Browserbase's Growth Engines

Welcome. We've got Paul, CEO, founder of Browserbase, here to talk to us about code to market. You know Hank, you know Gonto, and right before we started rolling, you two were fighting about SEO. You say it's fake. You say it's real. What was the thing you said that started this?

My controversial take is that I don't think that doing SEO for Google and thinking in humans make as much sense as it used to be.

And I disagree. I think SEO is the same thing as S-A-I-O. You know, like if you're putting out great content with good keywords that people are linking to, it's going to get pulled into the context of the model. Now, there's probably some middle ground between the two. I don't think we throw away your old school SEO strategies of putting out good content with keywords that rank. You probably need to make it a little bit more ingestible and it can't be as keyword stuff as it was before. Like the training data has to matter. But I still think if you're doing SEO best practices, it's still a positive thing. We don't need to throw away that rule book just to go make AI read our content.

What's the difference in the playbook though? Because I'm kind of neutral here. I would have sided with you, but I also know you're doing experiments on this. So do you have any experiments that are doing stuff?

We do. And I think like, first of all, I do agree that if you have content that is good quality, like it will work for sure. But like a couple of things that I think are different is like number one, now I don't care anymore about Google. It's all Bing because all of like ChatGPT, Claude, they all use Bing API and not Google. And then before I think the only thing that matters was being on the first and second position in Google because who will click anything else? But now ChatGPT and Claude read the first 12 results. So then I care about being in more keywords in the first 12 results rather than being in just a few in the first or second. And then I don't know, I've been doing a lot of experiments lately. Like we saw that if you have llms.txt, it works much better than having the sitemap. We're trying with questions, we're using Scrunch, like it's like a SEMrush for AI, SEO. And for example, if you put the questions in H2 or H3, it works better. If you have bullets instead of pros, it works better. So some of the things like bullets versus pros, it's the opposite to Google. Again, if you write really good quality content, it's still going to help. But as similar with SEO, I think you have to push it a bit more to be shown with ChatGPT and similar.

You know, it's funny, 10% of our signups come from AI where AI had recommended somebody use Browserbase, which for being 18 months old as a company, it's happened much sooner than I thought, considering how infrequently they run these training data, these training collections. So I think there's something there where it's like, I still always will just focus on quality. And I think if you're trying to squeeze the last juice out of lemon, then you're like, okay, you'll put your questions into the H2, not in the bullets. That seems like a more micro-optimization for me. But if you're a much larger company, you have a ton of content. But for startups that are under 10 million in revenue, just focusing on quality, you don't have much ammo to put behind all this stuff. I think you still can get really good results. And what's most important is that other people talk about you. If you can get outside of your domain, other domains, other blog posts on the Hacker News comments, like in the discourse, that really helps more content explode and more people hear about you, both through AI or SEO or also word of mouth. So in the end, there's so many ways to code to market. And I think you have to just try every strategy and see what works really well for your company.

I agree. I love the idea of like mentioning like one thing we did with one of our customers at HyperDX was starting to do deals with companies like, hey, if you do a list of the top 10 companies that do blah and others, we'll do the same and add you. And some of those cross references then help a lot to actually be shown.

I'm not excited for just the new wave of spam AI, SEO stuff. It's just like 2005 again.

One of the companies I work with, because we didn't want people to read that shitty content or good enough content, we actually created a path that was hidden blog. So the idea is like it's not linked from anywhere except for the bots, for the LMs.

It's interesting because like maybe going off of like these new marketing tactics, like I think being polarizing is really important right now, right? You know, especially as social media algorithms, if you are going on LinkedIn, X, Twitter, you have to be, I think there's a lot of incentive to be polarizing. I think in our space and developer tools, like being polarizing can actually be really bad. You don't want to be seen as like the really polarizing startup. I think developers kind of react negatively to that, but you also need to thread that line where people are engaging with your content, both positively and negatively. And I think we've done a great job with that at Browserbase. I think that we've like pushed the limit a little bit on like what you expect from a dev tools and infrastructure company, like what we do, but also not gone too far where we alienate our audience. I think I'll reference Cluely, which I think is an amazing marketing and copying company. Like if you read the manifesto, it's some of the best copy I've ever read in my entire life. Roy's great. But I think if Roy was in dev tools and Cluely was selling the dev tools, it wouldn't have had the same effect. I mean, you look at the developers who engage us and they really, really hate it. And that's part of their strategy. But when you're polarizing, you have to be polarizing in a way that doesn't alienate your core audience. And I think that requires a lot of deep customer empathy for what, in our case, developers really care about. And we've thought about that a lot. And I always try and push the limit a little bit. You know, if you're having a few haters in the comments, it means you're doing something right, but never too far where we completely destroy our audience for views in clout.

