Giving Agents Computers — Ivan Burazin, Daytona

you found like, we did you reach out or for Shift? Or I reached out to you. The reason was you we were just we were thinking about I was one of the cofounders of CodeAnywhere, the first browser based ID.

And so we're thinking a long time of, like, local hosts should die. Oh, yeah. Yeah.

And you had this article. And the local hosts. And then I reached out to you because of that.

And then we talked. And I was actually at a different job in learning about I was a head of, like, developer experience, and you were quite well versed in that. And I actually reached out to you among other people, like, how do we how do we go about that?

What are the key things and whatnot at this point in time? And you were nice enough to take the call. And I remember I was late on your call with you.

I don't remember. I remember because I was with my then I was thinking of a girlfriend or wife at that point in time. I'm not sure.

It's the same person, so that's great.

Sure. Like, really way back, me and my cofounder have been together. I've said this multiple times.

It's like we were married and divorced and married. Some people actually asked me, is is my cofounder my partner? Like, thought it literally.

It's not literally. But we have done multiple companies together. And to your point, we had this shift where we went from the CodeAnywhere to the conference called Shift and then back to Daytona.

We originally started stacking stacking servers, doing, like, virtualization in the early two thousands and, you know, routers and doing basically all these things at a foundational level. And that was a services company, which we sold to focus on what my cofounder actually invented, which was the very first browser based IDE. Right?

I say the first before us was actually Heroku. They did it for a very, very short time until they became Heroku. But outside of them, we were the only one, and it was called Cloud Nine?

There was Cloud Nine that came out slightly after us. There was Replit, which came out. When we stopped doing it, Replit came out, and they have been successful since then, which is great.

There was Nitrous IO. There was quite a few that existed in time, but it was, like, too early. But the interesting part is that we, at that point in time, because there was no Versus Code, for those that still remember Versus Code, for for there was no Kubernetes, and Docker had just started when we or not sure if it was even public at that point in time.

And so we had to build everything to the whole stack ourselves. And that was the key learning that we brought into and that we've been using in Daytona today. So it was super early.

There's about 3,000,000 people used CodeAnywhere. It was slightly it was angel back from other venture backed. We ended up paying everyone back because it didn't have that sort of scale.

But, you know, three years ago, we started something similar with Daytona, which is not what we were what we are today, but it was automating dev environments for human engineers. The basically, the underlining stack of CodeAnywhere. And then we who did a hard pivot last January to sandboxes.

And so here we are. Historic pivot.

but also in E2B, and then both of you pivoted into the same thing. I'm like, fuck. You invested invested in Daytona.

You invested in Daytona. But you were the first if we had not got your check, we wouldn't have done it. No way.

No. It was like, we have to get him on board first, and you were that kicker that we that got us on the No. Because you you were putting me on your pitch deck, man.

I was like, man, this is like a good trip if I don't invest.

It it runs AI code. Like, it cannot be simpler. Exactly.

But we're gonna probably have to change that because it's also a subset of what we do. Unfortunately, we really love this. Run AI code is super simple.

People interpret it different ways. I think we've given out five, six thousand of these shirts. People wear them with pride because it doesn't really market to about Yeah.

It's gonna market back to you. Back. It markets to the person about the person itself.

So I think we did a really good job on that one. But it is also subset of what we do because people, when they think about running AI code, they just think about these small, let's call it isolates, code execution boxes that, you know, you send some code, you get an output. Whereas what Daytona is today is essentially composable computers for AI agents.

It is the market calls them sandboxes Yeah. Which can be misleading. All these things on the Yeah.

Exactly. Because it it can be misleading because people usually think about sandboxes as a demo or a test environment versus a production grade environment. But what Daytona does, if you think of the laptop that you have in front of you or the computer that's over there or, you know, my wife is an architect, so she has like a Windows with three d graph graphics card inside to do three d rendering.

Like, computers or different compositions of computers. And our belief is strongly that agents today and going forward will need all these different compositions of computers to do different types of tasks. And so we offer that basically through an API.

Yeah.

when we had basically did a demo with I don't I think we discussed this as well. Devin was not public. You actually gave me access to Devin at that time.

So Devin I did? Yeah. You made I don't think I was supposed to.

Yeah. Exactly. So it doesn't matter.

Yeah. I I gave, like, three three friends access. Yeah.

Or it was a call call and you showed it to me. Doesn't matter. Yeah.

But OpenDevon was available, which is now called OpenHands. And so we're like, oh, this seems to be a thing. This is not public.

Let's take our for human automation of dev environments and take OpenDevon and launch that as a SaaS. And we did that. Not very many people signed up and used it, but a lot of people reached out that were building agents.

And they're like, hey, my agent needs a compute sandbox runtime, whatever you wanna call it at. I forgot what it was called at that point. And then we were like, oh, amazing.

This is a new market. Here is our infrastructure. Here's our product and go.

And what we found really, really fast soon was that people did not like what we had built. It didn't work. And I remember talking to people at the beginning when we're doing this, you know, the sandbox we're building for Asians.

People are like, oh, why is it different? It's the same thing. We have, like, EC two.

We have VMs. We have all these things. But we saw that everyone we gave it to was, twenty, thirty people.

They all said, no. Like, this is not what we need. This this sort of breaks.

And, basically, me and my cofounder not knowing a lot about because we're infra people. We're not AI people. So I basically took upon myself to, like, watch every single podcast that exists, including all of all of these and all that.

Sort of get up to date, read all the blogs, like, get understand what's going on. Do do you wanna shout out who who else was useful? Just just in case people are also looking.

So generally, we I looked at there there's a few of podcast different segments and different types. So there's you guys, no priors, Bill Gurley's was great. PG sir.

Yeah. While while it was around. So there's a few 20 VCs interesting from a different dynamic and some are different dynamics.

But there was But it was already about the compute market. It was already sorry? I guess you're you want you're looking at the agent infra market.

I was looking at the agent market and the AI market in general and sort of understanding who are the players, what the perception, and how that goes. And, like, obviously, you complement this with, like, going to conferences, going to events, going to meetups, reading white papers, like, doing all the things that you have to do to understand what's happening. And so when we figured when we sort of had an idea of what we had to build literally over the New Year's Eve, literally on New Year's Eve, I half vibe coded the first MC first minimal viable product of what Daytona is today.

And I went to sleep at, 3AM or something like that. I was doing I just put my, like, baby daughter and wife to sleep and, you know, happy New Years and go back just doing this. And I sent it to my cofounder, my CTO, and he saw it in the morning.

He's like, this is absolute garbage. I'm not sure this to anybody at all. But the idea is good.

