Anthony
00:00:00.000
I'm not a fan of the notion of vibe coding. I know that there's a lot of people that are really, really say, this is gonna be the way we're gonna be building stuff in the future. As somebody that runs such an Important operational system. And that's what we really are running here, is we're running systems. I think it's really important for all of our engineers to understand the systems that they're building and you only do that by then reading through and understanding the code that's generated and you can't remove that from the equation. Not yet, at least in my opinion.
Viktor
00:01:47.849
I mean historical order. Uh, calendar order. Order, yes. In order of appearance.
Darin
00:01:55.669
I cannot believe the B word crossed your lips. That just sounds like a horrible thing to me. Uh, but TCP, yeah. Okay. Well that sounds good. I'm gonna have to disagree with you. Uh, well, okay. TCP? Yes. Because TCP underlies one of these other things, I think DNS is a big deal, even though that's not TCP. That's UDP.
Darin
00:02:18.393
I'm saying Torrent hasn't mattered since, uh. Lime wire and Napster all went away. Yeah. Okay. Whatever. You may be asking why we're asking all these questions. Well, on today's show we have Anthony Eden on from D and Simple. Anthony, how you doing?
Darin
00:02:34.345
you agree or disagree with Viktor that blockchain is one of the most important bu building blocks of the internet?
Viktor
00:02:44.693
Are we not gonna write internet on ledgers now? Whoa, whoa, whoa. Uh, I'm confused. I thought you're forward thinking person.
Anthony
00:02:57.743
we can go on that a little bit. But the current one that they have definitely is U-D-P-T-C-P, protocols like that. The, the old stuff, the stuff that's been around for a while.
Darin
00:03:06.656
And guess what? That stuff is going to be around for a long while. Even during this, how shall I say, AI insanity. we have AI without TCP and UDP?
Anthony
00:03:18.823
I mean, you can, but I don't see it being anywhere near as valuable in a non-network world. But hey, who am I?
Darin
00:03:25.738
obviously we'll have to get into AI a little bit more, but an Anthony is actually the founder and CEO at DN Simple. and of course she took on immediately because somebody saw back in the late two thousands, 'cause we can say that, somebody saw a great value of this and they automatically threw a billion dollars. Right?
Darin
00:03:46.153
Wow. I mean, that's, it was right after, right after the dot bomb of 2000, everybody had billions of dollars laying around.
Anthony
00:03:57.103
I don't, you know, part of it is, I think it was old tech. It's old tech, right? Like the, the first round of people doing domains in DNS. They were around in the, that. The nineties essentially. including myself, I was actually part of one of the first, domain registrars during the deregulation of.com. I was in, uh, the CTO of one of the first seven registrars. And so we were doing a lot of this stuff even back in late nineties, early aughts. so by the time 2010 rolled around when I started the in simple. Essentially we were moving from the early phase, so Group one that had built a lot of the foundations for the internet to the next sort of phase of DNS and domain management. and that's where DN Simple started up. But the, the, there was really no money. Coming, at least to me, I, I, I do have a pretty large competitor that you've probably heard of, somebody called CloudFlare. They got a ton of money to take them, so they took that route with the VC funding, I intentionally wanted to build business that was not a funded business, that was built off profits and one that was stable and could support the lifestyle that I wanted to have with my family.
Viktor
00:04:59.525
Wait, I is saying that businesses should generate profit now, or first you're against, uh, Bitcoin being building block. Now we need to generate profit kind of. Well, what's next?
Anthony
00:05:11.615
I am just a contrarian, you know, what am I gonna, who I, businesses are there to be good things that generate profit for their owners and shareholders, but who?
Darin
00:05:22.665
It. It's interesting that you brought up CloudFlare. I'm thinking back. Okay. Hmm. End of 2025. They were down a couple of times or caused half the internet to be down. Used to would be AWS bring down half the internet. Now it's CloudFlare, it brings down the other half of the internet. Uh, if AWS and CloudFlare ever go down, we're all taking a break, I guess. Uh, but I haven't heard anything about D Simple ever going down. Okay. I, I want ask this question. 'cause DN Simple is a SaaS business without being a real SaaS business, but it's a SA business. What does it take to run DNS for somebody else? I mean, we understand, okay. Oh, I need A-D-N-A-D-N-S server. Great. Yeah. But then what?
Anthony
00:06:04.327
So a lot of it's about, scale and distribution, right? So if you're running on the small scale, you have DNS servers on one or two machines, you're probably running it in unicast mode, which means that every machine has its own IP address, and DNS packets are going in and out of those IP addresses directly to that machine. And so that's kind of the first phase that we see as you start to scale up. And as, as you want to distribute more Node.js around the world to improve reliability, you switch to any cast. You start building out a more redundant network, now you have to distribute changes out to the various edges. And this is, this is where things sort of cross over that line of, okay, I, I can do this just running a piece of software to now I need services that connect all this stuff together and get the data out in a timely fashion and ensure that everything stays up and is actually synchronized correctly. So we spend a lot of our time. Building and maintaining a system that allows us to do that, which is to serve the same DNS responses from many different locations around the globe with the shortest possible path to that DNS server.
Viktor
00:07:07.816
Does that mean that you need to be close to those destinations or how does that work?
Anthony
00:07:14.311
So generally any cast, the BGP routing protocol is gonna try to find the shortest path to wherever, uh, the, basically they can find to the nearest node that broadcasts an IP address. And so, yes, your proximity to the actual resolver that's asking the question does matter. It's a little bit more complex because a lot of people go through public resolvers like Google's 8.8 0.88, or they go through 1.1 0.1 0.1. Or these are public resolvers that are run by organizations. those also are running any cast. So you kind of have this. set of layers almost like an onion, where you're starting from the outside, you're going to the closest resolver. Then it wants to go to the closest authoritative name server to you. And so that's kind of how you get the fastest DNS queries is by having an authoritative name server that's close to a resolver that's close to the person who's making the query.
