We all know Ship Happens. Today, we explore the world of AI sandboxes with Vasek Mlejnsky, founder of E2B. He reveals how safe, scalable execution environments are reshaping the future of AI agents. Discover the origin story, the open-source strategy, and the vision powering one of the most exciting tools in AI infrastructure today.
Ever wondered how developers safely build and scale AI agents?
From early coding experiments in Prague to launching a secure, scalable environment for AI code execution, Vasek Mlejnsky, founder of E2B shares the story behind one of the most exciting tools in AI infrastructure today.
Discover why building in public, open-source models, and a focus on developer experience are transforming the way AI applications are built. He also talks about fundraising, navigating the AI boom, and the strategic move to San Francisco that helped E2B grow.
Whether you’re a developer, founder, or just curious about the future of AI, this episode is packed with insights, practical advice, and a peek into the sandbox where the future of AI agents is being built. Hit play and join us as we explore the tools, strategies, and vision shaping the next generation of AI!
00:00 Introduction to Ship Happens Podcast
00:27 Meet Baek Linsky: Founder of E2B
01:09 The Journey to Founding E2B
02:17 Early Projects and Developer Tools
03:22 The Birth of E2B
04:17 Building a Community and Open Source
08:28 E2B's Technical Details and Vision
17:56 Fundraising and the Path to Success
22:15 Inspiration from Silicon Valley
23:34 The Decision to Move to San Francisco
24:44 Openness and Transparency in Business
26:14 Navigating the AI Boom
29:35 The Importance of Startups in Innovation
32:30 Future of Software Development with AI
38:30 The Role of Open Source Models
41:53 Conclusion and Future Plans
Vasek Mlejnsky is the founder of E2B, a platform offering secure, scalable sandboxes designed for AI code execution and agent development. With a background in developer tooling and open-source contributions, Vasek is shaping how developers safely build and deploy next-generation AI applications.
Playing in the AI Sandbox: How E2B Is Powering the Future of AI Agents
Per Krogslund: [00:00:00] But it is essentially just a story of a couple guys who came together, got a good idea, built something, and then it started from there. Basically. How do you navigate this like, and it feels like there's a new idea every day.
Vasek Mlejnsky: I, I think if you lose startups as customers, you are basically losing the innovation.
Having startups as customers means like you are on the edge of what's possible.
Per Krogslund: Welcome to Ship Happens, a podcast where I sit down with some of the smartest people in the industry to talk about the skills, techniques, and mindset that helps you to ship amazing software. Today I'm talking to Baek Linsky, founder of E two B, A company that's here to revolutionize the infrastructure used by AI agents in 2024.
E two B raised a 11.5 million seed round, and this July they announced a series A of $21 million. Our conversation today, I would like to pick Baek Brain on how you run an AI focused startup. How he gets ideas and what other aspiring founders can learn from me to be.[00:01:00]
It's a pleasure to wel welcome you to the show, Vasic. Thank you so much for taking the time to talk to me today.
Vasek Mlejnsky: Hey, thank you for having me.
Per Krogslund: All right. So I, I want to just go straight into the basics, like, 'cause to get a little background on you, because I, of course, in my research interview, I, I went to your LinkedIn, uh, page and you have two, you have two experiences on your LinkedIn page.
This is founder of E two B and, um, selling Swiss chocolate in school. That's your entire, like, uh, professional history. Help me fill in the blanks, like what happened between Vasic studying and Prague to, to founding E two B.
Vasek Mlejnsky: Yeah. Long story short, my co-founder and I have known each other a very long time since high school.
Now it's, I think 18 years now.
Per Krogslund: Okay.
Vasek Mlejnsky: And, uh, between, uh. Finishing the school or the university and E two B, we were mostly either working at some, um, startups or like early stage companies for a few months quickly realizing we are [00:02:00] probably not the best employees we can be, um, pushing, like trying to push our own agendas and maybe not being super happy.
Um, and at least me I can say that like always thinking about starting something on my own. Thomas, who is my co-founder and our CTO. Gotcha. Um. We almost every summer we, we spend like hacking some projects together. So, uh, we initially, we were both in the same school. Uh, he then went to other city. He traveled for the whole summer to to, to my place in Prague.
And we just, you know, like kinda sleep on a couch, build something, um, and, and release it. And there was anything from like hobby projects, like whatever idea we had. Okay. And very quickly it just became like developer tools because we were solving all, always our own problems. And as developers, like the easiest way to find problem, find a problem, and developers have a [00:03:00] lot of problems.
They complain a lot. You know, so, so every time we complain, we're like, let's not only complain, it's also try to fix it.
Per Krogslund: Yeah.
Vasek Mlejnsky: Um, and we kept kind of building, iterating. Um, there were a few more like no table projects. Uh, I can go into them if, if we have time, but, uh, eventually, like we took a break. There was GPD 3.5 came out, and I think it was like a, some, there wasn't like function calling back then, but people were experimenting and realized, hey, like you can actually give it a tool if you prompt it.
