Bolt’s Eric Simons on Enabling Everyone to Generate Websites with AI

0:05Hi, listeners. Welcome back to KnowPriors. Today, we're hanging out with Eric Simons, the co-founder of StackBlitz and the makers of Bolt. A new AI tool that enables everyone, from developers to designers to non-technical folks, to build full-stack, real applications entirely in their browser. Eric has spent the last decade and a half thinking about how to make development more accessible. And since its launch, Bolt has taken off like lightning. Is it over for site builders? We'll talk about AI code generation, creative community, and if everyone really wants to build websites. Eric, good to see you.

0:37Good to see you too, Sarah. You have had a wild two months since you guys launched Bolt.new. Can you explain what it is? Like zero to 20 million of ARR? I don't think ever seen that sort of crazy growth. Yeah, I haven't either. It's been kind of surreal. It's kind of far beyond any of our expectations here. So for anybody who hasn't seen it yet, what is Bolt? Bolt is, it's kind of similar to like Chacham to your cloud, except you use Bolt to build full stack web applications. So you can come and just prompt if you want a landing page, a blog, or even like any type of full stack web app where you have authentication, you can log in.

1:20And you can use it effectively instead of going to like a web development agency or shop. You can come here, put in your idea, hit enter, and get a real production website for you. If you look at the world, there's 25 million developers, I think globally. And today, last week, we had almost 200 ,000 software composers, we like to call them, that use Bolt to build web applications. And we think that that number should be 100 million. And we're on this growth clip. It seems like maybe we'll get there sooner or later. Bolt is really enabling folks to build real software, not just kind of drag and drop sort of static sites, the previous era of how the web was made.

2:02There's a lot of code generation tools out there. You can do this directly in the core model products as well. What do you think people are finding special about Bolt? Yeah, totally. Yeah, what's special about Bold, and it kind of comes to the origins of our company, but in short, we've written an operating system in WebAssembly that can run in your browser. And that's really important because if you want to run dev environments, you need to be able to install arbitrary packages and run different tool chains, whether it's Next.js or VEED or anything else. It's very complicated and expensive to typically do this if you're going to use servers.

2:36So it's very valuable to do it in the browser because it's extremely fast. There's no latency. You're not paying by the minute for some cloud. What we've done has kind of married these frontier models with this technology we've been making. And when you kind of look at the other stuff in the market, there's, you know, like a cloud artifacts is, you know, probably one of the first things that hit the market that did a really good job of this, where you could say, hey, build me a UI and it will like do it. The problem comes when you actually want to build stuff that's more meaningful. Like, it's very good if you're saying, hey, like, yeah, I use a, you know, cloud, you know, every week for just kind of generating graphs based on numbers or whatever.

3:09Very good for that sort of use case. But if you want to say, hey, create a landing page where people can log in and do some type of functionality, you can't go NPM install Firebase or Superbase or whatever have you and plug all that up and actually deploy it. That's what Bolt specifically is uniquely capable of doing without any other setup. It's just all kind of baked. A common engineer, investor, tech person pushback. Hey, like these code generation tools are often the same. You guys have this web container technology that allows you to, you know, abstract away the back end and allow that to run locally without handling that developer environment mess yourself.

3:51There's this concept of like a GPT wrapper company, right? And so I think there were a number of companies that were less generously like some system prompts and like a well SEOed website. You guys open sourced your system prompts. So, and like a lot of the code for Bolt. Can you explain that strategically? As we were building Bolt, yeah, I mean, over the past couple of years, there's been a ton of more simple sort of wrappers that have come out, you know, around these frontier models. And the problem is like whenever the next model comes, whenever, you know, one of the AI labs eventually integrates that into their chat products or whatever, those companies, you know, kind of tend to go away pretty quickly.

4:35So when we were working on Bolt, I mean, one of the big advantages we have is we've been building this web container technology for five years and it's like pretty, pretty difficult stuff to do. And so when we were going to build, to actually launch Bolt, when we were building the system prompts to kind of use your interface around it, when we looked out there, there's not a lot of other folks doing this where they actually could open source their system prompts and kind of open source their products so you could see how it's actually made. And for us, you know, we just, we felt that it was inevitable that someone would get our AI model to dump out our system prompts anyways.

