Full Transcript

Eric Simons: (0:00) The the numbers we publicly disclosed, like, the first 8 weeks, we went from 0 to 20,000,000 of ARR, and and and we've just continued to grow since then. We actually wrote an operating system that runs natively inside of your browser that boots, like, 100 milliseconds. There's 0 latency, there's no cloud VMs that to get spun up. It gives you a very reliable experience that actually can scale. If you force people to go download a thing to their computer and set all that up, and then they get to use iD, the drop off is insane, right? Mean, it's like 90% or something, not more, whereas when it's native in the browser, it's frictionless. Then the actual AI augmentation, that actually even picks up from there, right? So there's an order of magnitude, more complexity of environment locally for any of these things.

Nathan Labenz: (0:43) Hello, and welcome back to the cognitive revolution. Today, we're simultaneously releasing the first 2 parts of a series that we're calling software supernova. With the makers of new and stunningly fast growing full stack AI developer products, Bolt and Lovable. Both episodes explore in different ways how AI's rapidly improving coding capabilities are beginning to tangibly transform the software industry by expanding the space of what can be built, changing how professional software developers work, and making it possible for people to create software without ever learning to code. My guest in this episode is Eric Simons, founder and CEO at Stackblitz, makers of Bolt, online at bolt.new, which has suddenly gone vertical after adding an AI coding assistant to the web based IDE that they'd spent the previous 7 years building. After announcing that they'd reached $20,000,000 in annual recurring revenue after just a couple months in market, the company has gone on to raise and recently announced a $105,000,000 funding round. In this conversation, Eric walks us through how the company's foundational web container technology works and how it's allowed them to scale so rapidly, explains the company's open source approach and shows where we can check out their prompts on GitHub, uses the agent to build a Spotify clone with just a few natural language interactions, showcases a few web apps that real customers have built, and describes how product managers are having especially notable success using Bolt to bypass design platforms like Figma and instead go directly to working prototypes. Video of the demo, by the way, is on YouTube, and the app design did come out looking really good. But you can take my word for that, and I think you'll be fine with the audio only version if you prefer. We also discussed how the economics of AI development are creating unique opportunities for Bolt users. 1 freelancer, Eric says, spent $7 on Bolt inference to deliver a $9,000 project for a client. To be honest, I have no idea how much of that is really happening in the economy at large or how long it will last. But I can confirm that there is at least some real AI arbitrage going on right now. Businesses that are used to paying tens of thousands of dollars for projects are now effectively paying thousands of dollars per hour in some cases without realizing it. In any case, I'm confident that the market, including the market for rapid response human help that Eric and team are planning to build on the Bolt platform, will rapidly bring prices down and send quantities soaring before we know it. And once again, I feel like the safest bet in the AI revolution is high consumer surplus. As always, if you're finding value in the show, please share it with friends, write a review, or reach out via our website, cognitiverevolution.ai. We always welcome your feedback and suggestions. For now, I hope you enjoy this unique perspective on AI powered software development with Eric Simons, CEO of Stackblitz, makers of Bolt online at bolt.new. Eric Simons, founder and CEO at Stackblitz, makers of Bolt online at bolt.new. Welcome to the Cognitive Revolution.

Eric Simons: (3:43) Awesome. Thanks for having me. Excited to be here.

Nathan Labenz: (3:46) You guys have made a splash to say the least in 1 of the hottest categories going right now, which is the full stack AI software engineer. People are prompting their way to all sorts of creations on your product and the revenue is popping off. You guys have raised a bunch of money. We're going to get it on into all of that. For starters, I would love to get just a kind of general sense of your AI worldview, because I feel like so many of these conversations are much more intelligible, much easier to sort of understand the details when you have the person's big picture in mind. How would you describe where we are right now in the AI moment? What is your expectation for like AGI or transformative AI? Are you somebody who worries about risks? Do you maintain a P doom? Give me the Eric Simons AI worldview in a minute.

Eric Simons: (4:35) I guess to me, what's been very interesting, you kind rewind a couple of years ChatGPT came out the first year after that, I mean, it's just people are like, AGI is going to be here within a year, dah, dah. I mean, there's lot of hype and understandably because there's this obviously a pretty big 0 to 1 moment sort of thing. And these models have gotten really good at certain things, but software is 1 that makes sense. It's getting a lot faster because code is text based. So it's easy to train these things. Code is deterministic. If you want to train an LLM to be good at law, you're looking at the cases dating back to the 1700s and how this judge felt at the time. It's not deterministic. But with code, can write code, AI can write code, evaluate it, did it do the thing or not? And then that's a new test case, right? So it makes sense that in the past year, that's where I think some of the biggest leaps have happened and that's going to continue. So I guess my pragmatic AI worldview is just the world software orders getting rewritten. I think the world software is actually going to be maybe the first major tectonic shift that's going to happen where stuff is just getting replaced by AI. Software is 1 to 1 getting replaced by being generated new. And I think that's going be huge because the bar is going be lowered for creating software. So I think to me is the most interesting, very tangible what is happening right now versus kind of, okay, postulating like where a year from now, 5 years from now. Yeah, I mean, and then at least to me, I like to take a more like, what's the pragmatic on the ground view of what we actually see moving in a meaningful way. And obviously the stuff around reasoning is a pretty big breakthrough, I would say. But again, where's going to have the most, where's it going to be a multiplier, particularly it's like software, Apply reasoning into software development. Boom, that's huge. Anyway, so I think that's kind of my general worldview. Holistically, think it's just going to be a good thing. Don't think that this is going to be like every other technology that's come before this, it tends to be a great enabler of a different future. It's things are reconfigured differently, but it doesn't strike me that what I see now is just people that couldn't code before being able to build real production products with software. Engineers that are fantastic today can leverage themselves 5, 10x. So the world needs more software. We've been constrained.

Nathan Labenz: (6:58) Yeah. It's definitely been pretty much universal, I would say among businesses that create software at all, that if you had asked, you know, at any point in time and still to today, you know, is there more you would like to be creating if you could then they're all like, yeah, we're totally constrained by how many developers we have. We're constantly making these trade offs, constantly prioritizing. I've certainly lived that life myself. But your, you know, your comments about reasoning, obviously we're in the 1 to 3 transition period and also the, you know, the R1 moment. It is crazy to think that we haven't seen it as the public yet, but the 3 results where they have the model in the top 200 coders in the world on competitive coding challenges is really a pretty crazy result, right? I mean, that's like beating almost everybody at OpenAI. Think they thought there was like maybe 1 person left at OpenAI who has a higher score in competitive coding challenges than the latest model does, which is pretty wild to think about. So how do you, I mean, I sort of am a big believer in like, we need more software. I've also kind of started to be a big believer in the concept of luxury software, I just saw the phrase for the first time yesterday, selfish software, you know, but, and those are kind of similar ideas, but basically like potentially you don't need to create software for a mass market. Potentially you can just create it for yourself or for your team, or, you know, you can create AI apps that spend an unreasonable amount of tokens because it's worth it to you, even if you couldn't necessarily take it to market in a scalable way. I would love to hear what your vision is in terms of how you think we'll be interacting with software, what the software landscape might look like in the future. Because I do also feel attention where I'm like, to some degree, yes, I want these personal things that are just made exactly for what I want. But then I also am kind of like, but also I do really want to use like well polished stuff with like good UIs, you know, it's sort of in today's world, it's easy to have that vision, but then it's also easy to get to a point where you're like, I tried making this thing for myself,

Eric Simons: (9:15) but it has a lot

Nathan Labenz: (9:15) of bugs and it actually is not, you know, I'm not exactly living the dream yet. So, you know, give me a little, put a little more color on like what the future of the software development and user experience might be.

Eric Simons: (9:27) No, it's a good point, man. If I had to like rephrase this, but you're kind pointing out there's kind of a long tail of like, oh, you can have this thing punch out software, right? But there's a long tail of like UX and even just support issues with it that certainly today are not addressed. It's like, even if you kind of look forward, it's like, okay, well, if you're constantly having to have, at least today our expensive models and maybe that price, it's just gonna keep going down. Like, I mean, so it's just really a question of how far out into the future is the long tail dealt with? Where does equilibrium hit basically? And I think you kind of hit the nail on the head is I think for certain use cases today, already makes sense, right? Like for people that are using Bolt, they're building internal tools for their company that are actually relied upon for critical business function or whatever. People are building real products that are like their startups and that sort of thing. So it's like, I think in both of those cases, you're dealing with people that have a keen eye for product. And I think the people that find the most success with Bold, and I imagine other tools in a similar space, they tend to be PMs and designers and that sort of And of course, engineers, but I'm going to kind of put them aside because they know how to write software. But the people that don't know how to code, it's like entrepreneurs, PMs, they know how to build great products. They know what that experience looks like and feels like. They know how to actually just sculpt the thing. And previously they could only write that code through someone else's fingertips. And so they, again, PMs, their job is really interfacing with developers and helping point them in the right direction on how to actually make the product experience great. Now with the AI, they're just saying basically the same words. They're just getting an instant result. Like, no, no, no, More of this, less of that or whatever. So anyways, I think for folks that fit that category, I think that they're already finding a lot of success today. Over the next year or 2, it's getting even more. When you kind of get to general end user, I think the point's extremely valid, where it's like to reasonably expect someone who has not been in the business of building products to really be able to build something that's going to be bug free and a great experience, da da da. I think there's going to need to be more, in short, under the hood, just more inference and more capabilities in these models and the agents to actually allow you to do that. Because right now, it's just very rudimentary. It's very rudimentary. And there's a lot of issues that people run into when they use Cursor or any other tool, then when they're non technical, it's like, when stuff goes wrong, it's like, how do you fix it? So that's, I think, certainly trending that way. It's like, I think a lot of the folks in that category, PMs, entrepreneurs, designers, etcetera, I think that's going to be the first big market unlock here. It's already happening. These folks are coming online writing software and like next few years that's going to explode.