Right. Um, and we took, we, we just experimented with it. And we put an agent at that time, something we could maybe call an agent into a, like first version of our sandbox and gave it like a tools to write and run, run code. Um, okay. And yeah, that, that's, that's how the first version of E two B came to life.
But actually for the first like three, five days, we thought we would be [00:04:00] building the agent, not the infrastructure. Um, but then we quickly, um, pivoted to, to offering the infrastructure.
Per Krogslund: All right. I guess that's a good strategy since there is a billion ways to, to write an agent today. Like every, there's a framework on every shelf and come one coming out every day.
Um, so one thing I also saw when I researched you is your, your Twitter profile. Is this a way you learn or a way you share things or like
Vasek Mlejnsky: it It was, for us, it was a way to learn what people think about it. Very early when we started building our things, like I, I realized that. Having an audience that can give you both, like either positive or negative feedback.
Yeah. Um, it's, it's unfair advantage, right? I think the worst thing is if you like really something, there's just like nothing. There's no like eco you, you scream, um, into, you yell into the forest and don't hear anything back, and that is just frustrating because you don't have any data points. Yeah. And so what, what I started doing [00:05:00] at some, at some point with this project was like I wanted to share.
Everything we were building and just like push the, the videos and the demos out, show it to people. Um, subreddits were great for the product. Hand was great for,
Per Krogslund: yeah. So you're building a little community around your, your product ideas and just trying things out. Um. So you also like each, we are gonna dive a little bit into E two B in a moment.
Like one of the things also with E two B is that it's, it's open source. Uh, so this thing of transparency and sharing is, is that something that's just like very important to you or it's more like this has just made sense for you to kind of build your company this way?
Vasek Mlejnsky: I, I think it, it, it, it depends on what kind of company you are building.
Yeah. So for developer tools, like for a lot of them, it's very natural to, to be open source and very common I think for infrastructure tools that, um, it's even more common. Um mm-hmm. And I think it also, especially, you [00:06:00] know, you know, being open source can, can have. Uh, you, you can have multiple reasons to be there, and even if, and if you want, you can, you can stop, host it, right?
Um, so we saw that as a combination of multiple of these, um, of these points that I mentioned, like building a community, building trust. Uh, I also believe like the infrastructure software like ours, uh, should be open source and, and customers don't need to, and users don't need to ask for like. Using it.
They just can, you know, fork it, use it. Yeah. Uh, change it, whatever they want to do.
Per Krogslund: Yeah. And I guess that that's also like the experience developers want today. Like, they don't want to talk to a sales person. They just want to like, see it and try it, and then they can kind of make up their own mind without someone pushing a sale on them.
Vasek Mlejnsky: And what we see right now, even with big enterprises, um, big customers. It always starts from individual developer and it makes sense, right? Like, because even the CTO of a huge company, or like VP of engineering, head of engineering, whoever they [00:07:00] are developers, they, they were at some point they were like huge.
Someone who's like there probably over the weekend just taking things together and like. No one likes to go on a sales call and just like listen to like getting the quote and everything, taking weeks. Uh, I just want to use the tool. I just want to evaluate it. Right? So I think mod the playbook for modern dev tools, like it's not a top down process where you go to execs and then you start pushing it on Developers and developers are unhappy because they hate the tool.
Now it's like having a top-notch developer experience, having great self-serve, um, and. You know, like accepting the fact that your users are actually smart and they can figure out a lot of things, uh, themselves, provided you give them the information. Um, yeah, and they can, oftentimes they can connect the dots and then only after they validate your tool and that experience is like super, super great, um, then you can start pushing [00:08:00] things top down.
Uh, and they even, they even want that at some point. Like the, the bigger organizations.
Per Krogslund: Yeah. Yeah. And I guess it was like specifically for like a space, like the AI space right now with dev tools that is moving so fast and a lot of people are like catching up. They might not be very now, but they, they are likely to be in six months because everyone is trying to catch up on this thing.
Right?
Vasek Mlejnsky: Yeah. Um, that's, that's true. I think like it's especially true. We see right now.
Per Krogslund: Let's, let's dive into the, like the technical details. What is E two B exactly right now? Maybe it's changing into something different later.
Vasek Mlejnsky: Yeah, yeah. So, so the way I I explain it is that E two B is a developer tool.
It's an infrastructure developer tool that gives AI agents cloud computers. So imagine you want, um, you have cloud, um, plot code, you have open AI codex like agents or any other agent that are running on your MacBook. And instead of them running on your MacBook, you want to run. [00:09:00] 10,000 instances of that agent.
And you want to run, have them run in cloud and you want to do it fast at the same time, securely at large scale. Um, you want to connect storages into it, uh, make sure like no network request leave or you know, about all the network request coming in or out. And, and that requires this like cloud computer for each instance of the agent.