And so he took two weeks and he Did it, like, look like that? Listen. It was rough.

Not yet. Not Oh, okay. It was my it was way worse.

But it was like a very it was a simplistic view of what it should be. Like, it worked, but it was not ideal. And so he went we went down the hole, which is his job as CTO, to go and he came back with this version.

We then called all the people that had said, like, this is garbage, you know, a quarter ago. And we set up these calls and we gave it to we just demoed it to everyone. And all the calls went long, every single one.

They were fifteen minute calls, and they all went like twenty five, thirty minutes or whatnot. And everyone said, we need we want access. There was no logging, just an API key because this was just a beta or an alpha.

And they said, oh, we want access. And we're like, sure. Yeah.

Okay. Thank you very much. But after, like, the next day, if we'd not send it, every single one, like, every call that we did, everyone came back, where is my IPI key?

Like, everyone wanted it. We're like, shit. Like, this is it.

Like, I've never felt so one, the the understanding to your point was like, most people thought it was the same infrastructure for humans and agents. We understood a quarter ago, it's not. We just didn't know what was the right primitive.

And then when we came, and we can talk about what that is, and we gave it to these people. I've never experienced I've done multiple companies in my life. I've never experienced this that people literally call you if you do not give them access.

Like, they want access right now. And so it's like, okay. They don't want this.

The thing that they want doesn't seem to exist or they have not found it. And they really, really want what what we want. And then when we understood that we're onto something, and then when you think about the size of the market, like the market for human engineers and enterprise is a very large market.

So think GitLab or or whatnot. But the market for every single agent that will exist ever in the future is just like, what is that market? How big is that?

And we're like, we are all in on this. And so that is where we made sort of the cut between the old products and the new one. Yeah.

But it wasn't composable at the time. So it was very it was basically just a Linux box that you could change that that you could define number of CPUs, disk, and RAM. Like, that that is what you could do.

But you couldn't have multiple operating systems. You couldn't resize it on the fly. You couldn't add a GPU.

You couldn't do, like, all things. It's just the just the first sort of variation of that. Yeah.

And was it bare metal from the start? It was bare metal from the start. And so the interesting thing that we thought about right away, so our Which, you know, give people the background.

What is the normal normal path? Yeah. So basically, most most providers run this on top of v VMs.

Yeah. And also Firecracker. Yeah.

They run a Firecracker on VMs. We also we can get we have multiple isolation layers, and we can do that. But the the common way to do it is that they one, that the state of the machine or the the the hard disk is not part of the sandbox itself.

And the other thing is they're not meant to last forever. So most of them are preemptible. Like, they can there there's a time that they can live.

And so our thought was when we're going into this is agents will be like humans in the sense of you don't want your laptop to be shut down until you're done with work, like and you want to close the lid and open the lid, but it's the same state. So you agents would want that, like, pause and come back. They want those two things.

But also agents really, really want speed. Right? Can they get it?

So when we thought about it's like we need something insanely fast, how to make it fast, how to make it long running and stateful. And so those two things it's like combining a lambda and an EC two. Right?

Those two things together. And so we didn't have an idea how others did it because we didn't know too much that that there was a market around this. It was more like, okay, this is what we need what they need.

And we looked at Kubernetes. It wasn't wasn't good enough for that. We look at Nomad.

It didn't enable that. And so our history and rewriting our own scheduler at CodeAnywhere is basically what my CTO came up with. Like, he's like, oh, the learnings from there, he brought it.

And the funny thing is our third cofounder, when he saw it, he's like, dude, what is this? This is like 2008. Like, we went back in time.

And he's like, exactly. And so the reason why Daytona is, like, super, super fast, and you see this on benchmarks, is we essentially we run on bare metal. We have our own scheduler.

We use the underlining disk CPU and RAM of the underlining machine, which means your IOPS are insanely fast because there's no, you know, there's no network between an EBS or something like that. But also the snapshot, the point in time, the templates are also preloaded on the bare metal machines. So when you fire off a sandbox from a template or a snapshot, you you're essentially directed to the bare metal machine where that snapshot is based on that NVMe drive, And then it literally just turns on that machine and it's local.

There's no network latency anything on there. And so that is sort of the specificities that we when we're thinking from first principles what a computer would look like for an agent, that is what we came up with, and that's what we created. Yeah.

I should maybe I don't know if you endorse this, but there's someone that does compute SDK. You guys do very well on there with, like, the TTI. Right?

I I guess. Is this a is this a is this a relevant benchmark for you guys? I don't know.

And it changes every day. So today Yeah. Arc Hill is Well, I don't know why Arc Hill has never heard of it.

Yeah. Arc yeah. So it is there.

But you are, you know, at least a third of the the next tier of performance, and then, you know, there's there's a lot of other better known names that are very slow to say. Yeah. We've been the number one by far for a long time, and now there's different there's different definitions also of sandboxes, different isolation patterns, different other things.

So Archel runs it literally on the s three. The data was very different, and they spin up a sandbox spin up a container for that. So it's a different type of thing.

Yeah. The definition of a sandbox is something that we can all Yeah. All need to get along with.

But, yeah, we're insanely fast on getting these things up and running. And so you can see even there that it's 0.1 to 0.

11. Like Close enough. Yeah.

Yeah. Yeah. I mean, what else do you need?

Right? So the benchmarks itself so in this in I don't think the benchmarks equate to market ownership or revenue or anything like that. And I've seen this with multiple benchmarks, not just in sandboxes, but in general benchmarks around.

It's table stakes. It's just like the But it doesn't reflect you definitely have to be up there and you have to be competing so that people know that, oh, this is definitely one of the top. Because this is only one dimension of what customers look for.

There's other things that, like, how many can you spit up consecutively. There's a feature set. There's support.

There's, like, all different things that people look at, but you definitely have to be there on the benchmarks. How how many people do people spit up consecutively? So we have concurrently, I guess, is the concurrency area.

So there there's there's three metrics that we look at. And so one is, like, time to spin up one. And so our time to spin up one is sixty milliseconds with network latency.

So request spin up reply, sixty the whole thing is sixty milliseconds. That is one. But if you wanna spin up 50,000 at once, we are now at about seventy five seconds.

So think about seventy five seconds to spin up concurrently 50,000. Some others, there's public data around this, like, take two thousand seconds, which is thirty minutes. Like, there's different variations of that.

And then there is the so it is speed of one, speed of, like, multiple, and then how many can you consistently have up and running. And so we basically have right now no limits to how much we can add because we basically own our own metal. But the biggest customer of ours does, like, about 850,000 every single day is sort of where they where they're just shy of a million every single day that they're running.