Nathan Labenz: (12:16) That's quite interesting and sort of anticipated another question I had in terms of just like, who are your target customers today? So it sounds like they are people who are gaining leverage and doing things they probably otherwise just would never see bubble up to the top of their to do list because they can now talk to a product like Bolt instead of having to talk to a human. And obviously the cost ratio there is massive.

Eric Simons: (12:43) Yeah. So our user base, like we see something like 40% developers, 30%, 40% are traditional software developers. 60%, 70% is folks that are like PMs, entrepreneurs, that sort of category. They're not software developers. In the developer category, this is a 10x multiplier on their ability to go and make prototypes or ship, create UI or whatever have you. So they're coming to us to drop in stuff from Figma or prototype out an app. They've been thinking instead of going and writing out a whole bunch of code or having instrument stuff in cursor or whatever, they just come to bolt. New, they drop it in, hit enter, a couple of prompts, they get something great, they pull it out to Git or locally or whatever, and they continue building on it. So that's a very concrete use case. That's 1 of the ROI stories. There's this guy CJ on Twitter. I don't know what his handle is, but he's got the most insane ROI that I've seen. But basically, so he's like a freelancer and his client had him build a web app. It was like some dashboard or something. And he spent, I think it was like $7 worth of inference on us. And someone asked the question I would have asked, had they not, which was like, how much did you build the client? And the answer is $9,000 So if you go to Twitter and you search Bold Arbitrage, there's a lot of developers, freelancers that are connecting the dots here, because the demand supply curve has not begun to normalize on that yet, right? So that's kind of that 1 category. We have a Fortune 500 using it to prototype instead of using Figma and that sort of thing. On the PMs entrepreneur side, folks are coming here where whether it's a side project, they want to launch a startup or a side business. They're using Bolt to build full stack applications with payments and authentication and database and all that sort of stuff. And even if they're at a company, again, if you think about the role of a PM, a lot of their job is writing Jira tickets, explaining how to make the product better, assigning it to somebody who's going to code it. They say they did it. The PM reviews the work. They give feedback. It's like, wouldn't it be great if they could just, instead of writing the Jira ticket, they just tell an AI to do it, get the immediate feedback when it looks good, they just commit it. And then the developer can pull it down and they had no idea if that was written by a coworker or an AI. It doesn't matter. And so we're seeing that use case a lot as well for both people that are starting their own stuff, starting their own project startups. But now we're seeing businesses start to come online where it's like, Oh, okay, this is a huge multiplier. Devs, it's a better use of the dev's time work on really tough functionality that the AI models are not great at or whatever, right? 0 shot at least. And the PMs can go in and make the sort of level of UX policy you're describing where it's like, make the buttons align, make the colors right. Like, make sure that this goes to that page when you click submit. Simple stuff like that. It's like, they can just do it and they can make sure it's perfect and then ship it. Right? So those are kind of the 2 segments of the audience we see in the sort of use cases that they're finding a lot of success with.

Nathan Labenz: (15:44)

Hey. We'll continue our interview in a moment after a word from our sponsors. In business, they say you can have better, cheaper, or faster, but you only get to pick 2. But what if you could have all 3 at the same time? That's exactly what Coher, Thomson Reuters, and Specialized Bikes have since they upgraded to the next generation of the cloud, Oracle Cloud Infrastructure. OCI is the blazing fast platform for your infrastructure, database, application development, and AI needs, where you can run any workload in a high availability, consistently high performance environment, and spend less than you would with other clouds. How is it faster? OCI's block storage gives you more operations per second. Cheaper? OCI costs up to 50% less for compute, 70% less for storage, and 80% less for networking. And better? In test after test, OCI customers report lower latency and higher bandwidth versus other clouds. This is the cloud built for AI and all of your biggest workloads. Right now, with 0 commitment, try OCI for free. Head to oracle.com/cognitive. That's oracle.com/cognitive.

Nathan Labenz: (16:57)

It is an interesting time for business. Tariff and trade policies are dynamic, supply chains squeezed, and cash flow tighter than ever. If your business can't adapt in real time, you are in a world of hurt. You need total visibility from global shipments to tariff impacts to real time cash flow, and that's NetSuite by Oracle, your AI powered business management suite trusted by over 42,000 businesses. NetSuite is the number 1 cloud ERP for many reasons. It brings accounting, financial management, inventory, and HR all together into 1 suite. That gives you 1 source of truth, giving you visibility and the control you need to make quick decisions. And with real time forecasting, you're peering into the future with actionable data. Plus with AI embedded throughout, you can automate a lot of those everyday tasks, letting your teams stay strategic. NetSuite helps you know what's stuck, what it's costing you, and how to pivot fast. Because in the AI era, there is nothing more important than speed of execution. It's 1 system, giving you full control and the ability to tame the chaos. That is NetSuite by Oracle. If your revenues are at least in the 7 figures, download the free ebook, Navigating Global Trade, 3 Insights for Leaders at netsuite.com/cognitive. That's netsuite.com/cognitive.

Nathan Labenz: (18:23) Yeah. That's that's interesting. So if I imagine not too long into the future, it seems like everybody has this vision and it seems like from your answer, we're not quite there yet to where people who have not been in the product creation business at all could start to get into the game. I imagine kind of a number of different stumbling points or barriers that they would have to get over. Some of which would apply even to today's users, the developers, and even for the PMs and even the developers, but others would be sort of easier for your current user profile, but harder for these like my dad or my mom who's never done anything like this, but my dad has plenty of ideas. So maybe we could just run through a few and you can kind of tell me how you're thinking about it. 1 is like just drawing out of people what it is that they want. I think when my dad sits down at something like this, he greatly under specifies what it is that he even wants and doesn't even know that he's doing that. Right? He's as like, he's so green in this area. You know? And then he gets something back probably from a lot of products where he's like, that's not what I wanted. And so I wonder to what degree there's a way to get over that. I also would flag, as you did, you know, when things are not working, how do you fix them? That has been probably 1 of the trickiest pain points. I'm really interested to hear how you are addressing that and, you know, what you think the sort of trajectory is there. Another 1 is deployment. And that's also, I'm sure, like a really critical 1 for your business because I've lived with this reality too at times. It's like, it's 1 thing if people can come in, make something with your product, export it and never have to come back. It's another thing if you become part of how they actually operate their business. So I'm sure that's something you think about a lot and deployment is definitely a major pain point. You might also have your own taxonomy of other challenges, but sound off on those 3 at least.