And that's what basically we give you, right? With like a one. Line of code, you can create a, uh, something we call Sandbox, which is basically a micro vm, a small vm, and uh, you can have, you can run your agent inside a vm. You can just have the VM almost as a, the sandbox as like a sidecar where like mm-hmm.
The code for with your LM exists on our backend and it's using tools inside the sandbox, um, and basically work as a. Literally like a, like a computer that you have in, in front of you.
Per Krogslund: Okay. [00:10:00] And the LM just see as like a function, it doesn't even know it's computer, I guess.
Vasek Mlejnsky: Uh, yeah. The LMS are now like, um, especially the latest ones.
They, they know like the concept of VM that can operate in like a secure sandbox. Um, so, so they. For them. It's like a, you can, there's a bunch of options. Uh, what, what you can do, right? So, so if you go like to most primitives van approach is like you have a file system tool, you have a code execution tool, you have a, um, uh, open browser tool, right?
Um, and all of these are doing something inside the sandbox through our SDK.
Per Krogslund: Okay?
Vasek Mlejnsky: The other other option is you can create, we can take. Cloud code, which is ACL I, you can put it in the sandbox and just interact with it, you know, remotely because it's somewhere running on our servers or, uh, deployed inside your VPC, uh, and you do 10,000, 10,000 of these instances at the same [00:11:00] time.
Um, and then you don't need to worry that much about the, the, the tools because that's already part of a cloud code. And you just say, Hey, you have, like, it's using just file system the same way it would be using file system on like a Linux laptop.
Per Krogslund: Yeah. Okay, so the initial selling point is isolation and security, but like down the line, when people's use maybe get a little bit more advanced.
Is this thing also being able to spin up like a a thousand agents and do a lot of like, work in parallel? Um,
Vasek Mlejnsky: it's a, it's a, it's a combination of things, right? So. Um, I would say security and well, at the bottom of any infrastructure tool, I think it's reliability. If you are not reliable, whatever we are building any fancy feature, it's useless because, like how I'm gonna use it, if it's, uh, you know, uh, if you are have a downtime two days out, out every week, um, or two hours, then um.
It start like as you go higher in like this pyramid, if you, if you can imagine, uh, it becomes more and more [00:12:00] abstract. So the next, uh, above it for us is gonna be security and isolation. Um, and, and this pyramid looks like a little bit, looks a little bit different for different type of a customer. It's gonna be different for a startup that's growing fast.
It's gonna be different for like enterprise company that's looking for. Stability, isolation, security, lot of observability and transparency. And, uh, there's a big part of everything's also dx, like making it super easy. Uh, to run a lot of these instances, manage them, know what's happening in them, get notifications like a, a callbacks when the instance dies, buy died, what was happening there.
Um, and the next step, like the parts two of this whole, um, sandbox stories like building products around the sandbox. Um, okay. Where it gets like really also really interesting. So this is like the next step of above the compute.
Per Krogslund: Okay, so there's more like specialized things. Is that where you, what are you gonna provide on [00:13:00] top?
Maybe you can't really talk about that yet.
Vasek Mlejnsky: Um, yeah, yeah, yeah, totally. So, so I would say like, fir, one of the first things is, um, persistance and storage, right? So you, when you start a sandbox, uh, you, the agent does the work there and you kill the, and then the sandbox. You kill the sandbox, you basically lose, lose all the state.
So how do you make sure that everything stays in? Like, oh, when you, when you kind of shut down the sandbox, you can resume it again like seven days later. Like imagine the same way you open your laptop the next day, everything is there as it was. Um, so that it means everything, both in file system. But also in memory.
Yeah. So, and how do you, and like this is relatively easy to do with one sandbox, one vm, uh, but very hard to do, uh, when you are running, you know, five millions of these per day, uh, hundreds of millions per, per month. And like at scale, essentially.
Per Krogslund: Yeah.
Vasek Mlejnsky: Um, [00:14:00] and to make sure unit economics works, everything is fast.
Um, so that's, that's one of the first things I would say. The another thing is, for example, um, secure, uh, secrets. So you need a way how to expose, for example, API keys inside the sandbox to the agent, to the L align without actually exposing the real value of those API keys. Yeah. Um, so kinda need to swap them in, in, in, in, uh, in flight.
Um. Another, another thing is storage. So, uh, basically the goal where, where I think this is going is like you have this infinitely scalable computer in a cloud for each agent, and it kinda morphs into the form what the agent needs at the same time, uh, at, at the right time. So as the agent is, uh, pulling bigger repositories, is, is running more complex workloads.
[00:15:00] The, the sandbox should kinda, uh, on its own, like increases storage. Like imagine it's like backed by something like S3 bucket and you can just keep uploading stuff as you go and you don't worry about it. Like if it's full, if it's not full, uh, and you pay as you go as as you are using it. Or it should like, increase the ram, increase the CPU, or decrease the CPU, uh, because like you want to, you want basically these workloads to be very optimized on what the agent is doing.