Darin
00:08:07.209
I am gonna go down so many d different rabbit holes here, but I'm, I'm gonna try to stay on one at the moment. Okay. You, you brought up quad eight, quad one,
Darin
00:08:14.685
which by the way, that's Google. That's CloudFlare. When you set up an internet service at home, I'm sitting at home and if I, if I'm a nobody and I don't know what I'm doing, I'm just gonna be using the ISPs, DNS, they're gonna be providing that. If I'm somewhat technical, I may do the quad eight or the quad one. Why would I not want to do any of those things? What is the right way to set up DNS at home?
Anthony
00:08:39.272
Oh, the good news is that there's, there's no right or wrong. The, best way is the one that works for your needs. Most people, like you said, they're gonna use the ISPs and a lot of ISPs these days will just use. 8.8. 8.8 or CloudFlare. So they'll just send that out. if you are doing something a little bit more advanced and you either want to control privacy, because every time that you connect to one of those resolvers, that Resolver now knows what question you're asking, and it's collecting all that data together, and that data's being used by the companies that run those resolvers. To make decisions like, oh, okay, well we see more traffic coming from this region, therefore we're gonna invest in that region. Or, we don't see as much traffic come from here, so we won't grow future investments there. they're also using it to do security analysis. Oh, we see these types of queries coming from these locations. What are we gonna do about that? If you want to avoid somebody else looking into that information, then you can run your own DNS servers at home, which. More techie people that want to have, say, a little home lab. We'll set up a raspberry pie running a little, bit of a DNS server on it, a lightweight DNS server, and they'll do the resolver part. From their internal network. it's the same kind of thing you would, you would've seen in sort of older, you know, small corporate networks where they would've set up their own DNS server, often running on Microsoft software, you know, active directory or something like that. And they would resolve all the queries starting from inside their own network. So still totally doable to this day. And a lot of, Folks who are into tech and who like to tinker will set something like that up just so that they can feel a little bit more comfortable about their privacy and about the information that they're sending out.
Darin
00:10:16.159
Until their significant others figure out why can't I get to the website I'm trying to get to.
Anthony
00:10:21.189
Well, we won't even talk about that. That is another use case that's actually pretty common, which is, basically privacy blockers in the other direction. Or not really privacy blockers, but content blockers. So corporate networks, often they will install name servers and use DNS as the first point to try to stop people from accessing sites. They don't want them to access.
Darin
00:10:39.595
That goes down a very, very dark rabbit hole. We're not gonna go down. Uh, now one of the things that DN DN Simple has is domain registration. Again, you said you were at one of the, at the time. The CTO had a company during the great deregulation of the.com away from Dun, dun Network Solutions. Right? That that was they, that was it. So if you wanted a domain prior to, what year was that? 99. Okay. Right when all everything hit the fan, uh, almost hit the fan. Uh, that's when things were warming up to hit the fan. now that we can get domains pretty much anywhere and get all sorts of top level domains, TLDs, I'm thinking about going back to the AI thing. If we've seen all these AI startups come along, of course they're gonna wanna get a domain name. What are people doing wrong when they're trying to get a domain name today, do you think? they're thinking through as like, oh, okay. Forget the marketing part of it. It's like, are people doing silly things when they're getting a domain name?
Anthony
00:11:35.355
Yes. So the first example would be, and it's not really that it's silly, it's just they don't really know when they need to level up. So, for example, you have some fairly high profile domains that will run on their domain registrar's, DNS servers that are not really like the, the registrar in that case is not DNS first. So they've taken something that's 24 7, 365, the most important, valuable operational. Element that they have for their, essentially presence on the internet, and they put it on something that wasn't really designed to operate in that fashion. And so there's a point where it's, it's fine when you start, like you register a domain, you're just starting out. It doesn't really matter. But the minute that you actually have production traffic running through things, you need a production system that can. Make sure that you maintain the uptime, make sure that you have the distribution fast response times and generally, domain registrars, especially older ones that maybe you're comfortable with, but who haven't really moved forward with the times, aren't gonna have the infrastructure needed on the DNS side to run the operational portion. And that, that's a big mistake. the other big mistake I think that people make is that they don't, put the proper security in place to ensure that they're not gonna lose that. Domain in the future. So for example, they'll register an account with a shared login with other people from their company because there's no multi-user access control on a lot of the older registrars. And so that we've seen problems where people will just be like, well, I have this domain and somebody left my company and now no one can access it. What do I do? So there are some pretty big challenges that when we don't think through the security posture and the protection posture on, domains, which are essentially intellectual property. So they should be treated like such.
Darin
00:13:19.873
I think one of the things that I've seen most of the time, going back to those. DNS providers that aren't meant to do DN or aren't built to do DNSI think of like a lot, not all, but some of the shared web hosting providers. To me that's just a, a classic. And I'm not saying they're not, and I'm not saying they are, but it's like, that would make sense because they're focused on running to the bottom of the barrel from a money perspective. So something's gonna give somewhere
Anthony
00:13:45.213
Correct. Yep. I see that. We see that as well. I see it. Uh, where, for example, you have, hosting control panels. That will let you spin up instances of DNS servers. So you could say, well, I'll just spin up an instance of a DNS server for myself it will work. But if that instance goes down, essentially it takes everything down with it, your web presence. And so that, that's a good example of it as well from the operational side. And that's how I separate these two things. I usually think of DNS as having the two sides. One is the operational side and one is the intellectual property side. both need to be. Sort of set up and maintained in a good, reliable fashion.