Eric Simons: (20:19) So on the first 1, right? People that have not worked as entrepreneurs or product folks before, etcetera, we have folks that are coming to the product and using it that don't fit those categories exactly. And everyone has to learn how to build great products. That's like there's kind of a general skill set of kind of learning how to really sculpt something. And what's cool about Bolt is, and again, just these AI generative tools is it's lowering the barrier to entry to come and learn how to actually build great stuff instead of getting mired in Like, you know what, my co founder and I, we grew up down the street from each other in Chicago when we were 13 and we learned how to code together. And we learned how to code together because we wanted to build products. That was the reason we got into building software. If this existed then, I don't know how deep we would have necessarily had to go or have gone on becoming software engineers. Because our end goal is to build great products. And so I think we have a ton of people that are coming today to BOLT and they're learning by actually working with the AI and how to best leverage it to build products and picking up the general, how do I build products or how do I build a business skillset, right? So it's funny, actually a couple of weeks before we launched, I had my mom who's 73 test this thing out because she's the I love her, but she would admit this readily. She's probably the least technical person you'll ever meet. And she built and deployed her first website on this thing without any guidance because it's just simplest interface in the world. It's a text box that you type in what you want, you hit enter, and then there's a deploy button where you get a live URL of the thing. That's it. There's a lot of use cases too that we've seen of just people building this for personal for weddings. For 1 guy who had tweeted us, his daughter needed medical donations or something like that. And he created a website to do that, which to me was kind of like, was touching, but I was also like, should I tell this guy that Wix exists prior to Squarespace? Because we're not the first people that have allowed you to build a website. But then it hit me, like for my wedding website in 2021, I used Squarespace because then my wife, I had originally wanted to build the site myself. I spent a week, like a day on a weekend on it, never came back to it because it just takes a lot longer than you want. So she made me use Squarespace and man, the interfaces for those things suck. These drag and drop builders are really complicated to use. And they're supposed to be targeting less sophisticated computer users than me. But when you look at Bold, this is it. So it's like that entire category is going to go away. Mean, there's no reason for drag and drop wissies when you have this sort of stuff, right? So anyway, so on how do you educate these folks? How do you make them successful? So I think there's 2 direct ways that you can do, and then there's kind of a rising tide thing that's going on. But basically, 1, education, like teaching folks. Instead of just the school of hard knocks of them prompting the AI and figuring it out, they create great resources. I think this is 1 of the things that's been very effective for us is we've invested a lot in our community because if you look at Midjourney as probably 1 of the best examples, what made Midjourney really successful was the community of people prompting, but then sharing those prompts and etcetera. And so that's kind of the same thing we've got with Bolt where we have a very vibrant community of folks that are sharing how they are doing it. And we're learning a ton from these people because they've used the product more than our own core team has at this point. They're the kind of the domain experts on the thing. So we invest a lot in our community, our educational resources, that sort of thing. That's like a long running sort of how do you kind of ramp people? The second thing is just we're not at a point tangibly, depending on if you're trying to build something pretty complex, we're just not at a point where you can actually, without any technical knowledge, just rely on an AI to get all the way there. And really what you need is just someone like DeGraut who's reasonably technical or is just very good at debugging these things or whatever to grab 30 minutes of their time and just kind of plug them in your code base. And so we haven't announced this. I think we're gonna announce it next week. I don't when this is coming out. But if this scoops our announcement, I guess it was heard here first. But so we're going be we're actually rolling out a program where within BOLT, you can actually kind of raise your hand and go, hey, I'm stuck on something. And we can actually connect you with an expert that we'd certified that can punch in, help you get debugged real quick and then get you back on your way. So just keep using the AI and that sort of thing. So it's just like basically a way to just get quick on demand help to get unblocked or get a professional opinion on it and then move on and that sort of thing. I'm not aware of anyone else that's done that model really well. So I think we will probably be 1 of the first. And I think the third thing is these models just keep getting better and we're making our agent better and better, etcetera. And the quality of the models has just a multiplier on the swing of everything that's leading up to that, like how much you have to learn, how many times you need to loop a human into the process to help debug you, right? So it's like, I think if you look forward over the next 3, 6, 12, 24 months, mean, it's just the better those get, the less you're going have to rely on the other things to actually keep proceeding. Right? But even today, it's like we've crossed this chasm where it's crazy valuable. That wasn't true a year ago.

Nathan Labenz: (25:26) Yeah. I think we've officially put to bed the no progress since GPT-four narrative that popped up for a little while there, but even before like the 1 sequence, was like, guys, I feel like there's been a lot of progress on just making GPT-four class models, you know, do the things that we actually want them to do in a much more consistently useful way. So I totally agree that the progress has been pretty amazing, especially in areas like code. So it's really, I think that the human in the loop thing is really interesting. You guys are literally exploding right now. I don't know what metrics you're tracking or how much time you have to even stay up to date on metrics, but do you have a sense for what your success rate is? If somebody sits down at the product and is like, I want to build X, how often do they actually get to X versus getting stuck somewhere that they can't get out of? And do you have a sense also for like what has moved the needle on that success? I assume that, you know, getting the model to do a better job is core to that, but I'm I'm kind of wondering like, you know, are you fine tuning? You did mention like choosing integrations. That's another thing that I think people are like very overwhelmed by, by default, right? Oh my God, I want to do payments, but what do I use? Yeah. It seems like you're being pretty prescriptive on sort of, this is the happy path that we think you, you know, We really think you should go this way. As I mentioned, that's been a driver of the success rate. But yeah, what is the success rate? What's driving the success rate? And what do you think will drive it in the future?

Eric Simons: (27:00) Yeah, good question. I think we're seeing a good success rate. I think that at high level, you measure this with just revenue growth. And for us, the numbers we publicly disclosed, in the first 8 weeks, we went from 0 to $20,000,000 of ARR, and we've just continued to grow since then. And so we haven't seen the ARR go down at all. So I think that when you dig into the usage metrics as well, it's just we're kind of setting new, more and more folks coming to the platform, the retention is extremely good. With a lot of these products in the AI direct to consumer space, the churn on these things can be insane. I mean, they can be like 70%. We're nowhere close to that. I think the acceptable is anywhere from like 20 to 50. We're hanging out like the 25% range, which is pretty good. So people are finding a lot of success in the product and find a lot of value in it because they're sticking around, they're coming back every day, etcetera. And I think that's to There's some You can kind of look this stuff up online, but as far as AI code products globally, if you kind of exclude Microsoft GitHub and etcetera, we look at the startups from I think free markets are a good way to just measure how much value is being generated for folks. In startups, Cursor is number 1 at 100,000,000, we're number 2. And then there's a whole category of folks that are kind of in the anywhere from 5 to 15 ARR or something like that. And so there's really only been at this 0.1 couple of, there's really only 2 tools at this point. They've kind of broken out for the rest of the pack and that's us and Kurz. I'm sure there's going be other folks that pop out and that sort of thing. But I kind of look at that and it's like, clearly something here is working. And for our product, a lot of this comes down to this technology we made that enables this experience. Can show it in a second, but basically you think about, okay, well, if I'm gonna come to bolt.new and I'm to type in words and get a full stack web application back, that sounds great, right? But when you actually go to build these products, the question is, Okay, well, are you actually running the development environment for that? Like where that full stack app is running that the AI is programming? How do you do that? It just so happens, this is like what our company has been working on for the past 7 years. We started out as a web based IDE. The traditional way that you solve this problem is you put it on a Cloud VM. Everyone gets their own Cloud VM that's using the service, etcetera. And that's kind of been the model of Cloud IDE since Cloud9 was the first. And there's never been a breakthrough application that has been used by lots of people that has this model where every user gets a VM for a couple of reasons. 1, the experience tends to not be great because you have latency. So every keystroke you're putting in has to be synced to a server. Result of this has to be synced back to you. Someone's got to pick up the cost of that cloud VM because these things are, they cost money. If you want to scale to 1000000000 people, that's 1000000000 VMs. So it's a good amount of capital. But the other problem is there's not 1000000000 VMs that you can rent on the planet. But there are 1000000000 devices. So there's billion laptops, iPads, whatever. And so if you look at how Google Docs and Figma work, they don't spin up a cloud VM for each user to actually render your designs or as you're typing the document, your keystroke show up immediately. And that's because it's using your CPU, your memory, your GPU through the browser as the interface to those parts of your device. And so what our company did is we actually wrote an operating system that runs natively inside of your browser that boots like 100 milliseconds. There's 0 latency, there's no cloud VMs that to get spun up. It gives you a very reliable experience that actually can scale. And so when you kind of look at the other approaches in the space, a lot of the issues folks will run into is the cloud VM gets borked. It's extremely slow, the latency. There's tons of issues that can kind of happen with these things because they're just running in the cloud, tends to be a lot more expensive. And whereas with Bolt, come here and 100 milliseconds, have a live environment, the AI is just punching code into it and you see the result instantly. I think that's under the hood, something that's maybe not obvious to folks that are visiting, but from a technical perspective, that's the crown jewel of our company. We spent 5 years building Web Container, this technology. And it just so happened that what we've been building with Frontier AI makes this magic experience that just kind of is impossible with any other approach. So I think that in a nutshell is the key enabler for us. And then beyond that, it's certainly like we're doing a lot of stuff around our AI agent, like a couple of folks in our team that actually built the Webkinear technology just so happens that before that they were working in ML and AI. And so when it came time to actually write our AI agent, they knew exactly what to do to go and build them, but also instrument it so we can actually leverage the data and the insights to fine tune and make it really reliable and not run into tons of errors for the end user because again, we want it to be a smooth experience. So in a nutshell, there's kind of a lot there, but I think that's the 3 60 of like, why is the experience this so different and kind of why has Bolt been a completely different trajectory comparatively to kind of the other stuff that's existed before us and after us.