5:12But also we felt there's kind of something missing in the open source world where we see, like we come from a web developer platform, right? Open source is key for innovation for everyone. And to date, a lot of these companies in the AI space have been looking at their system prompts and kind of their specific glue code as the secret sauce. And it just seems there's a whole lot that has been left on the table by not just putting this stuff out there and letting people fork it and improve it and contribute to it. If we're going to build a great business here, what's going to allow us to win is growing extremely quickly, building the best end-to-end product experience that really works incredibly well.

5:51It's not going to be the system problem. For those who have done web development, it's kind of like view sourcing on a webpage. You can go to google.com, you can view source. No one so far has kind of done that and built another trillion dollar company that took out Google by doing that, right? You can learn a lot, but it's not actually, you're kind of building this cohesive end-to-end product business experience is a totally different thing. Why is the community valuable to you? I mean, you've always been like very committed to the developer community and to open source, but have you learned anything from the community that improves, as you said, the end-to-end StackBlitz system or is there like an ongoing way that happens?

6:29100%, yeah. This has actually been one of the most interesting things for, I mean, StackBlitz, the company, we've been investing in open source a ton over the past, you know, five, seven years or whatever, have you. When we, when we put Bolt out in open source, we were really curious to see, does anyone find this valuable? You know? And the answer was like, yeah, actually. And there's kind of a couple of, a couple of key things kind of worth calling out on this. You know, one from a general community standpoint for, I would say just AI tools in general. One of the biggest problems that a company like us has, that's building an AI experience with AI models and heck, even the AI labs themselves, is educating folks on how to best use the tool.

7:09Because the problem with AI models is that they're non-deterministic. It's not like instructing someone, hey, here's how you send an email in Gmail, you hit the compose button, you type, you hit send. If you go to a chat product, prompt engineering is something that folks have to be educated on and learn how to use properly because it costs money every time you send a message. And so one of the interesting things I think we've done a good job of is really investing in the community and having it be a place where folks are sharing their knowledge on how to best use the tool. Because that's actually, you're going to see a lot of churn in your product if folks are coming, can't figure out how to use it and leave.

7:46And we're finding, we are learning, we are actually not even the experts on how to use our own product at this point, are power users. They actually know more. And so we're bringing them on live streams to actually show, hey, what's working for you in your workflows? And to the degree that we can upstream what they're doing by typing prompts into the product, we're doing that. So I think if you're building an AI tool, it's critical to be building out a community and actually be directly engaging with them and giving them a place to share their knowledge because otherwise, everyone else on the product, there's going to be a high amount of churn, which is exactly what you've seen happening with these AI apps.

8:24churn on these things can be like 60, 70 % for folks that are not doing what I'm describing. And that's kind of like, that seems to be the most common case with a lot of AI applications that are kind of in the earlier stages. It's just crazy how churn rate. Oh, I was just going to draw a parallel to what David Holtz had done and the MidJourney team overall in that I think that the fact that there are like obsessive, creative, really capable power users of MidJourney that are teaching the entire community how to use MidJourney models and demonstrating like what can be created is like a huge part of their position versus like the many other image generators that are out there.

9:08And so that I think is just an interesting parallel. Yeah, I think that's an excellent example and probably the best example that I've seen. There's surprisingly not a ton of, at least very visible examples of this, but I think it's going to be extremely important. I think for the companies, especially the startups, they really want to win big here. This is, I mean, just a critical community, more than ever, right? More than ever, I would say, community strategy and real investment is going to be key for success in building this type of product and user base. You guys are what, like 10, 12 people right now?

9:45Yeah, like 15 to 20. I don't know what the exact number is, but. 15. Okay, to 20. That's better. I'm a little out of date. I think the, like, the idea of like, let's have 200 ,000 plus users and growing be using this every week and tell us what works in a world where, like, there aren't evals that are, you know, from academia or standard that are useful in terms of like what real world applications can you build with your system, right? And so I think that virtuous cycle seems really powerful. And you guys also, like I think a big part of the Bolt theory is like make anyone a developer, right?

10:31Versus you guys are all developers. You also need to see how non-technical humans build with this stuff. Yeah, 100%. And that's like the majority of people using Bolt at this point are non-technical. And it's interesting, the most successful people that are using Bolt are actually people that have had to interact with or manage development teams. So think like entrepreneurs, PMs, et cetera, because it turns out kind of managing an AI is extremely similar to managing actual software developers, right? And one of the things you just mentioned on the eval. So this is actually the second piece regarding why we open source.