Um, and, um. And you don't want to overutilize, uh, or you don't want to like overkill the spec of the sandbox and just have like 32 CPU sitting there while the agent is basically running a, a simple Python server. And that's all.
Per Krogslund: Yeah.
Vasek Mlejnsky: Um, so yeah, like a, a lot of these things, it's like you imagine you have, you have like a, uh, uh.
A piece of infrastructure that can form into the right shapes ba very dynamically based on what the AI wants to do, what the agent wants to do. This is where [00:16:00] we are, like where we are going with with with each two B and our sandboxes. Yeah,
Per Krogslund: cool. Like it's, um, it's interesting to see how these, like cloud primitives we've seen over the last 10, 20 years and now they're morphing into this like, new use case.
And it's interesting how the, like the scaling challenge has become different in how we, we want to do this. Like we, we've looked at Kubernetes and like all these things over the last 20 years of like, really how, how do we handle scale? But I guess this is like an entirely new case, um, where the patterns look significantly different compared to how you would normally do it.
Vasek Mlejnsky: I would say. A lot of, for a lot of the cases, uh, the patterns are almost orthogonal to what we have right now. Yeah. So if you think about it, what we have right now, um, in a lot of infrastructure is there's like a very clear step where, um, you build the code, you deploy the code where you can optimize everything, right?
And it's, it's not bound by time. It's like bound by time that you don't want to wait hours to get deployed. [00:17:00] You probably want to have it in lower minutes. Um, but it's not like you need to have it in like a hundred milliseconds or 200 milliseconds.
Per Krogslund: No.
Vasek Mlejnsky: Um, but that's what you actually need with the sandboxes, right?
So we need a, uh, we need a sandbox to start right now, like in 250 milliseconds. Seven over the time we are pushing it lower. Um. And, uh, we don't know what kind of workload it's gonna run. We dunno what kind of code or what kind of files we'll get uploaded there.
Per Krogslund: Yeah. Cool. I was about to say that's the entire business model of AWS
Vasek Mlejnsky: Yeah.
Yeah. I mean, what we say, what, what we say, like when people ask us like, what's the big vision you are building? You say like, we are building new AWS we are building AWS for, for agents. It's just like we. Even users, not only us, even users have no idea what like, like how, how these records are going to, going to look like.
Per Krogslund: So I think also some of the listeners is probably also thinking about how do [00:18:00] I, how do I really get like an AI startup going and like, how do I get funding? How do I get people to believe this idea? Like at some, so I guess at some point you had like the confidence to, to ask VCs for money here, but like, how did you know, how did you know this is, this is the idea.
It's just like.
Vasek Mlejnsky: I don't think there was like a one point where I was like,
Per Krogslund: okay,
Vasek Mlejnsky: yeah, now, now we should like, this is it. Right? So, um, I, I heard this advice, I dunno where and when, uh, but like the best time to start fundraising if someone starts offering you money, uh, actually like, wants to send you money.
By the way, I think the fundraising is different, the process is different. When you are at pre stage C stage, series A is very different usually. Um. So now what I'm talking about, like, let's, let's say the, like the early, very early stage of like seed, pre seed. Um, and that's what actually happened with us.
Like we launched, um, we, I, I think I posted a video of, like a [00:19:00] demo of how an agent was using tools. Um, and it was like a. GPT-3 0.5 based agent, uh, met, uh, Greg Broman, red it on, on Twitter, and it kind of started snowballing. It was already open source. And then out of nowhere, like we started having investors reaching out and one of the investors were like, um.
Hey, like, do are, are you fundraising? Like, I want to send you a hundred K. Uh, and I was like, I guess we are fundraising. Like we had, we had, like, we had in mind, Thomas and I, like all the, the whole time we knew like we want to, we had like this hypothesis behind like why this can be fake. So it wasn't like we are fundraising just because someone is giving us money, but it's almost like the timing felt right because like someone was there actually sending us money.
Per Krogslund: Okay, because I, I was like, I was wondering like when I, when I looked at the story and so on, so, but it is essentially just a story of a couple guys who came together, got a good idea, built something, and then some personal Twitter saw and then it started from there, basically.
Vasek Mlejnsky: Yeah, [00:20:00] yeah, yeah. It's just like, uh, leaves out the power very like six to eight year years, like building project that didn't lead anywhere, but basically
Per Krogslund: that's always the thing you leave out, right?
Vasek Mlejnsky: Overnight success basically.
Per Krogslund: Yeah, the, the, the age year overnight success as it always is.
All right. So, um, but I think it's interesting in the sense that, uh, you're, I guess at that time you were also in the Czech Republic. Yeah. Like, you weren't like a, you weren't like a Silicon Valley in-crowd networking with folks on Sandhill Road, and like, you weren't, like, you weren't doing all these things.
You were just like picked up on the internet basically because your, your idea was so good that people shared it.