Nathan Labenz: (32:15)

Hey. We'll continue our interview in a moment after a word from our sponsors. Being an entrepreneur, I can say from personal experience, can be an intimidating and at times lonely experience. There are so many jobs to be done and often nobody to turn to when things go wrong. That's just 1 of many reasons that founders absolutely must choose their technology platforms carefully. Pick the right 1 and the technology can play important roles for you. Pick the wrong 1 and you might find yourself fighting fires alone. In the ecommerce space, of course, there's never been a better platform than Shopify. Shopify is the commerce platform behind millions of businesses around the world and 10% of all ecommerce in The United States. From household names like Mattel and Gymshark to brands just getting started. With hundreds of ready to use templates, Shopify helps you build a beautiful online store to match your brand style, just as if you had your own design studio. With helpful AI tools that write product descriptions, page headlines, and even enhance your product photography, it's like you have your own content team. And with the ability to easily create email and social media campaigns, you can reach your customers wherever they're scrolling or strolling, just as if you had a full marketing department behind you. Best yet, Shopify is your commerce expert with world class expertise in everything from managing inventory to international shipping to processing returns and beyond. If you're ready to sell, you're ready for Shopify. Turn your big business idea into cha ching with Shopify on your side. Sign up for your $1 per month trial and start selling today at shopify.com/cognitive. Visit shopify.com/cognitive. Once more, that's shopify.com/cognitive.

Nathan Labenz: (34:17) I've got, like, a growing collection of these stories of people who were building something hard, and then AI came along and made it like, you know, 10x more valuable to sometimes the same user they originally envisioned and sometimes like a entirely new user base. And it sounds like you've had a bit of both, but it's really expanded who can use the product. I've lived for what it's worth a pretty similar story at my company, is Waymark, where we do video creation for small businesses. We've been on this journey, I call it from DIY to done for you by AI over the last few years. But similar thing where we spent a few years building an in browser rendering engine for videos. And because we had couple and 1 in particular major early partnership with a cable company Spectrum Reach, we had to deliver TV quality specs. So we couldn't take any, like, shortcuts on the final thing. It had to be 30 frames a second and HD and so on and so forth. Was a little bit crazy to try to do that in the browser. And we ended up with a lot of work into that tech stack. And then it was kind of like, I don't know if you've experienced something similar, but we did sort of have a in between moment where we were like, okay, this thing's working pretty well. But then we sort of found like, people that are pro video creators, they still wanna use the pro tools. People that are not pro video creators were giving us feedback like, this is great. You know, it's easy to use. And we were like, so why don't you use it more? And they were like, I don't really do. I don't really have any ideas for videos. Or I did have an idea, but, you know, I sat down to try to write it and I kinda got frustrated and we're like, you know, could we make it, could we improve the product? A lot of things are like, not really, you know, it's just, it's on me, but you know, but it is what it is. And so the AI thing like totally changed that. Now the, you know, it's sort of accessibility in a web, you know, browser based sense did not automatically translate for us to actual end user value, but then layering on the AI totally transformed that. Haven't grown certainly as fast as you've grown, but it has definitely changed the trajectory for us too.

Eric Simons: (36:35) Yeah, I imagine. I mean, that's the exact same story. And like if you rewind 6 months, we had this awesome technology, developers loved it, but it's just like we couldn't get them to leave their local tools and therefore build a meaningful revenue generating business out of the thing. That's where we're looking at like, okay, well, we might need to start figuring out how to wind this thing down because at some point after 7 years, this cool technology, but we're a benchmark company. This has to make sense, right? And that what you just said is exactly what happened with us. And so hearing it's happened you, with I remember you mentioned this a little bit when we chatted a couple of weeks ago, but it makes me think there's going be more of this, right? Because that's the idea of like, if you think about people that are nontechnical, if you look at any profession, most professional tools, if you're going to be using a computer to do it, are local, right? Like video is something that's a graphic setting. Until Figma was local, right? And Canva, I should say too, for that matter. Development, video editing, video games, there's kind of this list of things that have not come to the browser yet. And because it's just really tough to get professionals out of the existing environments that they are in. Figma and the G Suite are kind of the only ones that have really broken through. But with AI actually lowering the barrier to entry, I actually think this stuff being able to live in the browser is a key part because the hardest part, if you want to become a software engineer, and this is when we started Stackblitz, the company that built a pool, the key insight that we had is like setting up dev environments has always been the worst part of this from when we were 13 learning how to code. It's like, you're not even learning how to code. It's like week 1 and 2, learn how to just get stuff installed on your computer and just get a dev server. So you're even coding. You're just poking and prodding your machine to get it to work right. And this happens when you join Meta or Netflix, first month is onboarding to just run the stuff on your computer. It's crazy. And so, yeah, I mean, if the experience of your product as a video editing tool, right? If you force people to go download a thing to their computer and set all that up, and then they get to use iD, the drop off is insane, right? I mean, it's like 90% or something, not more, whereas when it's native in the browser, it's frictionless. And the actual AI augmentation then actually even picks up from there. So there's an order of magnitude, more complexity of the SAP environment locally for any of these things. And then there's an order of magnitude on top of that of actually know how to code or operate these things. And so if you can kind of collapse that full stack, effectively, it seems like there's probably a good number of these plays that could be runnier where people that have been building stuff to be browser native by integrating really great agentic experiences into them

Nathan Labenz: (39:33) Yeah. It really helps. That model is incredible for sharing as well. I mean, it's the amount of times in the past where I've experienced like, okay, it's done, but where can I use it? You know, that is just an absolute nightmare. You know, it's like, well, it's on my local computer. Okay. But how do I use it? That, you know, well, I've got to deploy the staging. Oh God. You know, that that I felt that pain. Do you want to do a little demo or?

Eric Simons: (40:03) Sure. Yeah, I was just thinking. Maybe I can show you how this thing works real quick. So it's pretty cool. So I'm gonna go and pull this up. So if you go to bolt.new, which I love the domain, it's just the simplest thing to remember. So you go to bolt.new, we've got a free tier. So 1 of the cool ones, this thing is probably 1 of the best tools in the world for actually building beautiful looking user interfaces. So hopefully the demo gods play nice with you here. But we can just kind of punch in anything that we would want to make. So if I say, make me a music app that looks like Spotify. Let's go ahead and hit enter. This is like a pretty unbounded request I'm giving this thing. What you see here, this kind of looks like ChatGPT or something like that until this happens. And so on the right, this is what I was talking about with the core technology we've been building. This is actually running a full operating system here inside of my browser. It's already booted. It's already installed the dependencies for this project. So right now the AI is just writing out the code for this thing. I can just type in commands like it would. For those who are developers, mean, it's like a Unix compliant operating system we've built in WebAssembly that's booted inside the browser.

Nathan Labenz: (41:08) Made your own from the It's ground not like Ubuntu derived or anything like that?

Eric Simons: (41:13) No, we had to write it from the ground up. And the reason being is if you want it to be fast, if you want it to like boot in like a 100, there are things that will let you take a Docker image and convert it to WASM or whatever. But the image, the WASM file ends up being like a 100 megabytes or more. Runs 10 times

Nathan Labenz: (41:31) What's WASM? Like, I guess how much do browsers sort of anticipate or how much are they built to enable this? I know like the video story of this, there's like, in our case, it's like there's WebGL, you can use It's not necessarily meant to do 30 frames a second, but it can do stuff. Totally. What was that like on kind of creating a web container experience?

Eric Simons: (41:55) Hard. I mean, because the thing with browsers is that they're very strange with like what APIs, as you know from building on these things. It's not the sort of like really robust interfaces you typically got on your own computer if you were to build software there. And this is why you don't see a lot of these productivity tools that come to the web because if you want to do it right, you have to actually just start from scratch. And this is the same story of what happened with Figma. So back in 2012, my co founder and I actually had the good fortune of bumping into Dylan and becoming friends with him back then when he and Evan were starting Figma. And their first pitch for Figma was not a design tool. They were like, hey, we want to build a design tool, but what we have today is just this 3 d ball dropping into water. Let me see if I can actually find that. It's kind of iconic. Figma 3 d ball demo. So this is the demo. So Evan Wallace is the co founder of Figma. He made this demo of 3 d, this ball you can bring into 3 d water and see the actual ripples of it, right? This is back in 2012. And so this is a game changer, right? Because you couldn't do this in browsers before, but WebGL just landed in browsers, WebAssembly or the predecessor had just landed in browsers. And that's what made this possible. And so their pitch was design could not have come to the browser before, but now it can. Because look at what you can do. We can certainly build a 2 d rendering engine around this, around WebGL and WebAssembly, but we have to write it from scratch on top of these specific technologies. And that's what they spent the first year or 2 doing was just building a rendering engine that product could actually be built around. And so that for us was same deal when we went to build web container, which again is this, when you go to the code view here, this terminal, right? I mean, this is deceptively simple what you're looking at here, but we had to write something from scratch because if you want people expect a webpage to load in a second, that means that the size of this operating system is going to have to be like 2 megabytes, maybe a megabyte, and it'll have to boot in like 100 milliseconds. That is what we built here. And we had completely from the ground up and it took us years. We've been working on this thing for like 5 years at this point. There'll be 6 this year. But just a classic deep technology play that enables us to boot so fast. So if I hit refresh, let's say there's something wrong with this. Imagine if Gmail breaks, how fix do it? You hit the refresh button. That's again, the benefit of taking this sort of approach is if there's something wrong with that operating system, it just booted a brand new fresh 1, installed dependencies, spun up the server, and boom, we're going see the application in a second here that we just built. That's like, if you kind of compare this to anything else in the space, you're getting connected to a cloud VM. There's a lot of latency. If the thing gets broken, you have to go contact support and be like, Hey, you better, you have to hope that they have good support, blah, blah, blah, that sort of stuff. Right.