That's never our goal. What's been your role in that? Like as a CEO, what has been your role in setting like how you show up to the world on X on thinking other marketing strategy? Like how involved are you? What things do you care about and what things you don't care about?

Web, before we say that, just on the debate, I'm going to side with Paul just because he used code to market in a sentence. Yeah, we never asked you. Well,

that's how I built allies, right? Back to Gonto's question. What's your role in the marketing and go to market? Like how do you view that? How has it changed?

Yeah, I recently read Lee from Vercel's exit blog posts and he had an amazing run of that company. And he talked about how like he reviewed every single tweet for every single launch. And I thought that was like really impressive for a company that size. And I personally don't review every single tweet that goes out. We have like a go to market content review panel. But in the end, I think by being the probably the largest poster at Browserbased, people learn from that and they learn what the content is good or bad. And I very rarely had to give feedback and say, ah, this isn't really our brand. You know, let's take it down because I think when you set a good example and you get results as the founder CEO, that's, you know, putting content out there, I think people will try and learn from that. I think where I've probably spent the most time reviewing stuff is blog posts. And my team hates me for this because I want to have amazing blog posts. I want to spend weeks on them. And we just don't have the time to do that. Right. So I think I've had to let go of my blog quality a little bit.

I didn't hack for that. What was that? A couple threads in here was like one. Yeah. Like when we were a small company at Vercel, we had a committee. It was like me, Leerob, our product marketer and Rauch G like three out of four of us had to approve a tweet or anything in writing. And that was back when the company, you know, was going from 30 to 400. And then after I left, obviously like Lee was that guy. But for blog posts, I just do that deep review on the first two paragraphs because that's what most people read. They don't scroll. And then once I'm satisfied with that and the intro, I'll glance at the rest and then I'm like, I'm not going to be as nitpicky here for the sake of time.

I look at Fly.io as like a company has amazing blog posts and they built this. If you see a Fly.io post on Hacker News, it's almost always been the other front page. They built a brand of really strong blog posts and I'm sure they've had some misses, but maybe I'm just not aware of them because of the greatest hits are always surfaced. Well, it's so challenging to build a multi-channel marketing engine at a small early-stage company. I think it's actually a big misstep. It's like you want to have the best blog posts and the best video content and the best tweets and memes and the best LinkedIn. Like I've actually felt very intentional. We're like, we're just going to be focused on X. We're going to do a lot of great content. We're going to also have blog posts, but like we're gonna have one focused channel and really spend all our energy fine-tuning that. And as you get larger, as you grow from 30 to 50 people, you start to add supplemental channels where you start to get mastery. But I think in the early days, our single focus on a channel has been really effective for us and we picked the right channel for developers, which happens to be Twitter.

Did you chat with the team about it? Like was it a conscious decision? Like, hey, we're going to focus mostly on X. Like did you share it with the team? Did you work on it or it's more like implicit? And everybody was thinking about that.

Well, I think I was the team. I'm a solo founder, right? So we had a committee meeting and it was just me, myself and I, and I said, well, I'm going to do Twitter. That's where I had some following. And I think I was, I spent enough time on it where I feel like I knew how to post in the right way that would engage people. I think from that, we grew the team. A lot of the team has Twitter accounts that they use. And you know, if you look on Browserbase who we follow, there's, you know, 10, thousands of followers in there. So I think it was like the team was structured around this and we also just have learned how to master that channel. And that channel is nice because we can all engage with it together. Over time, we're going to do more. I really am excited for us to do more YouTube. I feel like YouTube was a really undiscovered channel for developers and especially for us. And the few YouTube things we've done have always gone pretty well. So that's something where we're going to have to get, you know, more.

What are your, because when you started Browserbase, your Twitter following wasn't that big,

right? Whoa. 2000 people. That was, it was, yeah, yeah. That's right.

We were about the same back then. Yeah. And now you're like 12 K. Yeah. Which is like,

you know, you're getting up there. I'm a micro, micro celebrity.

I don't want this to sound like condescending or anything, but like that's good growth. We've talked about this before Browserbases ability to pierce through the noise of Twitter, especially in the AI space is very impressive. So I'm curious, like, can you give us more of those tactics? Like what are you actually doing that like does that? And I'll give you like some things I've observed are like, you'll take a bet on a really high production quality video and you'll bundle lots of things or like any content that's, you know, normally you would make that content for the sake of that content channel. You're making it for the sake of Twitter is my guess. And just to amplify that, but like, are my observations accurate? What else are you doing?