We do have a request for half a million concurrent, which is literally half a million CPUs somewhere running.

The the most expensive thing are CPU. Okay. So the second one is yeah.

Then it's RAM, then it's disk. We actually don't charge Which is snapshotting. Right?

So no. And, you know, it's actually the it's part of it, but basically the size of your hard disk of your machine. So do you have 10 gigabytes?

Do you have 20? Do you have 50? Do you have whatever?

And then the the transference of that. Right now, currently, we don't charge for network at all at the Oh, yeah. Yeah.

You gotta fix that. Yeah. It is very much a it's a larger and larger part of our bill.

So we're working around that part there. Obviously, that is the the least expensive. Yes.

So the hard disk is the least expensive. So it's basically CPU, RAM for us network, because we don't charge the customer, and then hard disk is how it's spun up. But there's also different types of workloads.

So we basically split it up into two types of workloads in Daytona. One is what we call background agents or long running agents. Okay.

And the other is basically RLs and evals, which I put sort of together. And so they have very different patterns of usage. And if you look at the usage of a background and I'll just name names of companies, not specifically.

Open All Hands. Yeah. Yeah.

So like a background agents, a Cognition, a Lovable, like all these things are Harvey. These are all long running background agents. And so if you look at their usage patterns, their usage patterns are similar to human, is like follow the sun.

Basically, the usage patterns of that is like noon is probably the highest and midnight is the lowest, and then weekends are lower, you know, weekdays are Yeah. That's a fun question. How global is it?

You know? Is it very US centric or So The US is a large part, but we have currently, we have Asia, Europe, and Europe, and The US It's quite global. Yeah.

It's quite global. We have it all. It's interesting that our number I talked to you about this.

Yeah. And it's it's like keeps showing up. No.

It's quite interesting. We were quite shocked. I was like, oh, this is interesting.

Yeah. Yeah. And also one that's There's a reason I'm doing AIU Singapore.

I mean Exactly. Because I'm from We're there. We're gonna we're gonna be there.

Yeah. And it's interesting that Japan is in the top or, like, Tokyo is in the top, which is in all the tech cycles, it has never been. Yeah.

It has never been. So it's quite interesting that I think the Japanese just love AI. Yeah.

Yeah. It's it's that, and then it's Brazil. Yeah.

Brazil has always been in the even when I look if you look at like GitHub's data and Us historically with Goat Anywhere, it was always like US, Western Europe, and then you'd be have like India, Brazil, China, like that would be there.

all time, you know? Yeah. So the interesting thing is like we have those kinds of loads, but if you look at the researcher loads, they're quite different.

So what they are is like, if you give them concurrency of 10,000 or 50,000 or a 100,000 CPUs, whatever it may be, when they fire off a a run, it's just a 100%. And then it just runs, runs, runs, and then it stops. It's very the the usage pattern is squares, basically.

Right? And it's also not follow the sun because people will fire it off at midnight before they go to sleep, but then wake up. And so so it's very unpredictable, you don't know where that is.

So the shapes of the usage are quite different than we have had before. And also what's interesting is when it's sort of a fall of the sun, even if you have a high growth company, you can sort of predict your usage patterns and and have enough capacity for that because it's sort of it it grows in a in a way you can project. When you have companies doing sort of like evals and RL, they're super spiky.

So they're gonna come in. It's like, we're gonna use nothing, then can we have a 100,000? Right?

And then go back down. And then have thousand get back down? So it's very, very different.

Right? And so So do you want to lock them into commits? So that Yeah.

Yeah. We do. We so we have to lock them into some sort of commits to have that capacity because we have to have basically, we have to have the capacity for peak.

Yeah. Yeah. Right?

And so right now, Daytona's mean utilization is 15%, one five. Oh my God. So it's very low.

Because it's very spiky. But it's very spiky, but we get up to 90%. Yeah.

Yeah. So we have these things. And so what we're looking at right now as a company is similar to Cloudflare where you can like geo move things around.

But that works really well for basically the background agent where it's follow the sun. But this, it's not. Like, it's a very different shape.

Obviously, with scale, you figure these things out, but that's an interesting new problem that we have as a compute provider in the agent space. Yeah. And when we were doing the conference recently, and so we talked to, like, Nikita from Neon and the I should bring it up.

Parag from Parallel and whatnot. Everyone has the same problem, whereas the usage is super spiky. And this is something that has not happened before that you have these types of like, it was always it would the amplitudes were not this high.

Right? So it's quite interesting use case and problem solved. Yeah.

I don't know if we're gonna bring this up again, but let's just talk about the conference.

people that are building infrastructure for AI agents. Because when I think of what we're building, it is the agent is the primary user, what are the ergonomics and usage patterns of agents and so can do that. And what I found, this was a a theory wasn't proven, is that we all have these problems as I touched on to.

And I was as I was talking on stage, it was like, we all have the same underlining infra problems, which is this spiky workloads, unpredictable workloads that we've never had before in human compute or human infrastructure. And it's again, it's the same when I was talking to Parag or when I was talking to Lin, I guess. Nikita.

All that. It's Lin, especially, I was talking to her the other day as well. Like, the the it is a very interesting type of problem solve because I can touch on Cloudflare because there's a lot of, like, talk about that recently as how they solve that, which is they have a bunch of geos.

And basically, as users work in different places and depending on your tier, they can move you around the geos. Yep. And so that that's how they get to the higher utilization.

But you can sort of predict these. And if it's something and you'll rarely get a spike that is 10 orders of magnitude. Like, you'll get a spike.

Like, let's say, one of your customers has some ex like an exponential curve. What is that to to I'm make I'm using Cloud Verse as example. 10%, 20%, whatever it means.

I don't I don't have this data. I'm just assessing. It's truly not 10 x.

Right? It's truly not something there. And so how do you go out and solve this problem?

And we're all solving this in different ways. So you know So she also has the same thing. So yeah.

Yeah. I know specifically that, like, Neon had that issue as well. Like, how are we solving these spiky loads and things like that?

We can put it down in the screen.

yes. But they still have CPU and RAM allocation that you have to have Oh, yeah. And running.

As a CPU and RAM, you have to allocate that and have that ready. And so there's basically two ways to do it. One is you either over provision and you can handle the bursts.

Or two, you basically have I don't know if this is a term, just in time compute, which is like as your usage comes in, you can fire off requests for VMs or bare metals at other cloud providers and then get them So up and this is if you go above a 100%. Right? Yeah.