Viktor
00:14:21.125
How does that intellectual side works in practice from the legal point kind of. Okay. I, I take your domain. UU me. Is there, are we talking now about, you know, I dunno, couple of years process? Or how does that work even?
Anthony
00:14:37.710
so generally you have. Egregious abuses of intellectual property. So that means somebody trying to squat the domain name from a company that clearly has a trademark and
Anthony
00:14:49.110
yeah, that actually there's well-defined processes. There's an organization called ICAN that oversees. the majority of, at least the generic TLDs that we have on the internet, both the old and new, so that's comm net org, as well as all the new ones like, app and all these other names. So there are very well-defined processes. If somebody tries to take a trademark domain and you can basically say, no, that's ours, and you go through that, and that's pretty easy. In fact, so much so that they have short timelines in terms of months or less. Now if you have a squabble between founders who both had access to a domain and they're fighting over it, well that's a whole nother ball of wax, right? Because technically the company should own the ip, but if it wasn't properly put into the company holding, it may be under one of the founders, uh, identities. And they'll say, no, no, no, that's mine. Look, it's in my account. I have all the information on, I'm the registered domain holder, because they never switched it over to the company to be the registered domain holder. It's still a person, and that person now is the one that owns it. And then you can get into a protracted legal fight with that pretty easily.
Darin
00:15:54.465
Hey Viktor, you still own the DevOps paradox.com domain and I have everything else.
Viktor
00:16:07.455
Yeah, because, uh, it's so, so obvious that it's going towards a billion dollar, uh, company that we should be worried. Yes.
Darin
00:16:21.765
there we go. okay. So let's assume a person got the domain name right, and they picked DNS correct. D and Simple is obviously one of those. Uh, I would say probably most of the big hyperscalers are probably pretty good at this, like AWS Route 53, all, all the others. Those are all good. and let's assume that they actually get the IP address set up correctly for their website.
Darin
00:16:49.298
What else are they screwing up when they're actually setting up DNS or what are they forgetting that they really should have?
Anthony
00:16:55.607
the big common one are things around email deliverability. So especially in this day and age where email is the most common, asynchronous communication for businesses to, communicate with their customers, often they won't go through the steps necessary to set up. DMAC, which is uh, uh, essentially a way to define a policy for how you deal with spam, SPF, which makes sure that you're sending from the right senders, the only authorized senders can send, email on behalf of a domain and d Kim, which is essentially a signing algorithm so that you can actually say these messages. Our sign from their origin, which means that they are coming from who we expect they are coming from and they're not tampered with. so these are the types of things that we, of, that I often see aren't necessarily set up correctly. the other thing is often just sort of change management processes on DNS. So they'll set up something manually through an interface, but they'll have no record of why, why did we do this? well, some service over here asked me to add a text record. That I add and, and I did that five years ago. Should I remove it? Does anything need it? And so you, we see a lot of that as well where there is sort of chaos in the DNS records that are operational because there's no record of why anything was done.
Darin
00:18:12.146
Yeah, I'm thinking about pretty much every domain email and it's pretty much that way. Uh, I.
Anthony
00:18:17.091
Yeah. And we, we dealt with that in two separate ways. One is with the obvious thing, notes, like we basically made it so you could put notes attached to any record inside of Dan Simple. as you're making changes, you can add a note and it will essentially keep a history of those notes. So that's one example. the other way we've seen people do it is they'll take their zones and they'll commit them into a source control like GitHub, and then they'll use, there are tools out there, open source tools. There's a variety of 'em that can actually, then when you change it and you make the commit to GitHub, it'll. Push those changes up to the various DNS providers where your zone is living. So we've seen that as well. And then finally, you have tools like Terraform or Ansible, and you can actually make domain management and DNS part of your configuration management with tooling like that as well.
Darin
00:19:02.550
I will go ahead and pull back. One thing that I looked at at your site is because I, even though I was familiar with the name, I'd really never looked at the. The actual site before and I figured, okay, well they do domain registration, so okay, that makes sense. They do DNS hosting. That makes sense. And then I started looking, it's like, oh wait, I can go into D and Simple and manage my domains and DNS at AWS and at Azure and at Google. What gives, why are you being so nice?
Anthony
00:19:31.800
Well, because we we're, I mean, look, the, we're nice, but at the same time we see an op, a market opportunity and, and there's a growing need. For organizations to spread beyond a single cloud provider. So multi-cloud is a thing that happens. It's not everybody needs it, but when you need it, it's very, it's kind somewhat difficult to do when you've put all your eggs in one cloud basket. So, for example, if you have a, a, an enterprise where you're putting everything onto AWS, that's really great. But the minute that you have to start moving things. To another provider for reasons, whatever that may be, whether it's for redundancy, whether it's for, functionality that's available in one and not the other, whatever it might be, then you'll find that you're, it's you, you kind of run into a situation where things become more difficult. And in our space and DNS we know that we're not gonna be. We don't have one cloud provider and we say we only support that. The goal is let's, make it easier to go across various providers. And it's actually really common in DNS to spread a zone across multiple providers so that you have, again, more redundancy, and the ability to shift things around easier. So we saw that as an opportunity and we decided to, give a try at building what we call this domain control plane. And that's why we, we have the facilities to manage domains at other providers in addition to us.