Yeah. I mean, I could need to mentally break down the four biggest moments of our history, which have been fundraising announcements. Like we raised four times in 16 months, right? And each time we did a raise, we did a video or piece of content. And if you look at the video we did relative to the time, it was very much in the zeitgeist. Like we kind of caught each trend at the peak. So for a pre-seed video, it was me standing in our tiny little office saying my manifesto, there was an American flag behind me. And then there was a whiteboard that said secret plan in the back where if you zoom in, you can see the secret plan. And that was very much like, you know, early 2024, it was like founder led, like the founder videos worked as big back then, but when they were, they were very personified. People follow people, not brands. It wasn't really about the brand is about me and kind of establishment of me and Browserbase being kind of synonymous at the same brand. And it was very much like high level mission. And we hired people from that video. Early customers found us to get like a hundred K views. That was the pre-seed. Then the seed launch was a clip factory. It was like, there was a clip from iRobot, then a clip with Roombas with knives on top of them. And then it was like, you know, I think the hook was, we thought AI was going to save the world and all it does is vacuum for us. You know, like it was very much like Antimetal kind of did a good job of this, where they had that very much clippy video for a seed round and then had the Coachella lineup investors at the end. Right? So like that clip style, which now other people have started using later on, I think it's fallen out of fashion, but during the moment, like clippy, TikTok, brain rot type fundraising announcement was very big. And then for the series A, we did our first type of film. And we were actually a little earlier on this than, than others, but we had one of our, you know, film people, Nico come up and we hire an actor and we shot this video and it was like the next billion dollar company will be built on Browserbase. And our goal was to inspire. And we showed this kind of a narrative journey of a person building a startup on Browserbase. And at the end, they get a phone call from YC and they get in. Right. So our first experiment at film, which was really, really well received. And then for the series B, we took that even further. Okay. Let's go full production value. Let's go all in. We had a team of like 20 people doing the video and it was a separate style thing where, Hey, if AI has automated my work, it's kind of like my any in severance. And I, my Audi is benefiting from that getting promotions while my AI is doing the thing. And I think like each of those videos and themes have really followed a trend at the right time. What you want to avoid is actually being cliche. Like I would never do a clip video, like for any sort of announcement now, like that's so dead now. I mean, I actually think films are going to be dead probably in the next three months or so. Like I think it'll be kind of tired to see, you know, film YC companies doing films means that it's already kind of over. Where did you get the ideas

from? Like, did you see others and then like you saw very fast and then you copied them or what are you actually one of the first that were doing them and what did they come like?

Was it you? Yeah. I mean like, I think there's certainly inspiration, right? I don't think it's blatant copying, but I also saw that those channels worked. I understood why, like why did a series of clips from films and TV shows work really well during like mid 2024 what's very attention getting like there's like, there's so much video happening. You want something that's going to like tackle the ADHD brain. I'm just going to pay attention to it. And then people got kind of sick of that and they started preferring more like well-produced videos. And I think now that's actually going away. I think if I was to predict a trend, I think what Warp just did, and I think Kalshi did this as well, like the full AI generated video that looks really good for these ads. Like Warp did a GTA 5 like video that looks like GTA 5, but they produced it. Kalshi did like, what would you have bet that the Romans would take, you know, the city and it's like showing Romans via AI. I thought like that's pretty creative and I think we'll see a ton of that next. I love the Kalshi one. I think more and more. Yeah. And I think I bet I guarantee, you know, in the next 90 days, we're going to see a ton more of those. Right. And there's this neat

thing being in B2B, you can kind of see what's working in B2C and copy the styles a little bit. Like you can be a little slower in B2B and beat 99% of the other B2Bs.

Well, let me, let me add one asterisk there. I think if you're just copying the style, it's always going to go worse. What you have to do is understand why is that style working? And I think for us, like, so what is great about this? Like, why do I like this launch video? Like, I think of one of my favorite videos that have ever come out is when Deno announced, you know, their video. And it was so good because they were like, hi, I'm Sosa, I'm the creator of Deno. And it was like this like founder-lit video, which was funny. And I remember it because I remember someone plugged in a flash drive and it flew through a wall and it was just like, so, so memorable. And I think that worked well because there were so many micro viral moments within the video that you can like latch onto and repost and say, you know, every piece of content has to have like an LOL, WTF or OMG reaction. Your Buzzfeed reactions. Yeah. You got to have the Buzzfeed ones. When do you think

of all these things? Like I saw the same videos as you. I think you're terminally online like me, like all the time online looking at things. When do you start reflecting all the things that you see? Is it when you go running? Is it when you're in the computer reading the tweets? Like how much time do you spend a day on Twitter just looking at these things?