So overflow If you overflow like spillage or whatever Yeah. You you probably lose money on it, but doesn't matter. Right?

Not well, you might you might not that is a more cost effective way to do it, but it's a slower way to do it. Yeah. Because basically what you have to do is you have to like queue your requests, spin up these just in time compute, get it all ready, provision it, and then get your workload there.

And so if the time isn't important that much, that's fine, and you can do that. But if your customer and especially for, let's say, the RL training runs, the the reason why a lot of people come to us is because GPUs are more expensive than CPUs. Right?

So you want your GPU running at, what, a 100% the entire time. And so when you're running runs on CPUs, when the c when the CPU cycle is, like, down and spinning up the next one, you want that to be instantaneous so that your GPU doesn't go down. Right?

Yeah. And if you then have to, like, go out and provision machines, you're essentially telling the GPU that it has to wait, and that's incurring our costs. So there's things that you have to try to solve further.

Yeah. Let's talk about the different workload. Right?

now it's 50%? It will be this month, 50%. Yeah.

Let's talk about how different it is. Right? Like, I imagine, for example, a lot less dynamic code generation of, like, arbitrary code.

Like, here is probably all the same code, you're just doing parallel runs or something. Don't know. Yeah.

So you'll have multiple depends on the like, for each run, you'll have a snapshot. And for the most part, they actually do use our declarative image builder, which is like, oh, the agent wants these dependencies, these end bars. Yeah.

Yeah. The declarative image builder. Which is a very model like thing.

Yeah. So we build it on the fly, and then we propagate that snapshot, and then you can spin up as many sandboxes as you want against that snapshot. And then if you have to do changes, the model can or like, it could be also be automated.

It's like, oh, now for the next run, we need to install these things or remove these things or whenever to get a a task done. And then it goes off and and runs that. So, yes, that is something that it seems that they prefer.

The number one reason I found or should I say, let's take a step back. What we are competing against in that environment is essentially managed Kubernetes. Yep.

So EKS, GKS, whatever. That is what the vast majority run on. And anyone that has tried Daytona versus GKS, EKS is like, I'm never going back.

Mhmm. That has always been there's a few reasons. One is the ergonomics.

So if you have if you're using Kubernetes to spin that up, you have to essentially manage the interface interactions with that. Daytona, although it's a compute provider, it's more akin to a Twilio and Stripe from a consumption perspective than it is in AWS. Like, you have an API and SDK, it's quite, like, easy and seamless to get these things up and running.

That's one. The other is the speed to which we spin up, which we mentioned earlier, which is much, much faster, and the scale to which we can go to. We haven't got into features, but an interesting feature is that it's very hard to OOM or out of memory our sandboxes because we can dynamically on the fly Resize.

Resize, which is like impossible on almost any other thing. There are some technologies that enable you to that, but it's like a very hard thing. And so we actually saw this when the Terminal Bench team is brought us actually so thank you, Alex and the team.

They brought us into this whole space.

you know, a framework would just say, guys, just use Daytona. Yeah. I think it says it similar.

Yeah. Yeah. I was like, what is this?

There's all there's multiple there, but they also mentioned a few other places. Yeah. And so Daytona specifically, we have just jumping on themes here.

We do not pay them for this. Just say no. They just like you.

Yeah. They like us. Yeah.

And also a thing so Daytona has multiple isolation sets underneath. The customer doesn't have to know what they are. But basically, we have Docker, which is a container that's hardened with SysBox.

So it's Docker's isolation that is a security equivalent to a VM, but it's still a container. And that is the default. And they, especially in these training workloads, really like that as an interface to be able to use just a basic Docker container.

And we enable Docker and Docker, which for these RL runs, if you need to do a Docker Compose or Kubernetes, you can spin up a k three s inside of these things, which unlocks a huge amount of workloads they can do that you cannot do on other providers. So just on on that part is much more interesting. And so we went that through that.

We showed them that we could do that, and they enjoyed that quite a bit. They being the Yeah. They're not people.

Harbor people. The Harbor people. Do you know are they are they a company yet?

Do not know. Okay. Yeah.

Sure. There's a couple of things that we learned, which we chat about in the beginning, which and this has led our story, as we mentioned, we, like, talked to a lot of customers along the way, and we add more features and more tool sets as we talk to customers. And I think it's it's that the ecosystem is so small and or the the models get smarter where when we see one user come with a request, we know it goes on a road map if, like, three to five customers come with the same request in that week.

It's, like, very bizarre. It happens so many times, which is like Because they're all friends. They're Sorry?

They're they're all friends. They're on the same group chat. Yeah.

They probably. Yeah. Because and they're like, oh, can you do this?

I'm like, okay. This is interesting. We'll put it on a feature request.

And then the next one's like, oh, can you do this? Okay. It's all the same.

Right? It's all the same. And so what we try to do, and I personally try to do, I try to be on as many quote unquote sales calls I can.

I'm in every Slack channel. We literally have about a thousand Slack connect channels, something like that. It's an interesting there's so many interesting things you find out when you have all Slack channels.

You can also see where people transfer between companies. You see leave Slack channel and leave Slack channel. It's an interesting thing.

Also, just I digress. I feel that Slack connect is literally LinkedIn what it should be. Yeah.

You have a list LinkedIn charges you to to, you know, use your own connections, but Slack doesn't. Right? Slack is like, you do it for free.

It's the more lock in is great. Yeah. It's it's amazing.

It's why You're gonna pay Slack for life. Exactly. You're there for life.

So that's interesting. And so one of the things the the newer things we were talking about earlier is we made a big bet in and put a lot of investments on computer use. That is not seen publicly the light of day.

We haven't GA ed that yet. But we have Is there thing I can pull up? There is computer use there.

It's right up a bit. Oh, yeah. Okay.

Yeah. Cool. And what we have what we talked about and what we've seen publicly is there's this theme now about, like, the human emulator where and Elon from x AI has talked about this publicly.

And if you think about the models today, they're actually quite sophisticated and they can do a lot of work, but they still don't have access to all the tools. Like, I'm a strong believer that the most efficient way for an agent to work is essentially headless or through, you know, terminal or or whatnot. But if we if we look at knowledge work in general, there's about a 100,000,000 knowledge workers in The US, about a billion in the world, and knowledge workers.

And the salaries of them aggregate to 10,000,000,000,000 in The US, 50,000,000,000,000 worldwide. Something like that. If we look at the five most important sectors of that, so like healthcare and government and financial services and whatnot, that's about 56% of that.

So let's say it's about half of that. So in The US, it's about 25,000,000,000,000. And most of them, most of that work is actually still locked into legacy apps inside of Windows, which is not going anywhere for a very, very long time.