Nathan Labenz: (44:44) That is fascinating. Now that we're here and we've got this thing generated, first of all, did it just generate these images on the fly as well? Are you including?

Eric Simons: (44:53) Yeah, it's smart. So it'll go grab images that it thinks are gonna be relevant for you. And so these are the ones that it pulled to go But

Nathan Labenz: (44:59) these are not like the real album covers, are

Eric Simons: (45:02) No, I don't think so. Everything that we're pulling is royalty free sort of stuff or whatever the applicable licenses. But yeah, it'll go and grab good placeholders and you can tell it to use other stuff if you want. But this is like 0 shot. Mean, all I said was just make me a music app that looks like Spotify. And this is what it punched back, which is kind of incredible. That it's this good just on the first shot of the thing. And so like a lot of people, like I mentioned, a lot of folks are in the Fortune 500, often they're not saying, Hey, I want to build a full stack application from scratch. They're working on specific UI, whether it's for Chase's website for checking your banking details, or it's like on Netflix, clicking the next in the series button or whatever, right? And so this is like an incredible environment. Just be like, Hey, let me go hammer out some components that look gorgeous. But you can also continue to prompt on this thing. It's like, what would be a good example? Maybe I guess doing an audio file, you wouldn't be able to hear it on the stream. But let's say, I can just say, make the play button work and the seek track bar. Let's see what it does. But you can just keep prompting it. You just tell what you want it to do. And so you can, for example, if to we to make an actual music streaming application or something with this, we could go and say, Okay, here's the Spotify API, add OAuth and let me log in and play it through my Spotify account. And it's smart enough to actually just go and be to do that. And it's the sort of thing where you might get some errors back and you're like, we have a fixed error button you can collect. Maybe they'll pop up if something here goes wrong, or you can just go and debug it yourself if you're a developer or just Google it, etcetera. But it kind of the sky's the limit with the sort of things that you can build with this thing. I can show some of the stuff folks have built that's really sophisticated and full stack. So you can see here the play button is working. It does say there's an error. It actually is using the HTML5 audio API to handle. So that's the error thrown. It says, there's nothing to play. We would need to give it an audio file. It kind went beyond what I was expecting it to do here. So this is like, we actually have a tool that can actually play arbitrary audio files and this seat bar actually would work if we plugged it in and that sort of thing.

Nathan Labenz: (47:14) Yeah. I can imagine a, Suno or a Yudio. I don't know if those guys have APIs yet, but you're an API call away from generating your own music and kind of living a real hallucinated dream here.

Eric Simons: (47:29) I've seen people make stuff with I think Tsuno might have 1 because there's a guy who made a meditation app using Bolt and it goes to Tsuno and grabs it. It says, Make me some music that's Zen. And has you breathe in and breathe out with this. It's just crazy. I mean, it's like, when he tweeted, was like, I looked at the market cap of Calm and it was like, I think it's like $3,000,000,000 or something. I was like, I have his app pinned on my phone now. I'm like, Why? That's insane. Like, know, he built that in like 10 minutes. It's wild.

Nathan Labenz: (48:02) Yeah. That is disruption might be coming sooner than some expect.

Eric Simons: (48:08) Totally. And just like I said, maybe some examples of stuff that folks have made. This is actually made by a gal. She's a PM at a software banking company in Thailand. And so this is like her, she started this as her side project. But basically the idea with Viral Hooks, super cool idea, is effectively you can come here and sign up and this product helps you. I don't know if you ever had the inclination to be a viral TikToker or whatever. Yeah, I've never tried, but you've got to think about what it takes. You have to have a good hook to keep people watching. And so what she did with this is it actually reverse engineers the scripts of what popular creators do to make their hooks. And she made an AI agent that you kind of give it the content you wanna talk about and then it'll actually create viral hooks for your scripts. And so like a week before Bolt came out, she actually listed this project on Upwork. And I think the quote she got was from like a Ukrainian developer. Wanna say it was like 3 or 4 ks or something like that. And I think the timeline, it would take like 3 months or something in that range, which is pretty reasonable, kind of given the scope of the project we're talking here. And then the week after that, Boldt came out and she signed up for our $50 a month plan and she had the thing built and launched in 2 weeks with a full time job. So the ROI reduction there is crazy. I mean, it's like a 99% cost reduction. It's a tax faster delivery. And there's another case of this guy named Paul made this entire CRM using Bolt. And this has like an AI agent built into it. So I think I can zoom in here. You can actually like, out of it, if you ever use a CRM, these things suck. They're so boring to use. He has this AI agent built into this. So if you don't want to click through to do an action, Hey, I met with this guy. Can you log in? And it just does it. Like a pretty sophisticated application he made. It was the same deal he spent. Think he spent like $203,100 bucks on Bolt. The quote he got from an agency was $30. It was gonna take 6 months. And he had this done, I think it was in a month for this 1, which is it's just like wild. And so these are both like startups that have launched and are making money at this point, just like 2 tangible examples here. But yeah, I should mention both these folks are not technical, they're not coders. They made these things and completely without even having to tap developers to even help them finish it, right? Which made incredible products here. I got kicked out of 1 of free trial. I guess I'm logged in here. But I mean, this is real. It's like, I should buy this. I should support the Chelsea around. But anyways, this is kind of insane that you can actually, without any technical knowledge you can build real products like this at this point.

Nathan Labenz: (50:44) Yeah, it's pretty amazing. What does your pricing look like? Because you mentioned like a sort of monthly plan and then it sounded like there was sort of usage based pricing as well.

Eric Simons: (50:55) Yeah, totally. So this is actually 1 of the interesting things that I think we're kind the first to discover is a lot of people have followed too because it makes sense and it works. So a lot of these like cogen tools like GitHub Copilot or ChatGPT, they've had this thing where from the days of like, where you pay Netflix, like, Oh, should pay $20, kind of get all you can eat access. But if you eat too much, you have to slow down, we're gonna throttle you. And so what we ended up doing with BOLD is instead of trying to cram it into $20, we're like, Hey, you can just pay for as much inference as you need. And we gave people different plans. And so it starts at 20, it goes up to 200. Folks need even more than that. You can actually buy tokens or you have tokens, inference tokens. You can buy them as needed as well outside of your subscription. But you can kind of pick what makes sense to you. And what we see people do is they come in, they try the free tier, they love it, they run out of tokens, they upgrade to the $20 tier. Like, oh, this is good. This is building incredible stuff. It's totally worth $20. They burn through that. They keep upgrading, right? Because it's just it provides so much value. I mean, just for an engineer, 7 to selling it for $9. That 1 example. And folks are saving 99% compared to going with a freelancer contracting firm or whatever. When people start doing, they get to feel the math as they use the product. They just naturally upgrade to the higher tiers. It's like, this is totally worth the money, right? So that's for individuals. We also have team plans for organizations are buying it for their employees and that sort of thing. But again, it's all based on the number of tokens that you're using per month.

Nathan Labenz: (52:27) I guess a couple of questions there. 1 is if I'm doing my math right, your per token price is lower than the clawed SONNET 3.5 token price, which suggests you're probably not using either or you're counting on people to not use all their tokens in a significant way or you're using different models. It's just interesting when I was trying the product, I was like, wonder what model they're using. So how are you thinking about whether and, you know, in in, like, Cursor, for example, they do just, you know, to I can just choose my model. Right? So there's an interesting kind of different approach there where you are not making it clear to the user what model and you have to have sort of, you know, a pricing layer that is not fully transparent into even though they can see their usage, but they don't know exactly where it's going. How are you thinking about that? And, you know, can you tell us anything about what models you in fact are using under the hood?