Yeah, probably too much time. I don't like having an active reflection loop. Part of my job as founder and CEO of this company is to understand what's happening in our space. And that means I have to stay aware of like all the things that are happening. And I think over time I have a team that now brings stuff to me too. They say, this is really cool. And we go over, why was this cool? And try and understand those meanings behind it. And that's actually a great teaching moment where I can try and share more with the team. There's this kind of thing that maybe it's just unique to me. I'm sure a lot of people do it where like if you consume enough things and you're consuming them actively, there's a passive reflection where I'll see something early and I'll be able to say, oh, this is going to go really viral. And like if I kind of start checking back on that and I can predict yes or no, then I'm like, okay, I'm attuned to what this audience or this segment of people really like. And I really try and practice. I see, I sort my tweets by newest and I try and guess sometimes like, is this going to go really well? Thanks. Never really go that well unfortunately, but we'll work on it together. I love that spreadsheet. I actually used to

use always the following like ordered by when they were published, but then eventually I moved to for you because I think as a marketer for the tools, it's important to understand what things eventually went vital instead of just looking at the following. But your idea of looking at following and guessing, I think it's absolutely incredible. So a quick

principle for the audience. If you want your kids to be a founder, don't limit their screen

time. The iPad babies are going to be the next YC mafia. They're going to be the best.

Do you have tweets that you spend an hour writing? No. Like how, you never,

like fast Twitter or slow Twitter? We have both. My worst tweets are the ones I took a lot of time thinking about. My best tweets have always come out in like five to 10 minutes. Like

that's opposite of a lot of people we talked to. Like we talked with Colin from clerk.

It was like I spent an hour writing a tweet, but then I met with levels ago and he was telling me like, I just do a brain dump. Even if it's poorly written, like I don't care.

I'll post it. Yeah. I mean there's, it really just depends on the topic. Like I think if I look at some of those viral tweets I've had, I had one where we talked about building a healthy grind culture at Browserbase. I took a picture of our office and said like, oh, we're working on a Saturday. And people were replying like, oh, why are you working on Saturday? And then I quote tweeted the next morning while I was like having coffee and saying like, this is what's important to me about this topic. I think when it comes authentically out people in speech think quite a bit more. As soon as somebody sniffs a little bit of inauthenticity in your content, it just becomes really hard to engage with it. I don't want to engage with someone who's just like faking it on X. Like when I see someone putting themselves out there vulnerably, I think it's much more compelling. You cheer for the man in the arena, right? And when the man in the arena has a ton of body armor on, it's a little bit less exciting to cheer for. So I think you got to put some vulnerable,

you're not perfect content out there as much as possible. Authenticity I think is absolutely key.

I think something interesting about that though is I, so I think for a lot of people when they write in a context like email, especially they're inauthentic, their default is to become overly formal or wordy or whatever. And I think a lot of that carries over into social media and whatever. It sounds like you're the opposite of that. And so that's like, that's an interesting thing I think for people just to know and recognize about themselves. It's like, Hey, it's the first thing that comes to mind in this context, the authentic thing that I should just send, or is it the overly formal, you know, like dear Twitter followers, you know, and you should hash it over and really think it because there are two different types of

people that way. I've seen that's different on that sense. I think a lot about people who are speak to think or think to speak like there's people who to think about something they will speak with somebody and then rumble and then they will get to something and there's others who are thinking solely on and eventually they'll speak. I think that people who are think to speak are more formal, maybe spend an hour. The people who are more speak to think that have to talk to others to help to idea. Maybe the ones could like just put

something in one minute and then they'll see what happens. I think a lot of the desire to be perfect and copy comes from a fear of what happens if it's wrong. Like what if someone says you're dumb or that's a stupid day or whatever. I miss me all the time. They like to say okay, cool. Yeah, you know, like I think over preparing content comes from a fear of like what if someone doesn't like this and as long as you're like authentically okay with what you're posting and like I like what I, everything I post I like and believe. Like I try, I would never post and I don't believe in truly because then if someone rejects it or if I get like ratioed to hell, I can stand by that piece of content and say like, oh, well I really believe this and it gives me a little bit more protection from all of that. I will say on speaking authentically, all of my emails I've been writing with Whisper Flow recently. I just said that's a tool I really love. I love Whisper Flow. So like I just hit that function button. I just yap at my email. It really actually the emails come back really positive. Oh wow, that sounded so authentic. I do this for all of our customer refunds. Like when someone cancels, I send them an email and I try and just, I just sit there and I go through like 50 of them. I just like Whisper Flow a message about like, hey, sorry, it's gonna work out. Thanks for the feedback, whatever. And the response rates are extremely high because it just sounds like I actually wrote an email and people

connect with that. You talked a lot about launch events. You talk about like your pre-seed, your seed, your A, your B, how important are launches in the past and how important do you think our launch is going to be in the future as part of your marketing strategy?