Like, people just won't invest in that. How much of it? Our assumption is the following.

If, like, in the RPA market, which is similar market, but not the same, 25% of, like, these white collar workers' work is automated. If an agent is more sophisticated, can go through more runs, figure stuff out, let's say it's like 40. Right?

And so if you take 40% of that, you get to essentially, like, $10,000,000,000,000 Yeah. A year. That's a TAM.

That is the TAM. That that that is the TAM. So that's the TAM of the models.

Right? That's not our essentially ours, but you get to that size. And to be able to do that, you essentially have to give agents these computers with the legacy.

So computer use either Mac or Windows or Linux. Linux, we also obviously have and others have, but Windows specifically is something very, very new. And the only option right now is an EC two with Windows or an on on Azure.

Both of them take anywhere from three to five minutes to spin up. We've created an actual sandbox, so it's a second instead of milliseconds. But you have, like, point in time snapshots.

You have, like, forking. You have all the things that you have from a sandbox, but essentially enables you to hopefully unlock all this value. And so that's that's been our big push and bet what we've sort of like kept our ear to the ground.

What is sort of the next things in the market? Yeah.

deck Yeah. Recently, last month, whatever. And I'm like, okay.

Let's just let's just do automated. So, like, all our data's in, like, Clickhouse and HostHog and QuickBooks, like, where everyone else's is. And I'm basically, like, connected that all to, like, my Cloud Code, like, off and go Cloud Code or whatever.

Go off and, like, here's integrations. Go do that. It pulled out the first report, which was great.

It connects to Brex and all these things, pull out which was great. And I say, okay. Now pull out this, this, this, this.

And I kept getting, like, really well McKinsey style design reports, but the data said partial data. Mhmm. All the missing data, partial data.

Like, it can't access all the things. And I got so frustrated. And so I got I got, you know, my Mac mini virtual sandbox with OpenClaw.

I gave it its own account in our company. And then I went to all these services and created a read only account. So literally like an intern in your company.

And so I would say, now go and do this report, and we'll get the same. Or like, I can't via the MCB or the API or whatever. I can't get all the information.

Go log in. Yeah. Yeah.

And it will log into the website, then go in, export the data, it'll export the data and do the thing end to end. So even for things that have today APIs, not all of it is exposed. And I to get value, like I get immense value right now, but it has to be a computer usage, unfortunately.

And so I spend a bunch of tokens just on that, but I get the job done. And so if even a startup like ours and using all the hottest tools still needs a computer agent, what hope does, you know, Goldman have to have a headless. Right?

Yeah. Yeah. Why isn't Microsoft doing this?

Well, I'm pretty sure, like, Satya had a post yesterday. Oh, was like, agent needs a computer. I see.

I see. So they have launched something. Yeah.

They have Microsoft Power Automate.

sandboxes. So we will have macOS sandboxes fairly soon. The problem with macOS OS sandboxes is I'm deep in this.

I don't know how much interesting is this. MacOS has this problem. It's a licensing thing.

Licensing thing. Yeah. So one, you're allowed to run only two parallel VMs per machine.

So that's one. Two, you can only license to a different user every twenty four hours. So if you come in, and theoretically, if I wanna charge you per second, I charge you one second, I have to have it idle for the rest of the day.

Like, I can't have anyone else doing that. So the pricing will be different in the sense that we would have to charge for twenty four hours. And that's not even that's not even the most difficult thing.

But the thing above that is from a security perspective, they enable you to do memory snapshot, pause, resume, but only on the same physical drive physical machine. And so what you can do in, like, Windows world or Linux world is that I can move in the background your snapshot from one to the other and and manage load. Right?

Here, if you wanna do that, you essentially have to have your Yeah. Snapshot. Your your It's like a physical machine.

You you can't break it up. You can't move things around that. And all of that is that that that part is, like, from a security standpoint.

If it is written, they're like, I understand the security aspect of that, but it disables you from doing these agentic, like really scalable agentic workloads.

same same thing to macOS. Someone needs to do it. Someone someone will do that.

Someone will have some long running agents for a few days to figure this stuff out. But yeah. So definitely, we we're really close to offering something because people do want it, but the pricing will be different and the feature set will be sort of stringent.

Yeah. Nobody's gonna use this. I mean, like, the the labs the labs will because they want to They have to automate do URL.

Have to do URL again. But the point is with the RL part, if you if you do RL on macOS, then the next iteration of the model comes out, it will be able to use these tools significantly, then you actually need to run those that somewhere. So you're gonna have to have that later on.

although it's completely PLG and we have individual developers that use it, most of the users that use Daytona are sort of a b to b to c. So it's either b to b or b to b to c. So, like, in the researcher world, it's b to b.

So you're selling to labs and neolabs and things like that. But on the long running agents, it's mostly from a scale revenue perspective. It's mostly b to b to c where you have a app layer agent that uses you Like a mannest.

Scales. Yeah. Yeah.

Yeah. Like a mannest lovable type type thing. Yeah.

market. So I've, like we I've done multiple things. So the CodeAnywhere part of our career path are in the calendar was very much an end user developer product.

Yeah. And so that is great. It you can get a lot of developer love.

And I feel that we do as a company have a bunch of developer love, but it's a different type where it's it's people building these things. Again, it's more akin to a Twilio because you don't really run as a person, you wouldn't run Twilio. I don't know how many people remember it was like ask your developer or billboard Yes.

And whatnot. And people really love Twilio, but they only use it inside of like, oh, I'm building this app or service for thing. And so we're very much directionally to that.

And you also know that I used to work for a competitor for Twilio, so it's kind of ingrained, I guess, in my DNA. People don't know Infobip is that big. Yeah.

It's it's like. Because they're all American. They're like, whatever's in Europe doesn't matter to me.

Yeah. But, like, is that it's the same size or bigger? No.

No. It's it's it's about half the size. Half the size?

Yeah. Half the size. It's like huge.

Yeah. Multiple billions a year. Yes.

Crazy. Exactly. These are, like, really interesting and large revenue generating, very sticky businesses.

Whereas when you're selling to the when your focus is the end developer, it is a very hard sell because they're very price sensitive, very price conscious, very, you know Yeah. Around that. And there's very hard it's very hard to scale.

Your cap is the number of people that are willing to spin up first, I wanna spin that up, and then spin up multiple of these. Whereas if you're in the enterprise one, like, we know everyone's talking about, like, how many tokens they're spending, I'm spending. Like like, a lot of companies today are like, is our company, spend as much as you can.