Darin
00:20:51.576
I'll pull back the curtain. All of my domains are hosted. Hmm. Let me rephrase that. All of my domains are registered at AWS. The DNS hosting is somewhere else. 'cause I got tired of paying the. Uh, sorry. A WSI got tired of paying the Route 53 costs every month. It's like I just got tired of it. So I ended up, I think DNS hosted depending on where the website was. So, most of my stuff's either at Netlify or at CloudFlare, so I just used theirs and it's like, okay, for, for what I'm, it's not like I've got, you know, big request per month. So it's not like, not a big deal, but thinking through your part, it's like, okay, if I, if I went to die and simple. You were talking about the notes, it's like the note. If nothing else, at least I'd be able to know what's going on because I don't have API access directly to manage, even though I probably could at At Netlify or at CloudFlare. CloudFlare probably could. So that seems like a very cool way of managing things. We're talking about all the email stuff. You're talking about, records of what the heck is this? Uh. I'm gonna ask one that's probably going to either set the internet on fire or not. should websites have a www or no? Www
Anthony
00:22:04.191
that is a timeless question, isn't it? I am in the camp of, no, www. That's me personally. Uh, what we do is we just redirect the www over to the no www. So we, when I'm going around and I'm giving people DN simple's address, I see dn simple.com always, and that's what they, they go. Now, the irony is most people, at least in the last 15 years would go, oh, dn, simple. Open up Google, type in dn, simple and follow the path. Right now, they're like, oh, DN simple. They open up chat, GPT, they type in dn, simple, and they try to follow the path. Although it's a lot, a little bit harder because there's not the same type of, references via linking from chat GPT. It's coming now more and more, but so people like the domain names, even though it's underlying, most people on the internet are never actually typing that into their browser. If they're putting into a search engine or they're, now they're putting into a generative AI search engine.
Darin
00:23:00.408
Well, let's go there for a minute because since you brought it up, how has that affected your business? I mean, obviously the links or sometimes there, sometimes not. What as a business owner, what's that like to you?
Anthony
00:23:12.918
it's challenging because the traffic patterns, for example, from Google once when they introduced the, the Google AI sum summarization on top of search queries as being the first thing you see. If somebody in the past asked an informational query, we might show up in the first year results because we've spent a lot of time and energy to build really good documentation that explains not only. What happens at dn? Simple. But DNS in general, uh, I mean, we have this very large knowledge base of support articles that, that explain each different record type. What do they mean? What are they for? Those queries now are just answered in the browser as part of the search result. Somebody doesn't have to go anywhere or they're answered inside of chat, GPT or inside of perplexity or pick your AI tool of choice. So that means that the traffic has gone down. Now, the flip side of that is that the traffic that comes through. Is actually people that are interested more and more in the product itself. So we saw a dip in terms of signups, after the introduction of these types of tools in the beginning of the year. And we've worked hard to attempt to overcome that. And we've actually seen now progress going the other direction where we're starting to see an uptick in traffic. But the difference is now it's actually good traffic, more good traffic, rather than just. A lot of traffic bots following links off of search results and doing crazy stuff, right? These are actual people, and so in that sense it's, it was difficult to deal with, but I see some positives on the other side of it as well.
Darin
00:24:46.896
So you're not the CEO shaking. Why isn't, why are all our clicks down this month? Why aren't we getting more
Anthony
00:24:52.391
Well, I may have been six months ago I was, but, but then we did the time to do the research to figure it out and understood, okay, now we see what's happening. Let's see what adjustments we wanna make to improve the situation.
Darin
00:25:04.376
in running your company for roughly 15 years, I'm assuming you're up to maybe what, four or 500 people now? Does that sound about right?
Darin
00:25:17.606
20 people to run a DNS solution and a registrar for the world. That's not possible. I'm sorry. That's just not possible. That's not how it works.
Darin
00:25:33.086
Why so small? I mean, why? I mean, sure, you probably could have grown some, but I mean. I mean even, okay, I'm doing math. Okay. I'm, I'm, I'm happy it's 20, but it feels like 50 would probably be a little more sane. But is there just not that much support issues? Am I, is is it just because it's such a, a simple, again, we're using simple a lot. Is it just 'cause it's so simple and able to manage?
Viktor
00:25:55.341
me, let me translate what he's asking. He, he's okay with fan engineer engineers, but he thinks that you're missing through the salespeople. I think. I think that that's where he's going with the number 50.
Anthony
00:26:07.491
you kind of hit the nail right on the head there because we are definitely an engineering heavy organization, so we have been, since the beginning. I think we've run everywhere from a hundred percent engineers to. Uh, maybe 60 40. So 60% engineers, 40% non-engineers. Right now, I would say we're roughly 75 25. we don't have a big sales team. Uh, we never really have, it's been self-service. we have an engineering team that, in our case, we've focused on hiring really good engineers. But we've been very selective. So rather than trying to hire in quantity, we focused on quality and we've created an environment that is very, I guess I would say friendly to engineers who are senior engineers who, for example, have families. So they have kids that are trying, they're trying to be with when they grow up. So rather than getting in the mindset of, oh, we're a startup, you have to be working 80 hours a week. And you get one week of vacation a year, and during that week of vacation, you have to have your phone and computer on you at all times. Instead, we're more along the lines of, okay, this is a business. Let's run it where you can legitimately take time off. We expect everybody to take three weeks, at least off a year, if not more. And if you need to go out and take your kids to school, you take your kids to school, right? So we've created an environment that is. One, hiring high quality engineers, and then two, retaining those high quality engineers by actually creating an environment they want to work in and giving them problems they want to solve. So that's one of the ways that we've stayed small. we've grown a little bit over the years when we've needed to, but it's usually an inflection point where we see, okay, we, we are really missing something here. We need to, to add on people, and then we'll grow a few people and see how it works out. there's very much profit, like, and growing based on our ability to pay our team members not using debt or financing to do so.