Yeah, well they certainly can't all be fundraising events anymore. We, we, I think we're, we're certainly going to try and move more towards launching actual products and relaunching products. I don't think I'm a fan of launch weeks. I feel like that's too much. I like launch days, like singular days. With our series we launched, we launched like 25 different things, but people remembered the new product we launched and the, you know, the fundraising that we did. We did Stagehand Python, we did organization, we did all this stuff. So I actually think we may have launched too much on one day there. And there's a balance that we're still trying to strike in terms of how much we launch. But I do think trying to focus your energy around a single thing, you can only get the world to care about you for a few minutes at a time. And as much as you can, I really try and focus our, our laser beams on a singular day or moment, because that's when we can get our investors to support us, our friends, our employees, you know, our customers, they get an email that day. So I do love the singular day launch. I do love accompanying it with like a little growth hack. You know, for the series A, we had a plane fly over the city with a browser as banner, cheaper than a billboard. I'll let you know. And what's pretty funny is like, I don't think anybody really saw the plane. I think you saw this, but people were tweeting about it. Like probably like 50 people tweeted about it. We flew it over OpenAI's office and it lapsed, you know, and it was just like, okay, what happens here? Like let's always do something zany or something different with a launch to just try and find some interesting alpha in, you know, trying to pierce the noise, get in front of people. Exactly.

I love this point of like, okay, it's a launch day because we can get everybody to talk about us at that specific point in time. But something I think a lot about is sometimes like you do a launch, you get some users to come, but then you come back to baseline and it's the same number of people. How do you leverage the people who come through a launch day so then they can invite other users or you can use that frame for the future? Because otherwise it's not a big baseline, big baseline. Yeah. If you have a launch and come back to baseline,

you have a bad product. We have launches and they come back down of course, but it's never back to baseline. And that's because you don't have product market fit. Like you should be able to add additional users who will talk about your product and then, you know, continue to, you know, refer people to your products. So I think that if you have a launch and it comes back to baseline, you need to reevaluate your retention, your growth loops, like the usefulness of the product or even like the size of the market. I think it's like a hard reality to face, but it is genuinely true. You should launch something and see for the next several months, some increase in growth in aggregate. And for us, that's always been the case. I think it does. If you're a more mature company and everyone's already heard about you maybe doing a big launch and that will come back down to baseline, but that means you have to launch newer products. You need to have more reasons for people to re-engage with you. Growth does decay, right? And you do have to keep pushing stuff out every single month, every single week. You have to have some piece of content, but the product, product led growth, the product does lead to growth. And that's going to be the thing that people talk about. I don't really want people to talk about Browserbased for being, Oh my God, they have an amazing brand and marketing motion. I want them to talk about it saying they have a really good product that's important for our needs. And you can delude yourself a little bit and think too much about the marketing and brand. It's a very important part, but the product and the thing you're actually building that people pay money for, that is the core of the business. And I make sure that that's really, really good. And everything we launch is kind of, you know, in supplement of that.

If you do launches and then the baseline is higher, which I agree with, it's a good point. Why would you do blog posts in the meantime? Why not focus all of the energy in a launch?