11:10And one of the most interesting aspects that's popped out of it is that, like I mentioned earlier, there's not a lot of good open source AI tools today, like real world AI tools. And especially ones where the products that clearly are providing a lot of value to the degree they're growing quickly and both by revenue and usage. More like maybe one of the only of the few or something like that. What's kind of happened is that, as you mentioned, the eval suites were very good the past couple of years of kind of generally measuring how good are these models at coding. But the problem that folks are kind of running into now, when you talk about building a real world product around these things, the eval suites that exist today are very specific and not representative of like, hey, I want to go build a landing page or I want to build XYZ.

11:55There's nothing in there in those suites that you can actually test against. And so what's ended up happening, this is what's going on with our open source version of Bolt, is Bolt Local is becoming kind of like one of the main ways people are testing out new code gen models when they're coming out. Like I think there's one that was released by NVIDIA recently, and there's one called Quorum that was released recently. And some of the folks, I think over at Hugging Face, have been basically just one of the first things they're doing is taking Bolt Local and dropping that in, saying, okay, how good is this thing versus Sonnet 3.5 or whatever, right?

12:30And so that's been kind of an interesting, it reminds me for those that were like into video games in like the 2000s, if you, you know, can it run crisis, right? Like as far as measuring your PC performance, like it's kind of becoming that people are, you know, asking like, okay, can this run bold? You know, like how well can this run bold? They drop in like, okay, you know, yeah, can't really do da da da, you know? So I think that's to me, one of the most interesting things that's going on as far as the open source side in the community there for us. And we've got some stuff we're announcing on that end that's going to kind of further bolster this.

13:04But already there's kind of some benchmarks being set up around this thing where the latest AI models have a way to actually get tested in a real product that's actually pretty sophisticated with use cases that are not just, hey, can it write hello world? Yeah, yeah. Or this very well-specified SWE bench problem. I think it would be useful just to paint a picture of like some of your favorite use cases of things people have built that you feel like are real. Because that is definitely, I mean, correct me if you feel differently. That is a new development this year that you can use any sort of code generation tool to get to a useful application in any sort of end-to-end way.

13:50100%. 100%. Yeah. What's your favorite use case? What are people building that's cool? What's really cool to me is folks are actually able to build real world products. And so we've been online for just under two months now, and we've already had the first startups launch out of this thing. They've used Bolt to build their startup and are making money, like charging on Stripe or whatever have you. So a couple of examples just off the top of my head. One is from this gal in Thailand. She's a PM in a software banking company. and her company is ViralHooks.ai. And so she launched this project, you know, by herself on the side, just moonlighting it.

14:30And the product is actually pretty cool. So the general idea is, you know, when you make like a TikTok or something, I'm not a TikToker, but you know, I've had aspirations. When you make a TikTok, you need to have like a viral hook to kind of get people to keep watching, right? And so she's actually trained up some models, you know, for OpenAI or whatever, or have you to actually help you write great viral hooks for your videos and kind of reverse engineer how other great creators have done that. So you can go check it out, viralhooks.ai. And so what was kind of mind-blowing, and it's a beautiful site, like awesome product.

15:03And what was mind-blowing about this is a week before we launched Bolt, she went on to Upwork and listed this project, said, hey, I want to build this product, da-da-da, asked for quotes on the thing. She got a quote for$5 ,000 from, I think it was Devon, like Ukraine or something. Estimated timeline, two to three months, right? It seems like, considering the app I just described, kind of reasonable. Sounds cheap, honestly, for what it is. Yeah, it's like a pretty, like not a bad price and not an unrealistic timeframe. And the next week, Bolt came out. She signed up for our$50 plan. And in two weeks, she had built and launched the entire thing.

15:40The cost savings there, I mean, it's a 99 % reduction in cost from$5 ,000 to$50 ,000. and then a five times faster delivery, you know, two weeks versus, you know, months. And the alpha is just like insane, you know? And this is not, it's not actually like a one-off case. Like she was, I think the first person we had chatted with that had done this end to end, another guy named Paul, he launched an entire CRM called Chilled CRM. He's just been on a tear making a ton of different types of tools. But like, this is like fully featured CRM, like calendar, contacts. He has a chatbot built into the thing, you know, an AI chatbot, et cetera.