Anthony
00:28:01.813
I'm, I mean, ai, I, he, he, we were gonna, we're gonna talk about ai. We know it's gonna happen, right? And so AI is, is empowering the engineers we have to. continue doing more, but it's not really about more. It's the, I, I've been thinking a lot about this lately because I had spent a few years not writing any code. I was just in the role of CEO for the last few years and just this year I've started getting back into writing code again. Really just taking off little bugs and fixing things that, that I don't want the team members to have to spend their time on that are really well-defined. And I've been thinking, what is it about these AI tools that we find advantageous in our and how we do our work? And I realized that it's just giving me more time to think about problems and solutions rather than writing the code to solve the problems and solutions. and that's a big advantage when, when we can take an eng because engineers. There are some engineers that love to actually write code, like there's something beautiful about writing code and creating a working system, but a lot of engineers are far more interested in just solving problems. They like to see a tricky problem and they like to work through it, and so giving your brain more space to work through that, which in a lot of ways, delegating the actual code writing part out to a tool. I see that as very advantageous and I'm seeing what some of our team members are able to do now. That they have some of these tools in their hands that we couldn't have done two or three years ago without the same tools, without having more people and dedicating them to that.
Viktor
00:29:32.075
Man, I'm so happy you're saying that because in last months I entered into quite a few. Complicated conversations because I claim that the value was never in you typing the value was everything else. Typing in my head is chore. Kind of we need it. There is value in it, but kind of it's simply, you know, like cleaning your house. Right? I, I need to do it. but that's not really what generates value. Right. And, and I feel that with AI that's surfacing even more. I think that it was always present, but surfacing even more kind of, you know, solving problems, figuring it out. That's, that's the real, real job in a way.
Anthony
00:30:12.211
Uh, one of the key things that I, that I see still matters though, is once you've solved the problem, it's really helpful if you understand the solution. And so, for example, I'm, I'm not a fan of the notion of vibe coding. I know that there's a lot of people that are really, really say, this is gonna be the way we're gonna be building stuff in the future. As somebody that runs such an Important operational system. And that's what we really are running here, is we're running systems. I think it's really important for all of our engineers to understand the systems that they're building and you only do that by then reading through and understanding the code that's generated and you can't remove that from the equation. Not yet, at least in my opinion.
Viktor
00:30:52.381
absolutely not. I mean, actually you can, it's not that you cannot, it is just that the results between, I'm not going to call it five by call. I'm going to call it one shot. uh, AI doing something and you being behind the steering wheel is just tremendous, you know? Yeah, for no other reason because whatever you told it to do, even if it's perfect, you really don't know what you want.
Anthony
00:31:18.852
I'm a big fan of understanding the systems that I build and operate, by the way, humans can do just as much. they can do just as much to destroy the understanding of a system with manually written code as well. Believe me, as somebody who's written much of it, I could tell you I, I have caused many a dragons in systems and so it, this isn't saying that. AI is not going to do that same thing, but it gives us more, it does it faster than it, than we could do, thus giving us more time to think about it, to look at what it's trying to solve and to retry it if it does something terrible, it's good stuff in my opinion.
Viktor
00:31:54.623
we were talking about ai, how it affects your work. Is there a story, AI from the business perspective as well?
Viktor
00:32:03.708
of, yeah know, kind of like, does DN Simple or DN Ss in general or will that be affected by AI as a business? As a technology is whatever you want, you wanna focus it on.
Anthony
00:32:19.309
I think that there are already a few experiments on the IP side, more around sort of generating names based on ideas. So there's quite a few folks have written. tools where you say, put a concept into one of the AI tools, and it will generate you a list of 10 to 15 domains that may or may not be available, that are maybe good in that concept. And so we've seen that. It's an interesting experiment. I think that a lot of people search for a good name, so having ways to generate potentially good names. Yeah, I'm, I think that's a, a good idea. More interesting will be as these types of tools get applied to the, data that is potentially coming through, for example, in DNS at our edges. So for example, if it is, and we haven't done any of this yet, I don't see this in the foreseeable future because again, we have an operational environment to run. We have to focus on that first. But when the time comes, the tooling is, is that we're able to. Learn more about what is happening at our edges and train systems to generate potentially information about what's happening at our edges. When I talk about edges, I mean where we're getting the DNS queries in and sending the responses back out. That to me is something that we might see some interesting, uh, new techniques come out of. So it's a little hard right now. There are also, it's already being done with machine learning. It's been being done for years with machine learning. So generative AI might be just a different take on that.
Viktor
00:33:44.477
How, I mean I always thought it's very important, but now when we are talking, I started to think. How much does it matter to get a good domain if you're gonna be searched through Google and soon you will be found by, I dunno, check g pt, how often do people actually really go to, specific domain because you told them so, and when I say domain, I mean not the name, right? I know that, I dunno. Let's say it's called Viktor. Uh, Google will find it is if I type picture without comm.
Anthony
00:34:21.503
Yeah, I think, I think the answer to that comes based on how much competition you have for that name. So for example, imagine Viktor Viktor dot com, whoever has that, they're probably gonna have, they've had it for a long time. So overcoming the age and history of that name in China in terms of search results or quality of the domain. Is gonna be challenging. So that might be where the choice of name might matter because yes, you might be able to get, Viktor dot io, Or get Viktor dot biz. But it's probably gonna take a lot longer to competitively rank in terms of search NGINX and, and also in, generative AI search NGINX as well. having said that, th things are changing. Constantly with the models, with everything that's happening on all the, like with the competition between Google and Open AI and Microsoft and everybody trying to figure out how they're gonna make the next leap. and these changing models, uh, what I say today in terms of domain names might be very different a month from now.