You have to have every channel. Ideally you're doing all of it, right? It's not like a net negative to do a great blog post. I think what's really hard for small teams to do all these things well, and that's where you have to ruthlessly prioritize. You have to be realistic that like, okay, do I want to have a single person spend a week writing amazing blog posts? That might go top of HN or do I want a single person helping run the shoot to create a great video and pull the content together for the launch of this product on X? Like, okay, do both. I would love to do both. You don't have a team that says to do both yet. We will at one point. And I think these larger companies who have really great marketing teams are really smart about dividing their effort across channels and mediums to make sure they're saturating everything quite well. But for smaller companies, you have to pick and choose. So for us, we've swung towards a video world. We'll probably swing back to more blogging marketing and more content there. We're doing more clips on LinkedIn, trying to grow LinkedIn. And what I really actually love the most, our best performing content that we don't do enough of, and we will do more of it. I wouldn't do more of it personally is building things like one of my biggest browse ways tweets. I built a script that would go pay for your parking tickets in San Francisco. And I found some license plate on Reddit that had like 25 K in parking tickets. And I use that in the demo video. And I was like, my friend has 25 K in parking tickets. I am going to pay for their tickets using this script. That's like a WTF 25 K in parking tickets and like an OMG, like they just use the script to do it. And that did really well. And it also helped do the thing that is most important to us right now at Browserbase, we're building a new category. We're a category creator, browser infrastructure. And that really wasn't as big of a thing before us and our whole purpose of our company is to help AI browse the web, right? But when you're a category creator, you have to inspire people to use your product. So much more of our content in 2025 is going towards, how do we inspire the developer? How do we inspire them to use this thing? How do we help them say, you can build a feature that's going to help automate work on behalf of your customers using Browserbases infrastructure. And that inspiration has really shifted the way we're doing things because we want to show, not tell a lot more. And that's just going to change how we put all of our content out for the rest of the year.

I think a lot of goals, I don't even know the word in English guerrilla campaigns for channels where nobody's using them and then you get like a very high up, but then eventually they die. So I think when one channel works, which for you, I think it's the launch events. You have to be focusing on your second channel because eventually the launch will stop working. I think everyone's, everybody's doing videos, everybody's doing launch. Maybe there's another way to grab attention or another thing to do. Do you think about that? Do you worry about like, what will happen when my main channel dies? Like, am I thinking about the

next? We have so many channels. Actually, if I maybe I'll send you a screenshot of this later for the podcast. If you look at our signup attribution, it is very evenly split across like seven different channels. I guess it's like 20% YouTube, 20% Twitter, 20% LinkedIn, 20% search. And then like 10% I heard from, you know, AI and then the majority of the rest is word of mouth. So like when I look at my distribution of signup attribution, it's actually extremely healthy and shows that we are going multi-channel. Now I don't talk about like, Hey, we did all the Theo ads. I love Theo. He's an amazing creator. I love supporting him. Right? So like that has had positive ROI. Like we've done YouTube sponsorships of creators and that has a great ROI. Josh on your team at Largo. That's why I want to snipe from you guys one of these days. He's so good. Right? Like the YouTube comic creators are so, so good. And maybe I don't talk about like, you don't see the thousand little things that we do sometimes, but our growth team like Lindsey Gilson is so good at making sure we're covering all of our bases and continually to put stuff out in my mind. If our signup distribution was all just X, it'd be in serious trouble because it is healthfully split up.

It works out quite well for us. I'm a big believer on it's better to be different than better. You just mentioned the plane, which I love. I saw all of the tweets and everything. What other things have you done that are creative or unique or different? And do you think that doing something creative and unique is better than doing the same as the rest

but slightly better? I mean, something I'm very passionate about is like no dark mode on Browserbase. Our website does not have a dark mode and our dashboard will eventually I'll cave to that, but I actually think like having a dark mode website is so cliche in developer tools. Like it just comes off the same as everything else. And for our branding, it's very much inspired by consumer. The agency that did our branding Herf Paris, they're a consumer branding company. And I think that like the interplay between consumer and developer tools has always been quite interesting to me. When you're building a developer tool, you have to build something as easy to use as Snapchat, but as reliable as Boeing or maybe Airbus these days, right? Boeing's an interesting choice. Hopefully more reliable than that, right? It's such a tricky balance because there's so many developer tools. So you have to have your onboarding adoption curve being very short, but you really need to make something sophisticated for engineers that have all these use cases. And I think that I look to consumer companies in terms of branding because they create brands that are sticky. And I think our browser is branding is very sticky. We use red as our primary color. Now people generally advise against that. They say red is bad for your CTA conversion. Our conversion is fantastic on the browser's website. And I think that like when you listen to the traditional thing, like you need to have a dark mode for developers, don't use red in your CTA. It's like you need to have a contact sales button somewhere. Like I think more like when you listen to all the advice, you become one of the averages. When you take a first principles approach to like what is important to you as a company, you get to have a lot more uniqueness and it can pierce the noise. As you said earlier. Yeah. We talked

about being different is better than being better a lot. And like it comes down to even the smallest thing, like it's been known forever. The best time to send an email is Thursday morning, but that's when everybody sends emails now. So like you can't, you can't do that in the same vein. Are there things in your product? Cause you know, as you and other people have been building out products, are there things where you see people and you go, Oh, we just need to copy that feature. We need to add it. We need to have parody or the opposite where you say, Oh, they have a huge gap here. Like we can really differentiate

