that looks like it can be done at a fraction of the cost, which means more people get to do it. Yeah. And so, that excites me tremendously. I also think that, you know, more competition is going to create better outcomes. And I'm just excited to see how that works. Hopefully, I can stop staring at my phone, just talk at things. And, you know, also, hopefully, more people can have the experience that I've been able to have. Like, as you were mentioning earlier, you know, like solving your own problems with software. And that, to me, feels like such an opportunity for our society, you know, to basically allow more people to dream and, you know, not, you know, do my day. And it's going to be tough. And, you know, I just got off the call. I'm going to participate in a workforce development organization here in New York called Pursuit.org. And, you know, we're forming the AI Jobs Institute to help retrain the whole swath of society. Because it's going to be not even, right? And I think my brother read this book called, what is it? It was about the Luddites, the Luddite Revolution. It was Blood in the Machine. And he was telling me about it. And it was basically like, you know, this time in the Industrial Revolution where, you know, something like four out of five people were employed in England in, like, some sort of textile thing. And then, like, 15 years later, like, it was all gone, right? Like, it was, it was, most people were employed, like, on looms in their living room, you know, like, looming, you know, cotton or something. And, and the Luddite Revolution was really more about worker representation, which was, like, help, like, we're not, like, fighting the machine. Like, we're fighting that the government and the companies who are benefiting from this aren't helping us through this transition. Yeah. And I think we're going to have a similar challenge, which is, like, I think the opportunity is great, but it's not going to be easy for a lot of people. And society in general is going to have to, you know, work on that. And right now, we don't have a lot of strong muscle. I mean, in terms of helping people, you know, who are 45 years old and their job just got turned into nothing. And we also don't have good apprenticeship programs in other ways where, well, how do you create 10x engineers if everybody starts where? Where do they start? You know, and, like, how do you create fully empowered, full-stack leaders who can do all these things when they're young? Okay, like, you got to have programs, training, mentorship. Like, I always call it, you know, in the trades, like, electrical and plumbing and related. I've spent a bunch of my life being very adjacent to that world. You know, you don't let a brand-new electrician, like, work on your high-voltage electricity stuff. What you do is you, like, have a brand-new electrician or somebody who wants to become a master electrician spend time for years with a master electrician, right? And together, you know, like, you train the next generation of workers, like, so that you have another master electrician because you don't want beginner electricians doing work on complex systems or nuclear, you know, power plants or something. Like, I think we're going to need some more similar kind of construct with technology and other kind of – otherwise, I think it's going to be really hard for us to – I can already see, you know, my son's going to college next year. And I'm like, wow, like, what kind of job market does somebody enter into when the need for entry level is going away? So, oh, I just gave you my, like, super dystopian problems along with what I'm most excited about. But I think they're going to come together, you know. You can't look at them and be like, it's all going to be amazing. I think, like, there are a lot of possibilities. But we got some stuff to work on.

Jake Moshenko57:30

I mean, just like you, I, you know, I was starting to get into all of this stuff around the internet revolution in 2001. Everything was future. Everything was amazing. Everything was possible. The interesting thing between me then and me now was I think then when you're growing up, your entire life is unknown, right? Like, you can't see around any corner. Like, you're going to go to high school. Well, which one? What are you going to major in? You're going to go to college. Well, which one? What are you going to major in? Are you going to get a job? Where? Are you going to have to move? Like, your whole life is, like, the decision tree is just totally unmapped. And so when someone is telling you that, like, the future of work is going to change and the entire world is going to change, you're like, whatever. That's just, that's life, right? Like, that's how I live. Now, 25 years on, I felt like I had a pretty good handle on being able to see around the corners. And, like, yeah, I was trying to change the world in a way that, like, I was in charge of. Right. Like, be the change that you want to see. Like, I'm going to change the world in this way. And then it's going to be world, you know, sub one with Jake's addition to it. But now I feel like with AI, we're back in 2001. We can't see around the corner. But now it's a very scary thing versus, you know, me 25 years ago where I was like, oh, this is great. All things are possible. All things are possible. Once again, all things are possible. But we don't know which things will get done. Right. So, yeah, that's my take. I asked this question for everybody, and I think I kind of already got your take, but let's just put it down on the record. Are you an AI doomer or a futurist, first of all? Right. Like, do you think AI is going to take all the jobs and we're all going to be on UBI and miserable and live in Hubble's? Or do you think, like, AI will unlock better medicines and cancer treatments and spaceflight and intergalactic travel? Which way do you go?

Zac Smith59:17

In general, in life, I'm a pretty optimistic person. So I'm a futurist in this regard. I'm not a doomsdayer. I think, you know, as I've gotten older, there's more gray hair. I think it's actually made me, you know, just have to practice the optimism, you know? And I'm, like, excited for that world. Like, I think we have a lot to work on in our world. And I'm way more interested in, you know, seeing how we can positively shape that. So maybe I need to go back to Jake as a 17-year-old and have your, you know, starry-eyed optimism, you know, infect me more often. Right? Because I do think we, as a society of people, like a species, we're really good at, like, creating stuff. I think we're really good at figuring things out. And I think we're at this moment where we're going to have to look at a lot of problems at the same time. Climate change and use of technology and how our world is, you know, interoperable or not. Like, we're going to have to, like, a lot of things. How information, like, we've unleashed this huge internet, yet we don't know what's true. Like, you know, AI is going to just make that all way, like, we have some really scary, like, challenging, thorny issues to work through. But I've been kind of pleasantly surprised by, you know, some younger people around me and whatnot who have just, like, approached that with, you know, an optimism. And I hope to continue with that myself. So I'm going to go for optimistic. But I will say, as an investor, I've been putting more of my money into human things. And when I talk to my kids about where they're going to go to college, they're like, what should I study? I'm like, something that makes you a good human. You know, just do something that makes you better as a human being. And then, like, I think that will make you more valuable in the future, right? And I believe the same thing with businesses. So I've seen just crazy demand. We just got through the World Cup. There's a lot of technology in there. But, man, what do people love? They just love being together to watch the same thing and experience something very human, right? Which is awesome. And so that's where I see the flip side. And I'm like, yeah, nope, not everybody's going to be out of the job. We're just going to hopefully get a lot more efficient at a whole bunch of stuff and allow us to do other great things.

Jake Moshenko1:01:50

Yeah, you touched on something really interesting, which is I think that as a society, we've been trying to make ourselves more robotic for, like, the last 25 years, right? Like, get the same output from the same position on a factory line, time in and time out, right? And, like, you have these CEOs and VCs who are trying to, like, you know, max themselves in every way possible. Like, just be as robotic as possible. And maybe now we're like, you know what? Let's let the robots do robot things and we'll go do human things. We'll go do human things.

Zac Smith1:02:21

That sounds pretty awesome to me. Yeah. Yeah. All right.

Jake Moshenko1:02:26

Well, it's been an absolute pleasure. If people like what they heard, where can they hear more of your thoughts? Where do you like to hang out?

Zac Smith1:02:34

Well, I'm usually on LinkedIn. I'm not a Twitter ex-person. So if you go to LinkedIn, I'm ZSmith. So it turns out I got that early. So you can find me there. I post sometimes. You can go to datum.net. I've got my blog and other stuff that I put my ramblings on. And you can always find me if you're in New York. Stop by. Say hello. I'm downtown. Yeah.

Jake Moshenko1:02:57

Thank you so much for coming on. I love the conversation. Absolute pleasure, as always. Thanks, Zach. Bye. Bye.