Vasek Mlejnsky: Exactly. Because we were sharing these things online. And from previous projects that even though they didn't lead anywhere in terms of like, uh, financial success mm-hmm. Um, they were interesting enough for people, you know, to, to get, like, get to people and, and like you don't need any permission to reach out to people.
Right. You can literally just [00:21:00] like, that's, that's better. Or X is great for that. You can just, you're building cool stuff. Uh, cool people will follow you because they find it interesting and that that's what they're looking for. Uh, and you know, that's how we got to the first set of our angels, some first early investors, uh, and, uh, at the time.
And I think it's, it's definitely still true now, even more than, than before. We had bigger network. In San Francisco without actually being in San Francisco. Uh, thanks to Twitter, thanks to X Uh, yeah, because you're posting things online and, and like they're then, then compared to Czech Republic and, and, and Prague.
Per Krogslund: Yeah. So did you always wanted you, did you always know that you wanted to go to the US to do this? Or like, was it just like what accidentally happened and I guess also like the, the European VC market is not like comparable to like what's happening. Yeah.
Vasek Mlejnsky: Uh, so, so we, Thomas and I actually got to San Francisco for the [00:22:00] first time around end of 21, uh, with a different project.
And we were part of this online accelerator called Pioneer, started by Al Gross. Uh, we got here, uh, for like a month. It was end of pandemic, so the city was pretty much dead.
Per Krogslund: Mm.
Vasek Mlejnsky: Uh, but we, because, but it was like a. Sort of like small dream coming through because we spend a week thanks to the accelerator of like visiting CEO of Twitch, CEO of stripe, uh, CEO, of Airtable.
And, and like just hearing all these stories about how they build their company and just meeting these people kinda and being in the community where it, it, it, it changes how you're. Think about things, right? So now you are not thinking about how do I send out a hundred emails a month? Uh, how do I send out a hundred thousand emails a month?
How do I get like, not a thousand, uh, dollars, MRR, uh, but how do I get like a hundred thousand dollars MR Uh, it shifts [00:23:00] about how you think about things and also helps neutralize all the people, uh, that started these super successful companies. They're there. Just people, they, they're at the same point you were.
And it seems like hard to, it feels hard to believe because they are so much in front of you. And of course they are smart, but like, there's no reason if you are probably like reasonably smart. And I think especially like if you can survive long enough and like withhold the pain, uh, why you shouldn't eventually be, be able to, to do it as well.
Uh, so when we started E two B. We always had this idea of like, we probably want to get to San Francisco to Silicon Valley. We weren't sure first, like maybe the company will be based in Prague and we will be like traveling like founders. Uh, but when we started the first year, um, we were, we two co-founders then We were four people, had two [00:24:00] first two employees, and we were four people for like a year, year and a half.
So pretty long time. Um. We, uh, we went to San Francisco a couple of times and every time we, uh, have we were here, everything moves faster and like all of our customers, uh, are here and we're making the time as well. So, so it made a lot of sense. Um, you know, um. For us to, to be here and just for, for the company to be more successful.
And it's like another one of those unfair advantages that you can have, of course, you can build a big successful company from, from Prague. It's just like, will you have it easier? Uh, or harder? And like, do you want to make it like harder? If it's gonna be hard enough anyway.
Per Krogslund: So one of the things that it seems like that was like critical to how you got to where you are now is the openness and transparency and, and you also have a, a certain belief in dev tools and being in the open source space and so on.
Um, did you ever have this like nagging feeling of like, no, I should not share this, this [00:25:00] is like my idea and someone could steal it?
Vasek Mlejnsky: No. I think especially in our space, um, I always felt like, well, more likely you will die because, um. Of your own mistakes and someone stealing your idea. Like no one cares, right?
Like you. No one knows you. No one cares. And it's actually an advantage. You can try many different things. You don't have like big user base that you need to serve, so you can iterate very, very quickly. You can pivot to whatever you want and you can just like put things out there. But. You know, uh, we got a lot of, during fundraising, especially in the early days, we, we had a lot of questions about like, what did they for you and copy you?
And I was like. Yeah. If they think this is worth their time, we are probably already pretty successful business, uh, doing like a hundred million a r and, uh, you know, that means we did something [00:26:00] right and then I will worry about it then. Or like I have a
Per Krogslund: yeah,
Vasek Mlejnsky: idea close to like what to do about it, like close to that.
Um, so I think it's a, it's a little silly thing to, to, to worry about.
Per Krogslund: All right. So I wanted to talk a little about like the AI space, and now you're, you're in San Francisco, so I'm guessing you're feeling the full force of like, you could say like some, call it an AI bubble now, the AI boom. So I think like in being a startup in the AI space, like how do you navigate this?
Like, and it feels like there's a new idea every day. Like everyone's hyped about something and then all of a sudden it's not the thing,
Vasek Mlejnsky: the space is moving fast and you want to. Keep tabs on a lot of things, but not necessarily like, act on a lot of things. Mm-hmm. Right. Sometimes like the right decision is just like, wait, don't do anything.