Like, basically, that is where we're going. And so if you think about that paradigm where you're selling to companies that say, spend as much as you can to generate, you know, productivity versus, oh, I'm a single person. I have this much budget and I'm doing this thing because it's fun or it's helping me out or whatever.

Like, it is a different it's a different go to market, I think, strategy. Yeah. There's a lot of discussion.

It's been very good for you. Yeah. But I feel like it's maybe a drop in the bucket or maybe it's huge.

I I'm just checking whether it's like these are big trends. I mean, those things you work well in our favor to your point. Yeah.

Just because But they're kinda drop in the bucket. Right? I guess I think it's like sort of all the things come together.

Yeah. So there's so many things that that impact that. To your point, like, OpenClaw wasn't huge for us, but like having the agent SDK from Anthropic, so or Cloud Code Cloud Code was very interesting.

The reason why it was interesting is that a lot of let's call them app I don't know what to call them. App layer agent companies. Essentially, they are like, oh, I can create this new app, this new agent.

All I need, I just use Cloud Code, and I throw it into a sandbox, and then I have my interface to the human to that. And so that enabled so many more companies to actually offer this, and then they would pull on CyberBox. So that was that was interesting.

And to your point, like, MCP versus the CLI. I mean, the MCP is an interface against an API, whereas the CLI is like, you can actually go do things. Like Yeah.

This is the the difference between integrations and actually running scripts or data or analysis against the thing.

and to your point, we've talked to so many people over the last year. It's like, oh, like, why do you need a sandbox? Why do you need this?

Why this? And to your point, like, oh, I need sandbox for this. I need sandbox for that.

And so, oh, I need it for every single thing. And so, basically, what I what I and it sounds like a broken record. It's like, use a laptop every single day.

Right? And you are you end of one. It's just you.

But now imagine how and by the way, the the laptop, the computer PC market. The PC market is about equal to the cloud market. Mhmm.

So it's about a 150, 180,000,000,000 a year. Mhmm. Something like that.

It's about roughly the the three cloud hyper scalars is about equal to like Apple, HP, Lenovo, whatever, all it's a little bit less, but it's sort like that. And now imagine and that's just like, so how big is the addressable market? Well, how many people are there in the world now?

What's the last name? And it's called 8,000,000,000. 8,000,000,000.

And so let's say you can have two computer. Like, you have one personal and one business, whatever, like, so it's it's it's double that. And so that's 16,000,000,000.

Right. How many agents are gonna be running in two years and ten years and a hundred years? And for every single task, they will need one of these.

And so how big is down that market is essentially quote unquote infinite.

Yeah. There's a bunch. So we the the original product that we did was open source.

Yeah. Doing that was actually very good for us. There's basically a saying of what's the saying?

Like, companies that are doing really well measure themselves against, you know, free cash flow that are kinda okay. It's EBITDA. Then, you know, it's like But we're always like GitHub Stars.

GitHub Stars. GitHub Stars. So you go all the way down to GitHub Stars.

And so our original one was GitHub Stars. That's what we talked about. We're we're at the point where we're talking about revenue.

So we're we're we've gone up the stack on that. And so we Profit started profit. Profit.

We haven't we're we're Make it. We'll get there. But basically, at that point, did Starz and GitHub and whatnot, and it was useful.

And the original variation that we did, it we split the the core into its own repo, and it was Apache two point o, so very permissive. And then we basically would bundle that on the enterprise side with a proprietary repo. So it was like OpenCore, but it didn't fill out the repository.

The repository was very clean. When we did the pivot, we didn't have time to rethink this, and we wanted to we had this open source community. It felt a shame not to do that.

And so but we still did want to add some restrictions. So in the new sandbox product, we did add a AGPL three, which is, you know, it's a kind of a shortcut way to do that where you are open source. And it is true open source in the sense of what enterprise can use it if it if it wants, but you essentially can't make a competitor without open sourcing your stuff.

It's one of like three approaches.

pure open source believers agree that this is not full open source, and Yeah. I totally respect that. That is absolutely true.

But we did leave that. And Daytona, in its essence, everything outside of what's under a feature flag today, which is like the Windows stuff, GPU stuff, whatever, it is in this open source. It is there.

So everything is there. Like, our own scheduler, everything is there. So we are I've had some competitors say, like, you guys are actually open source, open source.

Like, you you're really like, you can actually see that. And, I mean, people do like that, and it has helped a bit, but it's actually more helped in the consumption of our cloud product than actually transferring people over. The reason is you can actually you send the repository to your agent when you're integrating Daytona, just has more context.

Yeah. It's like, oh, okay. This is why this is happening.

This is why this is something You could equivalently just have docs that you can yeah. Okay.

And that has been historically. The thing that we have found in this AI transition is, as we haven't talked about, Daytona's customers are everything from, you know, the single developer, the YC startup, to people say Fortune 500, I'll say Fortune five, like the biggest companies of the world. And and big Neolabs.

You you told me about Yeah. Why are they keep them anonymous? Enormous companies.

Right? Yeah. And because the market pull is so strong, we're able to circumvent these processes.

I'm not saying we go, we pass security audits, we pass all these things, but as you know, to your mention, like, Temporal way back in the way, they in our old version of Daytona, like, took us months. And usually at the end, they would churn off because just like, oh, you're too small of a company. Like, we don't trust you Yep.

Enough. Whereas today, we've had these large companies push us, like, they would push us through. Like, usually, when you would go through procurement to become a vendor of large companies, it would take you like two, three months.

We get it done five days now. And this is not saying that maybe we're great, but it's more, I think, a sign of the market where it is today. And so when you think about that, the open source is something that we, from a go to market perspective, don't think about that much because everything that we've created right now has been PLG through the cloud product, people signing up and just pulling us inwards.

I did. A little bit. A little bit.

Yeah. Sounds funny. Because, you know, I'm I'm thinking about, like, well, okay, what would it take to replace GitHub?

So there's a lot of things. I I I have thought about this, and I I thought I've tweeted about this, and I looked at some. I've actually invested personally in some.

Is it a entire? No. Haven't.

Yeah. So I I I've met Thomas virtually Yeah. And we've talked.

So I really think that and this was my reason for that. Because we have a bunch of background long run agents, and for our time, most of them are coding agents. Like everyone was building up a competitor to Lovable or or or Devon or whatnot.

What we saw from our customers was that they were all trying to figure out how to do versioning. Everyone is doing different ways. There were some some really weird ways where people were doing that.

And the reason was that GitHub as is was an overhead. Like, it wasn't fast enough what they needed. It didn't solve the problem that they needed.