Eric Simons: (53:16) We're super open about it. So yeah, use Solid 3.5 a lot. We're using other models too in various places. I appreciate the call out of like, Hey, this is cheaper than Cloud. Because people often are like, Man, this is so expensive. I'm just going to use Cloud directly. We're like, Go for it because it's going to be more expensive. It's like 1 of the benefits of, I think top 3 or 4 or 5 or something customers, right? Our upstream AI vendors is like, we can negotiate volume discounting on this stuff. And so that's like 1 of the benefits of having a subscription based model is that because we can project forward of like, hey, here's how much inference the user is effectively reserving. We can actually go and negotiate accordingly and say, hey, we know that we certainly need this much inference over the next months or whatever, right? So that's a huge advantage for us on a pricing perspective because we can price the thing aggressively. And 1 of the other kind of cool things about this is when people say, hey, this is a lot of money or whatever. And some people are like, This is so cheap. When you kind of understand the ROI, like, This is so cheap. For folks that are maybe just kind of learning how to build products, you're Wow, this is a lot of money. It's like, well, 1, you probably need to kind of learn how to ramp up on how to best control these AI things. But 2 is you can actually run Bolt locally. So this is a pretty unique aspect of our product versus pretty much everything else out there in this space or whatever. When we launched Bolt, we actually launched an open source version of it. If you go to, I think it's a Bolt. DIY, this takes you to our official repo for this. And so you can actually run Bolt using any LM. So you can choose the Gemini APIs, you can use OpenAI. This past week, DeepSeek landed in this. And so what's great is you can actually This is kind of like our testing bed of, Hey, what are the best models in the Bolt production application? What are the things that really show great promise, great results in our open source stuff? And we pull that forward and build it into the product. You can just plug in your own API keys and use this locally, etcetera. And I think it's gotten pretty wildly popular because again, there's not a lot of products, like a good AI agents. And certainly to the level of quality that we're talking here, most of this stuff is kept closed source. Because I think, what's your moat? I mean, for us, we've been building the core technology of the dev environments for 7 years. So it's going to take time for anyone that wants to build something through the fidelity and speed and accuracy we've got. But from our point of view, it's like there's only upside to put an open source version out there that people can build and fork on and improve. It's actually become a pretty popular thing that the AI labs use because they actually will go to Bolt. DIY to actually test out their new models to okay, how well does this perform in an actually real world application? Because a lot of these benchmarks for software engineering, the evals or whatever have you, it's like they're good to test against, you're not talking about a very dynamic test there, right? It's like a specific test case that it passes or fails or whatever have you. And so with our open source version of Bolt, they can actually come here and do things that are even a bit more qualitative. Like how good is the design that comes out of this stub 0 shot, right? And just how good is this an agentic experience of building a real application versus just a to do app that's not even styled? This is actually a very I think 1 of the Maybe counterintuitively, 1 of the big reasons that I think Bolt has kind of pulled away from a lot of the other stuff in the space is that there's a And you see this with a lot of popular open source projects. People end up consolidating around the things that are open source that they can fork and use and modify their own needs. And that ends up being a tie that lifts all boats and especially whatever products the company that's building the open source project are making. We see a lot of people try Bolt. DIY and then buy Bolt subscriptions, still use Bolt. DIY. So it's just a very symbiotic and cool thing that we're doing here because I don't see anyone else doing this.

Nathan Labenz: (57:17) I noticed the last commit was 9 minutes ago at the time that you loaded the page, which is couldn't have time that better. So what exactly am I I understand that when I go to the main web experience, all the code is loaded into my browser. I'm running this local container in the browser itself. What exactly am I downloading when I download DIY and am I still ultimately like using it in my browser or is it a different paradigm?

Eric Simons: (57:48) Yeah, great question. Yeah, what Bolt DIY does, it basically runs the open source variant of Bolt. New just on your local machine. When Bolt. DIY is doing the code streaming and execution, it's still using web container inside of your browser to do that. Really all Bolt. DIY is doing is letting you run that locally on your own device or on your own server or whatever have you. Basically, this is great if you want to have an offline experience running DeepSeek locally, You can actually use Bolt offline with Bolt DIY with DeepSeek locally connected directly in, etcetera. So we see a lot of enterprises starting to pick up Bolt DIY to run behind their VPNs and use whatever inference they've got behind there, etcetera, etcetera. But it's effectively just a way to host the web application of bolt.new and it's designed to easily point at any inference provider.

Nathan Labenz: (58:38) Gotcha. Okay. Yeah, that's cool. It is. And I don't recall seeing anything quite like that. I guess Cursor in a way has something similar where you can sort of subscribe or you can just, like, use their fork thing, you know, with your own tokens or whatever, but it's also it's always a local app. So it is it's definitely different too.

Eric Simons: (58:57) And it's closed source, which is, it's cool that they let you do that, but I think having it be open source just kind of A lot of people are interested in this stuff, but how do you actually contribute to something that actually has real adoption? Where you can actually see- So are all your

Nathan Labenz: (59:14) prompt templates and everything in there too? All the definitions of the happy paths people can go study, like how are you prompting it to make sure it's doing the Stripe integration correctly and all that kind of thing?

Eric Simons: (59:24) Yeah, we've got Let's see if I can find our We have our entire system prompt and everything in this guy. And we have some stuff that's in the bolt on news side, we haven't backported. Stuff we just might not backport, but we have been backporting a good amount of the stuff we are doing. Let me see, where is that? It's been a while since I've looked at this code base. My job as CEO sadly keeps me out of our code base for the most part. But somewhere in here, we've got the entire system prompt in this thing, which is kind of cool. Oh, there we go. Get system prompt. Okay. So let's pull this thing up. So prompt slash prompts. There we go. Yeah. So there it is. There's the whole shebang. This is like the main system prompt for us, like how Bolt actually works under the hood. There's pretty cool. Mean, and there's other prompts that are kind of chained and da da da and that sort of thing. But this is crazy useful. This represented some months of work for us with really smart folks building it. Where else are you going to find that? Unless you prompt leak 1 of the other providers or whoever that's like, this is forkable, runnable in the product. I imagine people have made PRs to this product.

Nathan Labenz: (1:00:32) So you have, I'm seeing here you have, it looks like partial support for Python. I didn't realize there was any support for Python. So it's like the node side, I can download random dependencies. On the Python side, I have like a sort of limited Python. I can't download it. I'm seeing here. I can't download third party, but I didn't realize there was any at all.

Eric Simons: (1:00:53) How about Yeah, we have some basic support for Python. So like the Node. Js ecosystem has, We've worked with that 1 for half a decade to make pretty much all the major tool chains run-in WebAssembly and that sort of thing. Python is not the beginning of the trail, but just it's a couple steps down the trail ahead of making everything work in Wassom. And so I believe if I actually go back here, I can probably in the terminal, let's see if we've got this a name. If I type Python, we should get back, I think the REPL or whatever. So we've got a version of Python that's running in WebAssembly here that we can kind of execute Python code in. I think we have some basic standard libraries and stuff baked into this thing. I know it's on our docket to get PIP added in. I think for a lot of the use cases we're seeing though, when we were in IDE, it mattered a lot more to support every language. That was our 10 year roadmap or whatever. And we've always unabashedly been focused on web applications. That's kind of been the point of The key insight was browsers don't have a way to build web apps. The web should be able to build the web. That's like if you look at every other platform, Mac has Xcode and Windows has Visual Studio. There's no tool like that built into browsers. That was the impetus. So we've always been focused on Node. Js and the JavaScript based ecosystem. So we've got Python. I think there's a couple other ones too. So the entire WordPress and PHP ecosystem now actually runs in ModAssembly too. So you can actually run WordPress in this thing. I think Laravel runs in it. And so I think that's the number 2 ecosystem behind Node. Js as far as compatibility is the PHP 1 in WebAssembly and WebContainers at this point. But I imagine, I think by the end of the decade, I think most of the major languages and ecosystems are going be running in WebAssembly. Yeah, we'll

Nathan Labenz: (1:02:37) just ask the super intelligence to handle that for us.

Eric Simons: (1:02:41) Totally, right? That's a lot of the issue right now is it's just how do you get the engineering hours to kind of really drill in and make that happen? Yeah. Agents.

Nathan Labenz: (1:02:50) Yeah. And as the market I mean, it's another difference between developers and everybody else. Developers care what they use. My dad doesn't even want to know. You know, he just wants to see something work. And in the ideal state for him, he's never going to even consider the code or what language it's written in. Best product experience for him. He literally wouldn't be able to tell you at the end what language things were ultimately coded in.

Eric Simons: (1:03:15) Totally. Well, I think that's what's interesting about the intersection of where we're at versus I think, I maybe Replit is somewhat ish in this space. I think they lean a bit more developed, like heavy developing than we are. But we're kind of at this perfect intersection of providing a great service developers and providing a great service to non developers, especially because we made the core of this thing open source, people are contributing to it. People want to see more stuff working in it, right? And developers, there's developers, want Python to work in this. There's Python developers like, I want Python. And we get this all the time. Like, I want Python apps. I want this. Want that. And so people are making poor requests. They're making the stuff work, right? And on the flip side of that, kind of upstream and even what you see working in Bold today is the result of the past 5 years of the tool chain that's running here is called Veet. That ecosystem said this running inside of WebAssembly and Web Computers is important to us because we want to be able to have these in browser environments. Let's make it a reality and they commit to it. And that's what's making this possible, right? The developers care. And so it's this 2 sided ecosystem where without developers caring, you can't enable all these people that you have never programmed before to actually leverage these tools. And so I think that's cool is we're right in the middle of both as developers care and they love our products, so do non developers. And so we're able to make this magic happen where again, you don't really see that happening in other products or tools or ecosystems.