Darin
00:35:18.537
We can probably guarantee that we'll be different in a month from now. Okay. While you were talking, I actually came up with three questions. One that we didn't finish before. We were talking about the dub dub dub. Earlier you said you automatically do a redirect old school. That means I have to go and put in a, a redirect somewhere for that. That's something that you include, I guess as a check box of if you want it
Anthony
00:35:37.107
Yeah, yeah. Well, we basically set, we have a record type that's called a URL Record, and we can say, I want to put this record on that redirects this other thing, and we have it set up so that you can say, I also want to have a Lets and Encrypt certificate. It with it so that it, it actually hits the end point with an HDPS termination and does the redirect so that you don't get into issues where, oh, it's, H TT P, therefore, it's marked as insecure on the redirect. it's a subtle thing, but it's something that we've done for a long time that helps make that redirect work as well as possible in all browsers.
Darin
00:36:08.345
And as far as I know, nobody else is doing it. So where can I put in my credit card?
Anthony
00:36:12.485
I think there's a few others that have started doing it, but uh, yeah. Dean symbol.com, that's where you put in your credit card.
Viktor
00:36:19.715
But I like, uh, I dunno whether that's common or that's specific to you. I never thought about let's encrypt on that level. Kind of in my head. I'm always, I dunno, let's say my application is running, Kubernetes should be accessible to Ries, that domain, whatever, let's encrypt trans there. Cool. Kind of like easy, busy. But is, is this you or kind of, is that the common thing with DNS registers kind of like, Hey, lets encrypt. Is there kind of no worries about it?
Anthony
00:36:46.340
It was not, it has not been for a long time. I think some are starting to come around to the value of doing it. We've actually been, supporters of Lets Encrypt since their inception. We were one of the earliest financial supporters of it, and, and even as a small company, we've conti because we believe that it's been really, really important. For the internet as a whole. And so my CTO came to me many, many years ago and he said, we really should be include lets and encrypt certificates inside of the, and simple. And we found a way to do it where at the lowest tier we could essentially, I mean, they're free, so you get them, you never pay for their certificate. But we had a way to encourage people to upgrade to our higher tier plans by supporting certain variations. For example, wild card certificates. Would only be available on a higher tier. And so when we built the Redirector in the first round of it, it didn't support HT PS, it was only redirecting over HDP. But when we, when we were able to do both lets and encrypt, and we were able to then say, well, well this is interesting. We can now push these certificates automatically out to our redirector. And we found that it was really useful. So yeah, it's been, it's been super valuable to do it that way.
Darin
00:37:56.381
How many edges, edge points? See, I'm making up new terms here. It's not an end point, it's an edge point, how many do you have
Anthony
00:38:02.996
So as of today, we have, 14 edges is how many we have today. So, there are definitely bigger ones. Cloudflare's huge, by the way, CloudFlare has hundreds of them around the world because again, they have that, those deep pockets to do it. but with 14 we've done fairly good coverage of Europe. Australia, uh, US and we're starting to, to go into South America now. the way we've designed it is we have those edges and then we have our actual origin name servers, uh, which are internal and there, and our origins. We have nine right now, and we have any cast running all the way through it. So even our own edges, we'll use the shortest path, any cast to get to the origin that's closest to it.
Darin
00:38:45.724
That's interesting. So. If you've got 14, you've got nine. You're covering obviously the full world, but full world from a speed perspective is
Anthony
00:38:58.274
It's good and yeah, it's, it's good. So for example, gaps we have right now, Africa, we don't have anything in Africa now we might. Spin up a a point of presence down in South Africa at some point in the near future or in the Middle East? we're starting just now to spin up presence in South America. there are certain areas of networking, for example, the Southeast Asia area, which can be very complicated 'cause they have a lot of countries that have very, disparate routing, like lots of different companies that are small players that are providing routing. So you have lots of hops. So those are the kinds of areas where maybe you might have a slower response time. but generally we've, tried to build it so the vast majority of our customers, the folks that are using us, are getting good response times, regardless of where their customers are in the world.
Anthony
00:39:46.930
So, no, we actually have a mix of on-demand bare metal, which is something that we've been able to start doing, in the last few years as, as well, I would say last. Maybe five or six years as on-demand, bare metal became more of a reality. Unfortunately, there's been a bit of a pullback in the industry from allowing like, so it's been harder to get at that. We also do run, managed hardware in some colo facilities, especially for our origins because those are beefier machines.