And to be fair, like, GitHub is for post your the inner loop. Right? It is is post your laptop.

Yeah. GitHub is the the point at which the auto loop starts. Exactly.

Yeah. People started using that for sandboxes, which is inner loop, which is usually, you know, it's it's on your laptop. Right?

And so that is not what it's made for. And then we had everything from people actually, the the the most interesting one is we had one customer that will literally take the entire code base inside the sandbox and every I forgot what the time sequence was. They would just dump it all into a JSON, and then push that to s three.

Yeah. And that's And make your own git. And it's it's there's not even diffs.

It's just a whole whole thing every single time. It's just every because it was super fast. And then they would go back and search and find, you know, sort of what the file was and right now and whatnot.

Because there's text file, there's JSON, like, they're they're very small, so that the the network cost is very low, and they didn't care, and they just did it that way. And I'm like, if people are doing this, that means there needs to be a new solution Yeah. To this problem.

Right? And so for me, it's quite interesting to look at who who is building these types of new things. Agent first, I think Git as is still exists in the future.

Maybe even GitHub exists, but there will be a whole new sort Exactly. Git is like the deploy artifact to kick off CICD. Yeah.

But then there's a layer before that that is like the agent collaboration layer. And so I think something needs to be said there, but on the other side, like, there's issues with another interesting thing is just like CI right now. So the amount of PRs being created is insane right now.

Right? In general. Even for you guys.

Right? Everyone's creating a bunch of PR. Like everyone.

And then all that has to go through CI. And then that's the bottleneck. Like, everyone is bottleneck.

Like, not just action like, not just actions, but like, go to any CI provider. You will not be able to if you have a high throughput of PRs, there's one company we're talking to, they do a thousand PRs a day. Which means like and they're just waiting they have just a queue on that.

Right? What? Do they use, like, BuildKite or I don't know what circle they you know, technically, your tech can be used for CI.

That's that that was the conversation. Oh, okay. That was the conversation.

Is that a serious conversation? So we'll we'll see how that goes. We've had quite a few conversations around that.

We're we are not a CI provider by any means. But what what is I mean, what's missing? Essentially, you could use a data say, Daytona sandbox in instead of whatever you use for, you know, your GitHub runners, essentially.

Yeah. Yeah.

Performances,

The thing that we found that sells people the most this is more maybe a day two thing instead of a day one thing. Sure. And we've seen this again and again.

So we have a bunch of case studies, and we have a bunch of them still coming out. They're all done by third parties, so we don't do the case studies. And it's actually interesting to watch those case.

I watched they're they're recorded. And because it's a third party, people are actually more open, and they will tell you, oh, we use this competitor. We like this competitor more or this thing or whatever.

And the the number one thing that people come back to us for is that our we have an insane responsiveness. In terms of your team? In terms of the team.

Okay. Insane responsiveness has been by far the now we can talk about, like, features and breadth of product and concurrency and CPUs and, like, all those things, but I feel that that would probably so if all other things are equal, that is very much a differentiator, I found. And I did not And is that entirely Slack or Slack plus email?

There's email there as well. There's calls, but only the vast majority is like on-site. So it's Slack.

Like, we have had customers like, hey, we have a problem. Can you get on Huddl? Like, we will get on that Huddl like in five minutes, literally.

I've done this multiple times. So Wait. Okay.

So how big are you? 25 today. How how do you do this kind of support?

This Verbs that we're insane, we don't sleep. 007. Have you heard the new thing?

You know? Yeah. We all have Slack exhaustion.

We're very, very tired. The thing that is unique I don't know unique about us, but unique I would say unique about any successful serial founder is that you're able to pull in people that you've worked with before. And so you can't do that as a first time founder, like, I couldn't have done that or not.

But of the 25 people in Daytona, I think about 13 of them, we have worked with seven years plus. Yeah. So it's like high trust, high throughput, high we know what we're signing off to do.

And especially these people worked with us when we were starting and we were actually hustling, you know, hungry for food hustling type level. And so those are the people that work with us. The now that the new segment that has come is almost everyone is sort of, you know, one one degree of separation.

So it's like someone that someone has known, and so they sort of come into this org. And we've had people that have, like, not fit into org as well. It's just like it's that type of culture where there is a high expectation of, like, being online, replying for these things.

And I do that first. You will if you ask any engineer, they're like, you never sleep, like, about me. And so then I do that as an exam I don't do it as an example.

That's just how I'm wired. My wife doesn't appreciate that. I can tell you.

My wife doesn't appreciate that. I told her about 996. She said, I wish.

It's like these these Chinese people are slacking. Yeah. So, like, that is something there.

And so I I think every company has their own culture, and that's something very, very deep, Ours and it's something that's come up again and again, and every single day we're reminded about that. And I didn't go out thinking that that is how I'm gonna build it.

and here. Like, lot of our team is in Croatia. A part of our team and a growing is here now in San Francisco.

And so I spend a lot of time away from a family, and that is hard. Like, that is a sacrifice that you have to. But going in, like, people say, like, on your deathbed, you're gonna miss some of those things.

The thing that and probably might be true. But I think that going into this, I already said, like, I know that this is gonna hurt, and everything has to hurt. By the way, I'm very much of a feeling that everything has to hurt.

Going to the gym hurts. Losing weight hurts. Like, everything has to hurt.

Right? It does. Like No pain.

No gain. It is literally but you actually have to enjoy the pain and just like, if you don't enjoy the pain, it's not for you. And so you get accustomed to that pain.

And so a lot of kids, especially, have a daughter and a son, daughters, the eldest, like, love her and do miss her when she's not here. But it's like, that's what I signed up for. And there is a plan and target of what I'm trying to achieve.

And now, hopefully, with my wife, which does support me, we can get ourselves together more so it doesn't there. But she takes a large part portion of that. And so if you have a partner on the other side that is okay with that, then you can do that.

But even if they do, you have to be okay with not being there. Right? Yeah.

This is my my vision for you, this this meme. Yeah. Yeah.

Yeah. So that's your kids in the future. Yeah.

Yeah. I think so. Yeah.

But we have to teach them that. Because Because dad, you know, built the compute sandboxes. Sandboxes.

Dad made sandboxes. Dad made sandboxes.

the startup ecosystem? Oh, the startup ecosystem. I know this was the the recent one, which is I think that and this is general, like, business.

I I feel that the it didn't come off, I think, well on Twitter. Something like this misread it, which is the market is adding premium to SaaS vendors that are reselling tokens. Yes.