Nathan Labenz: (1:04:46) Yeah. That's cool. I'm learning a lot from that. I really had no idea how far the in browser containerized world had come and just how many things you could do in it nor do I even realize that even from trying the app which I definitely did spend some time trying to make something myself but I didn't get that deep into

Eric Simons: (1:05:03) Good. Yes. I mean, that's the point. Anyone should be able to use this and be completely unaware. That means we did our job right.

Nathan Labenz: (1:05:10) What are some of the other kind of foundational technology pieces that you are tapping into? For example, you you mentioned Stripe, you can see Supabase in the UI there. I saw a mention of Vercel AI SDK.

Eric Simons: (1:05:26) Oh yeah.

Nathan Labenz: (1:05:27) You know, how did, like some of those are, I guess, fairly obvious choices at this point. They've certainly, you know, Stripe has certainly emerged as like a standard, but would you, could you run down kind of the list of the opinionated decisions that you've made and why?

Eric Simons: (1:05:44) Yeah. We've got, if you're using, there's 1 track that we're opinionated on, and this is really more for people that are not developers, who want have a smooth path where they're going to not run into a lot of errors. It's going to be really great. Even for developers that are just looking to quickly build a product or something, right? So kind of in that path, it's like Veat and React, which we've been for the past 5, 6 years we've been supporting. And like I've mentioned them before. So that's kind of your dev tooling and the framework or whatever. For auth and database and webhooks and that sort of thing, Subbase provides us. You can click this connect Subbase button. You can create a project or have you and just spin up an actual database for this thing, etcetera, hopefully. I don't know why this thing maybe my Internet's lagging. But So as you can have a database with auth, a way to actually create rows, etcetera, with the Postgres data is backing it. With that, you can actually plug in Stripes. You can accept billing, which you saw on the chilled CRM sites. This is powered by Stripe using it for the subscriptions there and that sort of thing. And the final thing that's kind of cool about Boldt let me go refresh the page and get back to refresh refresh page. This is the beautiful, wonderful thing. If something goes wrong, usually you can get back to refresh the page. Or maybe not. I might have borked my dev account before we came on here. But kind of the cool thing about Bolt is that, again, because we're running an operating system in the browser here, like you mentioned this before, actually, I didn't actually answer your question. You're like, how do you actually deploy this somewhere? Because that's often a challenge. Like, hey, you made something cool. And then how do you actually get it onto a URL and do something useful with an actually have a domain? We actually have a built in integration with Netlify, so you don't even have to sign into Netlify to do this. It's just in every Bolt project, you can just click this deploy button. And I'll flip back to the code view because this is cool. So this is actually running a production build, again, using my CPU. Normally, you spin up some cloud, you know, CI box or whatever, build a production website, deployed it live on Netlify, that's the URL, and I can share this with you. And if I wanna add my own domain right now, if I wanna connect this to a Git repo or whatever, I can click on this link, which will actually take me to Netlify and attach to my account. So I can myspotifyclone.com, it points directly at this thing. And anytime I deploy here and out, so if I want to make the backgrounds neon or something, we'll see what it does. As I make changes to this thing, it'll go ahead and update those. And once it's done making these changes, I'll make sure it's good. I'm gonna actually be able to hit the deploy button again. It'll run another build. It'll put it right back onto that URL. So if we had a production.com domain pointed at this thing, it would update live with that, right? So it completely removes the entire deployment part of the question for this experience, which again is like usually 1 that can be rather challenging to do.

Nathan Labenz: (1:08:34) Yeah. Deployment is such a nightmare. I'm very excited about things that make deployment easy and Replaid, which you mentioned, was 1 of the first things that I saw that kind of changed my paradigm there, but this is doing something similarly

Eric Simons: (1:08:48) cool, And which is the Replic guys did a great job where like they've baked deployment into the product, right? Which is I think for like certain types of apps is great. Like, I think what's nice about us just having a partnership with an actual deployment platform that's kind of what they do as a business. It's really reliable. You can attach domains to it. You can actually build real products and teams around it. And that's not to say that what having an in house deployment solution, what Replitude isn't useful because I think they've done a really good job there. But a lot of people want to actually host something that's got a real domain and that sort of thing. So going to kick off another deploy here. This is what it currently looks like for reference. Let this thing go ahead and do a production build again. So go ahead and do that. And if I go ahead and refresh, boom, you see our neon colors, right? So it's insanely simple to get this thing back live. But I think that's probably 1 of the most magic parts of the Bolt experience is just having a 1 click way to get this thing live on a real prod URL.

Nathan Labenz: (1:09:56) Yeah, that's cool. What's coming next? I mean, candidate ideas that I would be interested in your takes on. Obviously, O3 minutei should be coming soon, allegedly going to be another significant jump in coding ability. Agents in general, of course are like, you know, big buzzword, but when I've been doing this sort of AI accelerated app development, I've definitely noticed that a lot of my time is spent just copying and pasting stuff around and even more is like testing the flow that I just had the thing code. So I'm starting to kind of wish for like a cloud computer use to code it, or now we've got OpenAI operator, which is obviously unlimited, you know, access at the moment, but I do pay the $200 a month. So I had the chance to use it. I kind of wonder, like, are you expecting to sort of create like a an AI user, an AI tester? Because that that seems like it would be the next biggest chunk of time for me probably that would get taken out is if I could zoom out 1 level and go from today, I go, okay, what is my feature that I want to do? I often take that to R1 Pro cause again, I pay for everything and then, you know, have that do a sort of an analysis and kind of plan, make sure that my like overall roadmap for a feature is good. I'll usually put my full code base in there if it fits. And then it's like, okay, now I've got a step by step plan. Now I can take that to, you know, the the coding agent. Most of my experience recently has been with Cursor. That usually works pretty well, but then I'm like, okay, did it work really well or not? I I have to go find the edge cases. So I kind of wish I could zoom out 1 level up and be like, here's my thing. You go do it. And also you test it and, like, bang on it 5 times and fix the errors that come up. I feel like I'm already often 5x faster to do things, maybe in some cases 10x faster than I used to be. And if I could do that, I feel like I would be like consistently more than 10x and maybe genuinely getting into like 20 plus range for some scenarios.

Eric Simons: (1:12:02) Totally. I mean, I think the company that's kind of doing what you described is like Devon, right? Where I was talking with 1 of my buddies, Sean Wang, you might know, or you might at least see him on Twitter or whatever. He's like really smart guy and he's he's been Yeah, doing a lot Swix. Yeah, yeah, sorry.

Nathan Labenz: (1:12:17) Refer to I

Eric Simons: (1:12:17) him Yeah, by the thing that people know him yeah, yeah. Know, always call him Shauna just because I met him in person first before online, I guess. So everyone calls him Swix, but awesome guy. And so he's like an angel in the company and a good friend of mine. And so he and sat down a couple of weeks ago and we were just kind of chatting about kind of this, of like kind of the breakdown of different products in the AI software engineer space and how they're oriented. And I think at least my takeaway, and I think Sean would probably agree with it, I think was like, I think the problem with things like Devon, which is not knocking their pockets, I think for certain cases, I'm sure it's great. But I think the issue is when you kind of tell an agent to go and do something, and it's gonna do something that's like going to take a while. That's going to do a lot of things during that while. Therefore, it's going to spend a lot of inference and whatever during that while. A lot of things can just go wrong. And depending on how, at what stage of the process things went wrong and kind of cascade and then wrong doesn't necessarily mean, oh, an error happened, but it's like, got confused or it misinterpreted directions and dah, dah, dah, dah. And so I think that's kind of the challenge candidly for agents, think is when you want to start baking stuff on and it's going to go and do a whole bunch of things, the odds of that collapsing increases the more steps you add into it. Cause right now, things are just not the fidelity. There's not the fidelity that you would need to really send the thing off into the woods to go and do a meaningfully sized set of tasks. Doesn't mean that it can't work in some capacity, but it's like I'm Bolt. Coming back to your earlier question, what's the success rate? And I think from what I understand from what the more longer running agentic things like Devon, success rate is pretty low of like where people are like thumbs up, I'm super happy with what the result was, right? Whereas when you flip to something like Bolt, there's a lot more people, a lot more, which gets to the point where the ARR just continues to ramp because they're seeing immediate results. If the thing does make a mistake, it didn't go off for 20 minutes burning inference. You're just making a chain of decisions that are from the root of not what you want or whatever, right? So I think having kind of human in the loop where it's, they're kind of guiding the thing and verifying that it's not screwing stuff up, etcetera, is right now, I think we're tangibly, you can really squeeze an incredible amount of value out of it. And honestly, I'm a little hesitant, not We're doing R and D on the kind of the longer running stuff, but it's like I'm a little hesitant to really try and pull off of that just because I think humans are just now and for the foreseeable, it's gonna be an important part of the creation process. So it's kind of longer answer to your question. So as far as kind of adding some level of automated testing, there's certainly stuff that can be done around, hey, have the thing right unit tests upfront. So it's like, as it's building, it's ensuring that it's aligning with that. But again, I think part of the trick is, how do you write unit tests? If the users are not technical, how do you ensure that those unit tests actually are good? That they're actually going to be representative of what the end user wanted? That's the big question, I think, around this stuff. That's kind of a longer answer, but that's kind of like my current worldview on just, okay, how do you improve these things generally? And what are kind of the limitations that pragmatically and immediately we see. But we're investing kind of across the board in those areas at least.