if we double down. I really don't worry about competitors that much these days. There's a few reasons why, but for the most part, I've never seen a competitor out innovate us. And if they have launched something that I thought was cool, it was, it's been on our roadmap for awhile and we haven't done it yet. So I think we'll launch things that may look like, Oh, they're copying so-and-so, but actually just we prioritize things differently and they probably pushed something up because we didn't have it yet. But generally, no, I haven't seen any of a competitor that's smaller than us launched something that I didn't expect for us to build at some point, which is pretty sure to me, like you need to be, to be an incumbent in a category where you have to be an innovator. And I've always spent a lot of mental calories on what does that product going back to roadmap look like and prioritization is the hardest game in startups. Like why build this now versus later? What can you put on? What is critical versus what is nice to have? And then when I think about like larger competitors, I think about like when developers want to buy, you know, a dev tool, it's kind of, I always use this analogy, like buying a dev tool. It's kind of like buying a jacket. When last thing you got a jacket, you probably went to a store. You didn't just like look in the window and say, yeah, that jacket has sleeves pockets. It's red. I'm just gonna go buy it. You try the thing off, right? You put the jacket on how the sleeves feel. The same thing is true with adopting a new tool. Someone wants to get it into their ID. They want to try using it. They're going to try a little POC project. And we spent a lot of time making that as simple as possible, such that when you sign up for Browserbase, the first one we dump you in is a code editor in, you know, our playground environment and you click run and you see the code control a browser and you're like, Oh, that's how this thing works. And you're playing with it. So for us, I'm really thinking about like, how can we make it so that people understand and get to POC very, very quickly. And then I'm also thinking about how do we keep innovating to make sure that if a competitor is launched in that we haven't thought of before, I'll be really surprised,

but I'm still for that. It sounds like you've built in in your blood and sorry, I have another question. With Director and Stagehand. Well, with Director specifically, do you feel like that's going for a new or different audience? Does it broaden who can use your product and like, how are you thinking about going multi-product and how it affects your audiences? Because we've been talking a lot about that with, you know, just a lot of other

tools that I just won't list. And adding to that to the question, like from the outside, I feel that you're sort of following like the Vercel recipe that Director is like your v0 Stagehand is sort of your next and then you cover the product. Do you feel that relation? Like,

have you thought about it or not really? Yeah, I mean, I use that analogy all the time. And I think it's a great analogy because it's a good strategy, right? So, you know, maybe I'll talk through the products and like how they work together and why we're doing it. Right. So Browserbased, we're running many browsers in the Claude for developers to control with code. So headless browser infrastructure, headless browser is one that runs on the server. That is our Claude infrastructure. That's where we charge our money. You know, you pay for the info you use, right? Now, if we want people to use our headless browsers in the Claude, we may need to make it really easy to use them. Stagehand is our open source framework for controlling a browser. So you can use natural language to say like, hey, on this page, click the buy now button and it will generate the code to control the browser to click that button. Right. So open source framework, MIT license, a million downloads per month. Companies like Intuit, CrowdStrike, Coinbase use Stagehand to automate the web. Right. But when you're building a framework, so we built the browser infra and then the framework to make it so it's easier to use the infra. But when you're building a framework in the age of AI, nobody writes code anymore. Right. You need to have something to generate the framework code. Well, that's where Director comes in. Director is easiest way to generate web automation scripts. You can go in and say, hey, we're doing my passport and it's going to go like, go to the passport website. And as it's going, it's generating Stagehand code that runs on the Browserbased infrastructure. And these are all different entry points to the problem we're solving where we're going to help people automate the web. Right. Maybe you already have something you just want to buy some infrastructure. Hey, you have a low level option for you. Maybe you're a developer, you're getting started and you have cursor or Claude Code set up. You could just start with Stagehand. Maybe you're a PM that wants to see if this thing is even possible. Director is a good starting point, but I'll tell you what I've had PMs start using Browserbase. I've had engineers start with Director. So I think it's hard to like when people create ICPs like ICPs are not the absolute like there are people outside the ICP who use different products. It's ideal. It's ideal. Exactly. So for Director, we want to make a great entry point to understand what's possible when you let AI automate the web and Director is a really nice interface for that. And of course, when you're building a framework in 2025, you absolutely need to have something to generate that framework. If it's not integrated into Bolt.new on its own, or if it's not a very lovable, are they going to generate your framework code? Talk about AASEO. If I want to have more Stagehand in the world, I need to make it extremely easy to generate Stagehand code.