Yeah. Because like it's like it's gonna be different in three months or, uh, like this won't be relevant in, in, in three months. I think grade indicator, um, in, in like especially developer tools is [00:27:00] when after, for example, some announcement, um, or, or some, like a new type of a. Approach. You see like the community of other developers building around it, uh, like forming around it and building things with this approach over this product tool or whatever it is.
And that's like really good sign. Like there's like first science of a community like you have, if it's an open source tool, like you start having think issues on, on, on the repository. Um, it's in, for example, by the way, this is. Uh, I think like issues are open, issues is, is much better indicator than any stars because issues means that people are using it and running into problems.
Yeah. Stars is like, I can click, uh, on a button that that doesn't mean anything. Yeah. Um, and so, um, and, and just like obvious, like posting about it and posting about, uh, like showing demos is, is one part of it, but, um. I will [00:28:00] just repeat what everyone is saying, like there's a difference between like a cool demo and a full product.
But I do think that, uh, you know, the, the, the delta between these two was like almost enormous, starting beginning of 24, but now the delta is so much smaller. Um, and like when you have a cool demo on Twitter, I definitely noticed less like a. Over hype demos on Twitter. Yeah. Where someone is promising something that like, actually you tried and doesn't work at all.
Now either people are more careful like telling you, like kind of walking the line of like, yes, it works, it's great. But for these type of use cases, uh, for other types, like, yeah, we have work to do or they actually tell, they. To say like, yeah, it, it works. And you try like eight, uh, out of 10 cases or seven out of 10 cases.
It works depending of course, like what type of tool it is. Um, seven out of 10 might be good or, or, uh, terrible, right? So, [00:29:00] um, really depends on what, what you're building. But I generally, I think like, um, seeing the demos now and people like posting about these things, uh, is actually a good indicator. It's much better indicator than it was like a year or two years ago.
Per Krogslund: Yeah, so the space is maturing. People are getting a little bit more realistic sense about what you can do with this technology. And, um, maybe there's even less hype, maybe there's more people in it, but, but the hype is maybe a little bit lower now,
Vasek Mlejnsky: actually. I, I think definitely, I think there's still a lot of excitement, uh,
Per Krogslund: yeah.
Vasek Mlejnsky: At least in the, uh, like a investor community. Uh, also it's always funny, you know, uh, so, so we have this advantage of like, we are, uh. We, we, we have one office in San Francisco. We have another office in Prague. Um,
Per Krogslund: okay.
Vasek Mlejnsky: And, uh, it's always great to kinda step out, step out from the Silicon Valley bubble,
Per Krogslund: get out, get outta your bubble.
Yeah,
Vasek Mlejnsky: yeah, yeah, yeah. And, and then you see, uh, then for example, we go, when I go to Europe [00:30:00] and, and we have few customers there or just meet a few companies there. Um, like what type of problems they are. Solving and what type of perform? The startup that's like on the edge of what's possible solving in sf.
Uh, very different. Like, there's like a difference of one to two years, right? So, uh, enterprise customers are like integrating coding their operator while, uh, software, uh, sorry. Startups in, in San Francisco are, uh. Doing forking of, of agent instances and, and like paralyzing multiple agents. Yeah. So very different style of, of, of thinking and, and where their, their mind space is.
Per Krogslund: Yeah. But I guess it's also healthy from like a, you could say go to market perspective, that you actually know what you can say. Like, SF is like an edge case. Like it's a very specific kind of market and it's also like comparably small to, you could say, Europe or the rest of the world. Uh, so actually knowing what, like, you know.
More like you say, normal companies are doing with this technologies, but we're also good.
Vasek Mlejnsky: [00:31:00] Both are important, right? So, so, uh, the startups here actually show you what might be possible and what, like the, the more like normal, regular common companies will be doing in the next six to 12, 18 months. Yeah. And they help you to push and build your product to be state of the art.
Uh, and they move very fast. So we also, uh. Uh, you also need to like move very fast. And, uh, but, uh, then if you know you are making the right decisions and, and, and, uh, building the product the right way, you will. When, when the enterprise customer comes to you for the same feature the startup was asking you a year ago, uh, or the type of a product, you will be ready and, you know, like this can scale to a reasonable, um.
Levels. Um, and oftentimes the big enterprises, they get inspired by those, uh, startups.
Per Krogslund: Yeah.
Vasek Mlejnsky: You know, they see, uh, how they're [00:32:00] pushing CV art. They think, okay, we want to build something like this on our platform, what these startups are using, uh, in their tech stack. So I, I think it's super important. Like I think if you lose startups as customers.
You are basically losing the innovation, uh, like having startup as customers means like you are on the edge of what's possible and you keep up with the market.
Per Krogslund: Yeah, you have to be relevant in the space. And I think if you're not relevant with startups, you won't be with enterprises later. They're looking at that space for experience.