16:15Um, same deal. Like he actually, he's been running web dev agencies for, I think, like 20 years. And so, you know, to, to build the CRM he made, it was like a$30 ,000 quote. Um, he did that on our$200 a month plan, you know, in one month. Right. So again, same sort of cost savings, et cetera. So I think a lot of, a lot of folks, uh, you know, especially in the web dev shops, et cetera, they're just, they're able to punch out incredible web applications faster than ever before for clients and are able to charge the same price. Right. And so, yeah, there's one tweet I saw online where one of these folks was like, this is the most incredible arbitrage opportunity in web development ever.

16:57And it's true. I mean, it's unbelievable. So I think what's really cool is just seeing people be able to take their ideas, launch them into reality for a fraction of the cost way faster than ever before. I have, for a number of reasons, been long-term skeptical of like no-code tools in the traditional sense, right? Like a GUI-based editor for people to build simple applications or more complex applications in a closed ecosystem. I'm just like, ah, like for anybody who's coming from engineering, that's really scary because I'm basically trapped in your platform without the ability to leverage the entire developer ecosystem, the open source world, like frameworks, anything we might need, because I don't know where the bounds of your system is.

17:52Like, how do you, you clearly believe that there's some version, I mean, it is working, but like there's some version of no code and development for non-developers that is going to happen. Like, what changes? Like, why should it work? Yeah, good point. I mean, and to be clear, like, you know, like six months ago, I shared the same viewpoint. And, you know, and maybe even like three months ago, I would have shared it. But there's some key things that have meaningfully changed. Just from a technical perspective, AI cogen models, there's a tipping point. Specifically with Sonnet 3.5, there's a chasm that's been crossed here.

18:27As far as AI models have gone over the tipping point of being good enough to really write real applications that are like production grid. And it's only going to get better from here. But up because, you know, kind of some inside baseball, earlier this year, I think like in February, we had the idea for Bolt. We tried to build it with some of the, you know, frontier models available at that time. Wasn't possible to do. The models just did not spit back quality, accurate code that was constantly breaking, which ruins the experience. It doesn't work, right? So we put the project on the shelf, and then once we kind of got an early preview of the new Sonnet stuff, we're like, wow, this is okay.

19:06This changes everything. And so I think, you know, if you kind of think about these no-code site builder things that have existed to date. The only reason that these existed, they had to make custom WissyWig GUIs and stuff is because how else can you get an end user that's not technical to turn their idea into code? The best middleman tool or interface to do that to date was like drag and drop Wix style sort of stuff, which comes with all the problems you just mentioned, lock-in, how do you expand this? Like, how do you, you know, it's like you want to actually do, add real development to this at some point.

19:44How do you do that? And so these things end up as these walled ecosystems that can't really get mainlined into building real stuff on top of over time. That changes now because of this tipping point in the AI models. Now the best interface... I mean, we have people coming to Bolt that... With the week we launched, we had a salesman, I think from Dallas, that he tweeted us and said, thank you so much for making this tool. I used this to make a website for my daughter because she has a medical condition. She has to find donors as she travels. And so I made this website for her so she can send it ahead of her travels.

20:25And it was an incredibly touching use case. But my first thought was like, should I tell this guy that Wix exists? There's other things that can do this. And then I realized Wix and Squarespace are really complicated to use. The only time I've used Squarespace was to build my wedding website back in 2021. At first, I wanted to do it myself. I'm a developer. And so I was like, honey, this is important to me. I need to build this thing. And I spent the Saturday on it and it wasn't done. And I'm running a startup at the same time. I have other things going on. So she finally just bought an account and said, make it in this thing.

21:02And it's pretty complicated. Whereas you compare that against both, it's a text box. You say, hi, I'm having my wedding on this date. Here's the details, dot, dot, dot, dot. Here's the RSVP. hit enter boom zero shot there's there's a production website ready for you right and like to degree my my 71 year old mom built and launched her first website ever two weeks ago yeah and so it just kind of goes to show like it's way simpler right to build a real and then the stuff that's being the code that's being punched out is is uh you know the same stuff that developers would work with it's like next js or remix or astro or vid or whatever have you So, and what's actually happening right now in the community is as folks are trying to do more and more complicated stuff, they're raising their hand in our Discord or on Twitter and they're saying, hey, is there anyone who can like come and help me debug this or build this thing out?