And I think that's incorrect. Why? Why I think that's incorrect is that if you look at one, your pricing depends on what the price is, if it's public market or if it's private or whatever, you're saying, the person that's reading that, that the reacceleration of revenue is equal to the old revenue, which it's not.

Not even close. Because one, you had on SaaS, you had typical SaaS margins, whatever it was. Right?

Yeah. Stickiness and all these things. Now what you're doing is you are saying, here is my agent, and I have whatever the margin is.

It's way worse. Right? Yeah.

And now you're using Anthropic or, you know, or OpenAI or whatever through me, the the SaaS model. And then we as a community are saying now that is reacceleration. And so one, I think that's wrong because it the first, it's not the same.

The the makeup is not the same. The other thing is and go back to, like, what what I mentioned earlier is, like, the the CUA and how I set up OpenCLI and whatever. I don't want your agent, essentially.

Because what happens right now, we have a problem that, and this has historically been, you have data siloed in, again, Clickhouse, QuickBooks, it's all siloed. And now you're giving me an agent that'll give me the data, but it's still siloed. Right?

And so now I have to, like, take that data and then get another agent Just expose the data to my Yes. Just expose it. And one thing I have to and so I'm like, just expose everything and charge me for that.

So charge me for consumption of API. So you'll have old seed based pricing for humans. Yeah.

Yeah. Charge me for this. The number of agents will skyrocket, and essentially, you'll have more usage and charge for more if your product has value.

So, like, there's arguments some of them do have value, which have database, not database. We can get into that. But some of them really do.

And I was actually shocked that the first person to do this was Benioff.

Was a tweet, I think, three days ago, where she said, every product in Salesforce has been exposed via API. Wow. Everything.

And I'm like, now I understand why this person has built the this insane. Kudos to him. Amazing.

It's like, thank you. I don't know if you listen to me or someone else. I'm like, thank you for so much.

This is the direction of the world. And so if you can get real acceleration against that, against consumption of API, that is actual revenue, and that is actual real acceleration, and that is where value come from. And I think that there will be a cold shower when people understand, like, no one's actually gonna use and pay for these agents and tokens, and that wasn't actually really acceleration, but it'll drop back down.

Yeah. Yeah. Yeah.

I I mean, looked like, obviously, I think generally correct, and I agree. I think but people are going to try to become an AI company. No.

No. Absolutely. And I I had nothing against that.

And I this is no to be very clear, this is not a downer on anyone that's building this thing. Everyone has to get to, like, get the revenues, get the multiples, get the valuations, do what you have to get to the next step. Absolutely agree.

But we as a community are now, like, saying, oh, this is, the magical way to get up. This is not. Like, that that is not what is happening.

Right? No.

Would you go into offering GPUs? Oh, yeah. We will.

But enough for inference. Like, essentially, what we think about is, like, the GPU sandbox. So if you think of, like, if you have a GPU in your computer, that is what you have a GPU in the sandbox.

So there are workloads that do need GPUs. Again, I always go back to three d rendering because it's the easiest one to comprehend. But, like, if you wanna do any type of RL on, like, CAD or or something like that, you will need a GPU in the sandbox.

And so that's coming out as well. How about owned data centers? Owned data centers.

So we run on collocation providers, bare metal machines. Data centers, we technically can run on that or our own data center. Like, that's how we architected it.

Today, from a gross profit margin perspective, doesn't make sense for us to get in that. You have to raise a large amount of capital, a large amount of risk for, like, single digit percentage points. So today, that doesn't make sense, but we are fundamentally architected so that we can do that if we want.

Yeah. I mean, you're a large customer of these guys now. Do you see any opportunity?

what are we what are we gonna be by end of year? It's an insane number. I'm sort of scared to say it out loud.

So, like, it is it's very big. Just the the sandbox market on on and we there we talked about this in general. The entire infrastructure market is running 40% plus or minus month over month.

Everyone is growing 40% month over month. That's also a hot take. It's like if you're not growing 40% ish, it's not that it's just just the market.

You might as well you don't have to come to work. You'll grow that amount, basically. I'm half kidding, but, you know, that that's where it's going.

And so where does it end? We will see. The thing that I think about from from at least a CPU perspective, GPU is even crazier.

From a CPU perspective, it is like, there's a high probability that actually owning the CPUs beforehand will be a a go to market tactic. And it will probably because I you as you do probably talk to a lot of GPU providers, their growth is hindered by the amount of GPUs that you have. Right?

Yeah. Right? It's just like it's whatever NVIDIA decides to bless that day.

Yeah. Is that how much that's that's how much they're gonna grow. Right?

And so we're the CPU market in general, be it like something like Railway, for example, or Vercel or whatnot, or deployment or like the sandboxes, they're still CPUs. So, like, each is is growing at the pace of their the market and what their, you know, plus or minus of that market. But it's still not constrained by that.

And so my thought is, like, for for all of us in this market, databases fall into that as well, because database is also running CPUs. And it's like, we all have to grow as fast as we can so we can get enough of, you know, CPUs tomorrow from Intel or from NVIDIA because they have now CPUs and everyone else later on. So it'll be interesting when we get to that.

or new what's new Stripe? Like, what's the what's the analogy that Yeah. Is most appropriate?

It's interesting. There's, like, analogies of, like, so the you know, there's new Cloudflare, but new Cloudflare is new Cloudflare. Like, they're actually doing a really good job about, like And Cloudflare owns networking.

No one can fight. Like, come on. They're doing really well.

No. What I was gonna say in the sense of their whole agent portfolio Yeah. Is actually really good.

And I should say there are some technical limitations, I think, personally, around, like, everything's under constrained under workers. Like, workers is their thing. But from a go to market vision perspective, I think they're actually really, really good.

I think they actually get it unlike some other companies. And to your question is like, what is gonna be there will be an equivalent. Everyone says like an AWS for AI agents, but your answer like, it might look more like Stripe than AWS in a sense.

So there will be a cloud built out specifically for agents. And so that cloud will have sandboxes, and it will have web search, and it'll have databases like SQLite or Neon or or whatever specifically for agent and other things. We are not at the end of the new infrastructure primitives for agents.

There are more coming. Yeah. So people think like, oh, there's nothing else to sit.

There are more. Like, we have some ideas about the next ones. We don't have time to do them.

But there are definitely more primitives that are being built out for agents, and there will be, I think, a cloud that runs all that. Yeah. OpenAI has said AI cloud, Vercel has said AI cloud, and you are potentially also one or the other the prospective AI clouds.

I think it's a very big prize Yeah. To win. Well, thanks for coming on.

Thank you for having me. It's been amazing.