That's totally correct. So, um, so specifically for now, you, you're a dev tools company, uh, infrastructure company for developers. Um, so your opinion, like you're saying again, this is the, the extreme opinions in this space of like, either AI is replacing us all, or it's like a copilot that's gonna make us all better.
But like, if you look at the SDLC, like how developers build software, we've. We've had two decades of, yeah, everything is cloud native, Kubernetes containers, cloud, cloud, cloud and so on. And, and, and people [00:33:00] have adjusted to that. If, if you look at any like, bigger kind of organizations, that's kind like how they built software.
Um, but with agents, um, and all these new things coming out, how you, how do you see the SDLC. The lifecycle of software changing.
Vasek Mlejnsky: Yeah. I mean, um, trillion dollar question, right? So
Per Krogslund: if you know, I was saying, yeah.
Vasek Mlejnsky: Uh, this, this, this, an idea. I won't tell you. No, I'm joking. Um, I, I, I think one thing to keep in mind is that, uh.
While the space is moving super fast to actually transfer these businesses, like, uh, especially the large ones, takes a lot of time. Like just look at the cloud. And I think if you look at the like latest numbers of how many. Large organizations have transferred into cloud. It's mm-hmm. It's double digits percentage.
But I think I, I don't, I don't want to come up with a run number, but I think it was like fairly low number, right? Like you, I was surprised seeing it that like there's still a [00:34:00] very large percentage of businesses that haven't transferred, um, to a full cloud. So, uh, I think this is, that's one thing to, to keep in mind.
Um, and in the current ai, I think it's. Little dangerous to have any predictions for anything like more than eight months. Uh, you can, like, I'm trying to be too specific. You will just build the wrong thing. Um, so, you know, it's, it's just like, it's good to think about these as, as primitives like from the first principle, like LMS are getting more capable.
They're capable of understanding more code. Um, at the same time, you know, the, the inference, oh, sorry. The, um, inter. Uh, infrastructure layer of different tools are becoming more complex and ready for these LS So I think kinda, kinda, uh, the space is sketching up like the tools like VR building, like are allowing the [00:35:00] LMS to use their full potential and agents.
Um, and I, I, I honestly try to stay away from prediction like.
Per Krogslund: Hmm.
Vasek Mlejnsky: Uh, we will just be calling an agent, uh, through interface, like chatting interface in the next six months. Um, or we will, you know, just be coding from our phone or maybe there will always be like, human or we won't be coding. Right? Or there will be like a always human in the loop and, and it won't get better.
I, I, I do not think it won't get better, like the, the models. Um, I, at the same time I'm not like. A researcher. Uh, so, so, but I think what's important is to look at the primitives, look at what your customers are doing, users are doing, and listen, uh, listen to them. And then have a, like, gut feeling, uh, gut feeling [00:36:00] where this can go.
That should be like 20% of the roadmap and leave something for the experimentation. Um, because like those experiments, some of those experiments like. Can be the winners. Uh, and, and actually can, you can, um, can show you where the space or can, can, can be the right beds, where the space can go. So I think there's, it's a, like, it should be like 80 to 20, like 80% you follow your customers.
Um, because together you are discovering this whole new blue ocean world. And 20% is like you have your own bets because you've seen so many customers and so many use cases. You think about it 24 7 that maybe, like, you probably have, you, you should have a good idea about like where the space can go and you want to try and experiment with things and that's where you should like 20% of your roadmap you should dedicate to that.
That's how we are thinking at a, at a, at a team.
Per Krogslund: Yeah, I think that makes a lot of sense. And I guess like if you're, and as you say like being ultra specific, you might overlook interesting things that people might be using your primitive for and like, oh, that's actually, [00:37:00] I didn't think about that. That actually expands like the use case to use A-C-I-C-D or testing or doing something against production or whatever you like down the line that some other person might think of by just having the primitive, right.
So
Vasek Mlejnsky: Yeah. Yeah, yeah. Um, and you know, these ideas can come up with like. Very small things. We were just talking to a customer, they say one or two birds, and then kinda all clicks together because you had other a hundred conversations and now you have a full map in your, in your mind. Um, but I, I, I think like great things like.
You don't build the great thing right away and you are just right. You, you, you build a small version, you iterate very closely with a few customers, with a few users. You see like early inklings of like, it's, this is going the right direction, and you keep building on top of that. So, um, I, I think it's just like being very close to how people think about things.
Like what, and are they using it? This is just strong indicator of, and, and those are just like basic. You know, startup advice that you can find. Yeah. [00:38:00]
Per Krogslund: Yeah. It's gonna be interesting to see what people are gonna use the, the stories and the memory and like the, the different things you're gonna do, like the scale that you're allowing them to do.