Nathan Labenz: (1:15:52) Yeah. I think that's really interesting. In preparing for this, I lined up 4 different products and basically tried to build the same app with all. And they were Bolt, Replit, which I've had the most experience with, Devin, and Lovable. And I would say Devin definitely stood out as being the most different experience from the others in terms of just the the paradigm. And it is for the reason that we're just discussing where it sort of just keeps going. And I feel like I do want that, but I don't wanna watch it. And at least as it was at the time that I tried it, which was, you know, 10 days ago or whatever. So these things are changing fast. It was, you know, with the other 3 products when I came back, literally just like rotating through them and being like, what's the status of this 1? What's my next instruction tab? And, you know, so I'm like the outer loop, you know, agent of each 1 is kind of doing its inner loop. What I found was a little weird that the Devon experience was like, I didn't know where I was, you know, when I came back. So if I had watched it, I would have known where I was, but I didn't want to watch it. Of course, I wouldn't do other things. That's, you know, 1 of the big promises of this. It does feel to me like there is a way to solve that. Like, if you imagine a good, you know, intern or junior person who's helping you with a project like this, the piece that I feel is kind of miss, I do want them to take that initiative. I do want you to test it before you bring it to me. So that's definitely an expectation for a human developer I'm working with, right? Like don't bring it back to me, even having tried the happy path. I can expect that there could be some edge cases that you haven't sanded down. But if you haven't been down the fairway and confirmed that it works, you're not doing your job. At the same time, you know, I want like a summary of that. I want to kind of know, not in a, like, I have to go back and read through your experiment logs, but I need you to tell me like, this is where I'm at. I did this. I tested it this way. I confirmed this is working. I got a question here And that sort of interaction layer, I don't think has been nailed, but I do think I want that like several extra but I don't think I haven't seen anything yet that sort of brings me that human like experience of like, here's kind of the update, boss, you know, on on everything that happened since last time you were here. And then they'd almost need to trigger that on, like, refocusing of the browser tab, really, or some sort of signal that, like, now is the time when the user's back for an update. Because I was coming in and it was just kind of midstream and it's working and I'm like, I don't know where you are, where you've been, but I do think that will get, you know, with time, people will will improve that. And I do think I'm definitely willing to wait longer and pay more inference tokens if I can get just like bigger chunks of things done.

Eric Simons: (1:18:40) It's definitely going to happen, right? Mean, I think it's And that's kind of the thing that we kicked off with the pricing model where it's like, you can buy more. Was a result of, I mean, Sonnet 3.5 basically gotten AI codes into the point where it's good enough where the more money you kind of put into it, you can actually have some realistic expectation of value that it's going to actually be able to give you back on it. Before that, that was not the case. With the previous kind of frontier models, we tried building Bolt a year ago, like in February '4, with the Frontier models at that time, and they just couldn't do it. The code was too unreliable. The design output was ugly, if it even worked. But I think as these things just improve 100%, I want too. Because I do this all the time with my own team where I'm like, Hey, can you go ship da da da? And they're brilliant folks. So I've worked with them for a while, they kind of have a sense of what I'm looking for or whatever, and they'll come back with, Hey, here's kind of where I'm at or whatever. But if you can deliver on that, oh my gosh. Yeah, I mean, that's worth every penny and way more actually. And I think the trick is really just the accuracy of the user experience. I don't think we're far. I feel like we're probably a model or 2 release away where the Fidelity is at a point where you can meaningfully start doing that for at least some types of workloads. You can have longer running agentic processes that you can actually trust to actually do what you want them to do. I don't think it's far, but it's an important part. I think it's definitely like, there's a lot of value, I think, that'll be created from that end of things, you know?

Nathan Labenz: (1:20:20) Yeah. It's a fun, it's a tricky balance for you and all these people taking different approaches to this because it's like to varying degrees, and it sounds like there was a time when you really needed to turn up some business metrics, but everybody wants to deliver value to customers and grow today and have momentum. You want that positive feedback that's going to make everything better from including hiring and just whatever. So you need to be sort of in the sweet spot, but you also need to be getting ahead because we know that like the sweet spot is definitely moving. That's definitely been a challenge, you know, that I've experienced again, firsthand. It's like, how do you do both? I don't know. That actually might be a great closing question. That's about all the questions I had, but I would be interested to hear how you think about balancing the today versus tomorrow. And then you could throw in anything else you want to, if there's if you want to put a bad signal out for hiring purposes or anything like that, we'll do Yeah, that

Eric Simons: (1:21:14) dude, it's a good question. I mean, I think we're pretty laser focused on just making our AI agent incredible today. We've got irons in the fire on a handful of kind of R and D stuff. I mean, I think it's, you know, There's a very clear line of sight for our product being able to enabling folks to build incredible software. It's already happening today, and it just keeps getting better, right? Even with the current state of the models. And so I think that's primarily where our focus is. And there's a part of this too where there's just tsunami waves that keep heading every 6 to 12 months with these models. And there were some products that were in this space. I think there's a couple of them. Like Replit and mean, Replit v0 and Lovewell. Lovewell had rebranded, relaunched or whatever. But all these guys were kind of around before Bolt existed 6 months ago or 9 months ago, whatever. And when Sonnet 3.5 came out, those product experiences, they had to completely reinvent them because all the work they were doing on those agents for previous models just didn't matter anymore. And in some cases, just I mean, they stripped, pivoted their product offerings. So I kind of look at that and I look towards the future and it's like, how do you ride tsunami waves into the shore and then paddle back out and catch the next 1 versus getting wiped out by it? Think it's a delicate balance. Just like pro surfers, it's not an easy thing to do. It takes having a realistic perspective of, hey, what's the probabilities of how this is going to land and what the next 1 could be and aligning yourself to catch the thing, right? And sometimes you just can't, of course. So I think that's kind of how internally we kind of view that, how we hedge the bets here as we look towards the future, just because we've kind of seen the result of 1 or 2 of these things at this point, how that's kind of transformed how people's products were oriented. So we're pretty open minded about what will happen. Yeah, and to that end, yeah, like we're hiring. Yes, we just announced we raised a series B like a week ago. So we just raised $105,500,000 and we are looking to aggressively expand our team. We've scaled to tens of millions of dollars of revenue and we have like over 2,000,000 registered users. We have a team of 20. We are severely understaffed for the demand we are seeing. I think last time I checked, have, I don't know, like 50,000, 60,000 customers, something like that active. So we are hiring aggressively. So if you go to stackblitz.com/careers, we'd love to hear from you. I think we're working on some pretty cool stuff. And anyone interested in AI and the web and just enabling both developers and people who've never coded in their life to build great software, we'd love to hear from you and work with you.

Nathan Labenz: (1:24:06) Cool. Well, has been excellent. I really appreciate the time and the product demo and some of the deep tech stuff that I didn't know previously. And you guys are on an unbelievable trajectory. So definitely check out, what was the career space again? It's bolt.

Eric Simons: (1:24:22) Yes, stat. If you go to bolt.new, a link at the top that says like we're hiring team, click that. Or you could just go to stackblitz.com/careers. That's like the direct link to the thing.

Nathan Labenz: (1:24:31) Alrighty, cool. Again, this has been amazing. Eric Simons, founder and CEO at Stackblitz, makers of bolt. New. 20,000,000 ARR in just the first 8 weeks, a $100,000,000 raised, and an exciting future in front of you guys.

Eric Simons: (1:24:45) Yeah. Thanks for having me.

Nathan Labenz: (1:24:46) Thank you for being part of the cognitive revolution. It is both energizing and enlightening to hear why people listen and learn what they value about the show. So please don't hesitate to reach out via email at tcr@turpentine.co, or you can DM me on the social media platform of your choice.