Darin
00:40:13.086
My last question and then, we'll, not, not the last, last question, but the question that I'm catching up on Viktor was talking about, you know, how, how are the developers working? So I imagine you have developers that are mainly for your self-service web, that kind of work. What other skills do your engineers have? I'm thinking more, I guess more towards the backend, like the origin servers. What, what are they having? I mean, I'm assuming you're running bind or some variant of bind,
Anthony
00:40:41.106
So we're not, I'm gonna surprise. Yeah. This is what I'm gonna surprise you with. We actually run an erlang based name server in our authoritative name servers. I actually wrote the first version of that. Hmm. It was like 2015, I think is when I wrote the first version of it and we launched that, I think it was in 2017, and we've been running that ever since. And then we use go as a language to sort of transport messages around from our. Core central systems out to our various, what we call our zone servers on all these different edges. We use kind of like a spoke a hub and spoke design for getting messages out. and then we have, uh, Ruby and rails for our web applications. So most of our engineers are comfortable in. One or two, if not three of those. So the LAN is the one where we have the fewest people historically, but LAN's actually had a bit of a, resurgence thanks in part to Elixir and Gleam and other languages that were built on top of the LAN beam. and so now we're seeing, we, we are seeing more and more people that are interested in the Erlang space. so that's kind of cool and fun. that's one of those areas where we encourage team members to. To go look at the code and try to understand what it does. And that's another thing that we get with generative AI that we couldn't get in the past. You can ask it to like, explain what's happening in this area of code and it will do a pretty good job of explaining what's happening and the interactions with other parts of your code. I had our CTO who came to me and showed an example of how it taught him a particular code path inside of our er lang systems. And he doesn't know er lang. I thought it was, that's pretty impressive that it was able to do that. And so, uh, yeah, so those are the skills. And then of course, networking skills. We have a number of team members who are pretty good with understanding routing, BGP routing, understanding, general system management and how to, so we use a lot of containerized systems, how to get those spun up, how to build deployment systems and just, it's a big mishmash of skills and we kind of hire generalists who. Are curious and who have been operating internet infrastructure, internet technology for a while.
Darin
00:42:49.228
So what you're saying is if Lan was good enough for telco switches, it's good enough for DNS.
Darin
00:42:58.131
That's very interesting. So with ai, you're saying, I agree with you using generative AI to help explain something. Would it make sense to have it take those comments and stuff 'em in, or whatever it came up with as comments in the code? Would that, or is there, there's gotta be a better way to do that because I don't want the comments in the code, but I want, I want them tied together somehow.
Anthony
00:43:20.321
It's a good que I haven't really thought through what, one of the things that we lose, I think in some cases with these AI because it's, it's non-deterministic, is if we don't keep a history of the conversations that we've had, and if we're not evolving those conversations alongside the code, then we lose track of what it told us because that history could potentially be gone. I think maybe in the future as. people get more comfortable using it, and as the tools get better at retaining the history, uh, we'll see more and more tooling that comes outta this as like, walk me through the history of this part of the code. And it will be a combination of what it told us in the past. Maybe looking at GitHub repositories for example. That's one way that we've, like what we've started doing is, we'll, when we get output from something like Cursor that says, here's a summary of what I did. We'll use that summary because it's often fairly clear and detailed, and we'll put that in sort of the issue or in the poll request. And so now we have a history of it inside of GitHub. And then we can use tools like gh, which allow you to connect to GitHub's API pull that information back into Cursor and we can say, okay, cursor, like, tell me where are we with this thing? And it can say, here's the history of it, here's what's changed. You know, there's, there's gonna be a lot of really interesting things that come out of just. being able to connect this data together in tools that we have locally and have it run and operate fast enough and with a large enough context window that can actually give you a, a really good idea of the history of code and how it's evolved as well.
Darin
00:44:52.071
You bring up cursor. Is that your only or are y'all testing a bunch of different ones? Um.
Anthony
00:44:57.168
I think right now we, a lot of us have settled on Cursor because it's, it's kind of easy and it allows the use of different models, and we have the ability to put some com like controls on it that are applied across the organization. We've also experimented with Claude. We have some team members using Claude. They like that interface a little bit better, and they find that it does some things better. the biggest challenge with Claude is that their enterprise or their business pricing is very, very pricey, like $200 a seat. per month it's expensive, whereas cursor's far more, I say cost effective and does pretty much, it's evolving just as well. And you get the whole IDE right? Rather than this sort of command line interface For now, we're letting team members use both to see where it goes. but it took us a while to get here because we, we were very risk averse. And so we, first wrote the policies about what team members are allowed to do, make sure that we had the right configuration. So for example, they can't train. Models in any way. Like we make sure that's all shut off before we get into, we don't send customer data into a model where it could potentially be trained on it. Like these are the types of things that we had to think about before. So we spent a good, I'd say six months to a year just thinking through the risks that we had, both for our customers and our own business before we even said okay. And we ran some trials as well from team members who were really into the technology to make sure that we, we tried to avoid. Pointing gun at foot and pulling trigger, before we started deploying it more widely with the team members.
Darin
00:46:25.413
Well, even back then, say six months ago, for how much has changed in six months, again, if we look back at 2025, I, I went offline in February. Came back in June, that's when the Ag agentic stuff came into play, which was a completely different thing. now it's just, you know what, what's next? Things are moving. Can you remember, 'cause you've been around long enough, 'cause you talked about the nineties.
Darin
00:46:51.461
Can you remember a time and you went through the dot bomb, which you remember that, this does not look the same as dot bomb to me.
Darin
00:47:00.956
I mean the the height. Well, okay. You know, I'm, I'm hesitant to say anything about a bubble because it doesn't, it doesn't feel the same as the sock puppet, let's just put it that way. it's moving so much faster. The hockey stick has been so much compared to basically just Web 2.0.
Anthony
00:47:15.708
Yeah, let's take it into two pieces, right? There's the technology itself, which is fascinating, and it's just, it's interesting. It's evolving quickly. there's a lot of capital being put into it. So that's one side of it. And then there's the business side of it, right? And the business side of it maybe smells a little bit kind of like, what's what happened in the dot bomb? or.com bubble in the sense that companies aren't making anywhere near as much money as they're investing into this. And the level of investment is so much higher, than we had it before. There's so much more capital. Uh, there's some really good podcasts that are talking about this right now. I think from the technology side, we've definitely made some, interesting leaps. I also think we're not anywhere near done with it. there's a lot of researchers who feel that the generative AI path is maybe one direction, but that it's not actually gonna get us to, true ai, which is there's people that really want to get there, so they're taking different research paths. So I think we're, really at the early phases still of this but look at how much it's already, like, it's change. It's, it's changing a bunch of. Ways that people work even now in its infancy. So I can imagine that in a few years it's gonna still look dramatically different.