That's gonna be interesting to see what happens
Vasek Mlejnsky: and, yeah. And, and, and just like bet on LMS are getting better. Right? That's, that's our, that's what we build this company on. It's like. The more LMS are getting better, we are getting better, we are more important and needed and we are getting better and more used.
Um, yeah. And if you figure out this flywheel, then I think you're on the right direction.
Per Krogslund: So the models keep getting better. Do you think like, again, this is like asking you to look into the future more than six months, but like we have the big foundational models now, like open AI and in Tropic and so on, but we also have like this undercurrent of a lot of like open source models and like small language models that is really.
Like they haven't really caught on, like, it doesn't seem like it's something that a lot of people are adopting, but they do exist and they get better and better as well. Like do, like, it's more like, where do you see this going? Like, do you see maybe these foundation models is not so important in the [00:39:00] future?
Vasek Mlejnsky: I, I definitely think like, as, especially in the last couple of weeks, uh, the open source models have shown that they are going to be extremely important. Like the, you know, the new model from, uh, cursor. Uh, from wind for recognition, like they are both built on top of open source Chinese models. Um, and I, I would like, I think you shouldn't bet against open source, right?
So, and like constraints, having, having, having constraints in a world like force you to be, um, more creative and, and get better. So I, and that's kinda, I think what happened with the open source model. Like you do not have the funding, you don't have the. The g all the GPUs, you, you, you, you would need. Yeah.
Uh, you don't have like a big company usually like, or like company at the scale of open AI or push behind it. And so you need to fi figure out what are your unfair [00:40:00] advantages that you can actually find a, a tech vector, uh, on the space. And, uh, that's why I think like the open source model really super, super interesting.
Uh, because they are kind of being born from the necessity, um, and, and from, from like, struggle, right? And that, that means they're figuring out something that, um, if like, if they figure it out, like they, like, let's say like you can be more effective on like less powerful gpu or like, uh, more effective per one kilowatt.
And that's like if, if you figure this out and actually it works, like there's gonna be huge transfer advantage compared to any, um, you know, AI life.
Per Krogslund: Yeah. And it's gonna push the other labs as well to, to, to figure that one out as well. Right. So,
Vasek Mlejnsky: yeah. Yeah. And, and, you know, uh, it makes total sense that the open source models are lacking a bit, like if you need such a huge capital investment.
Um, but I, I, I, I would, I would say, like, what do I, [00:41:00] I think like being open source model is like. It's like open source software, like you want to get everywhere. You want to come, like you want to have your model to be distributed everywhere and everything built on top of that. That's, that's a win, not not having everything closed and guarding it.
Um, I, I, I do think that's, um, that's more dangerous and that's also, uh, less like you will be, it'll be harder for your users. Uh, it, it's, it, it, it, it supports the innovation and, and, uh, it supports the international less and, and like the, uh, experimentation of other, other developers and not necessarily revenue.
I think that can be, that's a different question, but, uh, in terms of like innovation that I, I think is, um, easier if you have open source stuff.
Per Krogslund: So this was a super interesting conversation, va. So I think if, again, thank you so much for, for being here. Um, I really look forward to [00:42:00] seeing what's gonna happen with E two B and all your interesting things in the future.
So is there any, any last things you wanna, like, either recommend, anything you're specifically excited about for the future? New releases, plans, so on.
Vasek Mlejnsky: Um, so I definitely recommend checking our carriers page, uh, because we are hiring a lot of people. Um, so go to etb dev slash carriers. Uh, we are hiring both in, uh, San Francisco and, and Prague, uh, to, in our offices.
Um, uh, you know, the sandboxes are taking off. It's like very clear. This is, uh, one of the piece and, uh, core pieces of the infrastructure for our agents. Um, so. The growth we have seen so far. I, I think only like the fraction of what we are going to see, especially this year and, uh, sorry, next year. Um, so it's, now it's like I'm not, I have, I'm biased, but now is like the best time to join the company.
Per Krogslund: Yeah. Thank you so much for your openness and your transparency and sharing all these things. What's going on with the E two B and with [00:43:00] you. Um, so that is a wrap for today. And again, thank you so much for joining me and if you wanna follow Baek work, you should. Definitely check out his Twitter profile.
Check out each of B websites. Do you have a GitHub profile as well? Other people should check out?
Vasek Mlejnsky: Yeah. Yeah. It's github.com/E2B-Dev.
Per Krogslund: Cool. We'll put all this stuff in the show notes, and as always, the show is brought to you by Docker, the open platform for building, shipping and running apps.
Thank you so much for tuning in, and we'll see you again next time. Thank you.
Vasek Mlejnsky: Thank you for having me.
Per Krogslund: Thanks for tuning Into Shape Happens. Brought to you by Docker, the platform trusted by millions of developers to build, share, and run applications anywhere with dock teams, ship faster, stay secure, and collaborate without friction from code to cloud.
'cause when you build with docker, ship happens ready to make shipping simple, start your next project@docker.com and that's D-O-C-K-E [00:44:00] r.com.