21:53And folks are like by the hour saying, hey, yeah, book a time with me and I'll come and help you develop this thing, et cetera. So it's kind of this really beautiful mix of the best of both worlds that's happening, right? So I think that to me is what's changed. What's changed is like AI code gen has gotten good enough where you can go and take your ideas, put them into your fingertips, hit enter, get a great result. And for things where you need to bring in actual professional developers to tidy up or fix bugs or really expand, you know, more difficult capabilities, they can because it's like any other code base that they come into, right?

22:28So I think it's a very, very interesting and kind of mind blowing point in time because I don't think anyone saw this coming, you know, years ago. Yeah, I find that very inspiring because I think that there are plenty of entrepreneurs or even just individuals who like are not one of the 25 million professional software developers in the world, but want to make software or have a web presence of some kind. and uh and this is the first time they can do it in a way that like to me makes sense where i'm like okay if you if you succeed or you lots of engine actual like existing professional engineers use bolt too but it is because it's not a dead end right because like you can go you know iteratively do development or even use the the stack blitz ecosystem of developers or whatever over time which is which i think is like step function different yeah and for existing days i mean it's like this is just like every other developer tool or innovation you know the past 20 30 years this is just allowing them to focus on the actual high value work that they do you know it's just it's it's kind of not worth their time to punch out a ui you know um and so that's that's what they're coming here to do is just rapidly iterate on on uis and pull in data etc and um and some folks some developers just using this as a primary way to launch their startups or whatever or if they need to pull it into Cursor, they do that and they can bring it back to bold.

23:52They can kind of use the best of both worlds. But certainly for non-technical people, this is huge. That was an interesting thing that we learned actually is there was a large number of people that have been downloading Cursor that are not developers because it let them meaningfully dip their toes into kind of like clicking accept change, accept change, accept change from the AI. and when we first launched Bold, I mean, there's still coming on all these YouTube videos that are like, Cursor and Bold, it kills Cursor and we're like, they're two different products, you know? But to non-technical people, they aren't.

24:29It's actually like, this thing solves the problem of me not being able to code and et cetera. Maybe we can back up a little bit and just talk about like stacklets and the story as a company. Yeah. I think it's very funny when companies suddenly have overnight success right because it's like oh well they were you know notion was working on developing their point of view and trying different ideas to refine it on for five years and then they made notion um and you guys have been working on this for you know five plus years as well like uh can you talk a little bit about like the the origin of stack blitz and you and pi and like when uh when you decided to do both yeah so i i co-founded stack blitz with one of my childhood best friends his His name's Albert Pai.

25:12He and I grew up in a settler of Chicago together. And when we were 13, we had ideas. We and I were always very interested in computers. We were building PCs. And we wanted to learn how to write web applications. This is like the mid-2000s. So like for our 13th birthdays, we asked for, you know, the O 'Reilly books because they're like 200 bucks a pop instead of an Xbox, you know, and then we learned how to code together. And really, it was painful. I mean, at that time, there was not like Code Academy and all the stuff that's for free online. There wasn't really online communities around these things.

25:43But he and I really wanted to, we thought we had cool ideas for products or whatever. And we really wanted to build them and launch them. I think that's really, I think, that's why we've been building stuff together for 15, 20 years. It's been about that. Coding was really a way to, a necessary part of how you bring these things to life. Anyways, fast forward, Albert and I have done a couple of different startups over the years. But back in 2016, 2017, we had this realization that browsers had gotten really powerful. We've been building web apps at that point for a decade and a half or so. And we had this realization that browsers had gotten really powerful.

26:25And it had hit this new inflection point where you could actually basically run an operating system in a browser tab that was really fast, etc. And that was really cool because that means that you could actually use the web to build the web. If you look at every other platform that's ever existed, that's been an important capability of every platform that's ever succeeded in a meaningful way. Windows can build Windows apps. Macs can build Mac apps. The web does not have a built-in way to do that. So there's kind of this nerd instinct of ours where we're like, this is important. This seems very valuable to solve, right?