Viktor
00:48:28.055
I mean, this time last year you probably only heard about, Hey, somebody mentioned something inside. I don't know what, what that is. That that was a year ago, man.
Darin
00:48:40.160
So with all these AI startups, we were talking about the business side of things. You've been running your own business now for 15 years. what reality check do you need to tell all of these new AI founders that think they're just gonna cash out, uh, and, and be happy in the next year?
Anthony
00:48:54.849
the reality is, is that most businesses fail. that's the reality of business. The vast majority of businesses in any type of business will fail. there's not a lot of moat for a lot of these businesses because essentially they're repurposing models and interactors that are provided by somebody else. the twist that they put on it it won't exist as a moat. It'll just be a feature essentially for something else. That's actually one of the things that I. am Still looking to, I don't think the shoe has dropped in terms of the real revenue generators that are gonna come out of things like Perplexity and ChatGPT because none of them have really started touching what happens when they get into the buying process. I know. Chat, GPT is starting to, they started experimenting with Instacart, like, okay, can we do this? But apparently it's pretty awful right now. Well. The ones that figure it out and they, they become where you are working with an agent or working with, querying a model and the buying process for whatever's you're buying happens sort of automatically in that flow. That's gonna change a lot for a lot of businesses. And it can be good. It can help. Consumers who want things to be able to get the actual things they want instead of the stuff they don't want, or vice versa, can put a massive amount of power in a very small group of people's hands and now all of a sudden they can move markets just by showing or not showing some product. And that's already, we're starting to see that, that was around your first question. You said, how does this affect us as a business? There is a very, possible case of. You know, some business pays a lot of money to shut everybody else out. In particular concepts. We're not even talking about search queries, we're talking about entire concepts, and that could consolidate power into a very small number of businesses even more than we have today.
Viktor
00:50:45.571
you mentioned mos, right? And I, I assume that you're, you were referring more to those. Startups, businesses that are just wrapping you, check GPT, things like that. But on the other hand, I feel that the moats are bigger than they ever were in terms that, I mean, I dare somebody to, to come up with the idea to create. To compete, uh, on model level, for example. Right? Because, and when I say mode now, I don't mean technological mode, but kind of Yeah, you just need a trillion dollars. Kind of like good luck,
Anthony
00:51:22.098
Yeah. The for, yeah, for the biggest players, the moats are already in place and they're just getting wider and wider. What I was referring to is more the moats that a lot of the smaller startups who are going into the AI space, and they may be trying to tackle, uh, a vertical. So the moat can't be ai. For them, it still has to be the vertical. So in other words, if I'm gonna start a business, and my concept is I want to take AI and mix it with auto repair, I better be really, really good at auto repair and I better have data and knowledge and experience with auto repair, and I need to then figure out how to, not ai, like the AI part of it, somebody else is solving that, right? You're gonna be using their tooling.
Viktor
00:52:05.431
I'm not even sure about that, to be honest. I feel that the situation is similar, like coming up with sus and going to reinvent every year, praying to whichever date you pray, that actually that's not gonna be killed during the keynote. Right? 'cause those, like, you know, recently Cursor released their own model. And I have it so clear, kind of like you are on the real threat, kind of like cursor can disappear like this, right? if a tropic decides just kind of like, Hey, my model's kind of like my id,
Anthony
00:52:44.789
the risk of having a big player come and eat Your lunch has always existed though, right? I mean, and you're right. With SaaS it's been something where. we are a perfect example of that. when DN Simple first started, none of the cloud providers were doing DNS. Then Amazon comes out with Route 53 and people were asking me, aren't you afraid? Like they're just gonna destroy you? And my answer was, well no, because this is just one of 500 other products that they're maintaining and, and building and growing. It will be good because it's Amazon. They're good at building distributed systems with AWS, but it will also be finite in the amount of investment that they're gonna put in it. And it's going to be kind of generalized and generic. It's not gonna be. Great for one particular audience or another particular audience because it's Amazon. They need stuff that is broad. And so it allows a niche provider like us just to have a focus and to evolve our focus because we can continuously move faster we're kind of like the little mouse that's running around the elephant's leg, We're staying alive just by, keeping on the move. And so I think if you're gonna be a small business, regardless of who your competitors are. You have to be agile and nimble and, keep moving. And again, when I look at some of the things that the tooling now is allowing us to do when combined with good processes that don't get in the way, I think that allows us to continue moving a, in an agile fashion with the team that we have.
Anthony
00:54:11.822
our biggest thing today really has been we have, a couple different types of customers that we see that are our sort of ideal customer profiles, we are really trying to double down on talking with them and seeing what they need, like what are the areas that they need us to focus on for them. And so in the next. Three to six months. Our goal is essentially to take all this feedback we've gotten. Uh, I was interviewing customers over the last couple months and sort of figuring out, what we could do with that and using that to help those niches operate their DNS better. we'll keep doing the things that we've been doing. Most of our customers just want us to operate in a stable and consistent fashion. So we continue doing that while also trying to add on, uh, new things that they want. So it might be new record types that they wanna see supported. It might be new ways of organizing domains inside of DN. Simple. There's a lot of different things that people are asking for.
Darin
00:55:06.678
That, that, that's just amazing to me. I, I, this conversation has taken me all sorts of different directions. We can find DN simple@dnsimple.com and all of Anthony's contact information will be down in the episode description. Anthony, thanks for being with us today.