Director is one of those paths. For the stage framework. Are you thinking about like, do I want my infrastructure to have special things so I can do things that others do not like

Vercel did with Next or not really? I think there's a lot of value in us running Stagehand closer to the browsers. I think that we built a lot of Stagehand to be great on its own. You know, we support local browsers, bring your own browser, but there's a lot of items on my roadmap where I'm excited about how we can do more interesting things like caching people's station actions, or how can we help you improve your prompts? We just launched observability for Stagehand and we have this really nice observability thing where you can see how long it took your AI to click this button. What did it think about? What options did it consider when it was going to click that button? How many tokens did it cost? So with Stagehand and Browserbase, we do think we can build a really compelling infrastructure layer that works best for that framework as the office of the framework. But we don't think we can, we're going to do that and also like, you know, kneecap the framework while doing that. I think you can do both. You know, I think the core stuff we've done in Stagehand, which has been reused in many other libraries, is still something we want to keep open source. Now the rails to make it run well on a browser, I think that's something that's less viable to open source, but valuable to our customers who use it with Browserbase. So I think you can balance that quite nicely. I think like server actions is what you're referring to with Next and stuff. I think that server actions are great and like you can do server actions on your own if you want. They don't have to do with Vercel. If Vercel did server actions and then like remove that part from the open source corpus, I think it'd be a little bit less friendly, but it makes sense that you want to improve your framework to help your customers that are already using the platform. And that's how we always approach it. With Stagehand, I'm always thinking about, I care about the Browserbased customers using Stagehand first. Those are my number one priority. And then I care about all the open source people second. The people pay us money are going to get the most support with Stagehand.

So this is actually your second startup. What's different this time? I mean, how are you thinking about it differently? I mean, obviously we're most interested in the go to market, but like Browserbase is growing much faster than the first one, right? So what are the big lessons learned that, you know, any technical founders listening and their heads of marketing should be thinking about?

Yeah, well, I was CTO of my last company and what we were building was prosumer software for live streaming. So you can kind of imagine like Canva for live streaming. That company, we ended up selling it to mux video API company. And that was a great outcome. But I think the trouble with that company was like prosumer software is extremely challenging. It is so hard. You're stuck between free and very expensive. You're trying to find a way to make that work. And we had the best product on the market, but we certainly did not have a great market to go build. So with Browserbase, when I was approaching the company, I really thought hard about business, a great market. And like one thing I did early on was like, okay, let's say, you know, how often are you buying something online? Geez, could be every day, every day. More than like once a day. Often it's often five times a day, not five times. Okay, we'll call it three times a day. Stripe is a $96 billion company and they power the transaction infrastructure of the internet, right? How often are you getting like a text every day online? I guess fairly often, like seven times a day, several times a day. Twilio $16 billion company powers the SMS infrastructure. How often are you using the internet for work? You're clicking a button, you're filling in a form every minute of every day. Okay, so majority of the day Browserbase is attacking that market. I think when you think about how often you're doing work online and how much of that work could be automated by AI, it's a very, very large market. That level of like sophistication of thinking about how big is this market we're playing in? Can we play in it? I have experience in dev tools and infrastructure. That all went into the kind of formation stage of this company. I think upfronting a lot of that work allowed us to execute very, very aggressively for the first year and a half. And I think that's what I like trying to advise founders is like, you should be putting more thought, like you should write a 3000 word essay about your idea before you start the company. Yeah, which you did. Yeah, which I did. Exactly. I mean, that's like the baseline. Like if you don't put that much thought into building this thing, you're going to have something down the road that you figure out as much as you can upfront that you're going to be a good spot. And then I think I've just spent a lot of time at great infrastructure companies. Like I think about mux and they have amazing branding. You know, they have great developer relationships and they built this like, you know, a warm and fuzzy brand in dev tools, which is very challenging to do. I was early at Twilio. Twilio did some of the best dev rail stuff you've ever seen. They invented the five minute demo and usage based pricing and all these different positioning things. So I think by studying great companies in the category, but just adjacent sectors, and then also having kind of like first principles thinking about why this company is going to work really well. That is the recipe for success to really kind of go out and build a great company and do it in a straight line. We don't have to kind of pivot and zigzag. We see all

the time, you know, companies come into us for advice and they just don't have the time. They never thought of it. They did follow the advice of like, oh, build something you would love to use or that you're passionate about or that you would find value out of, but they didn't do those extra steps. And that sounds like it's made all the difference

for you on Browserbase. I talked to a founder this week who has the opposite vision to you on what will happen with the browsers. And I'd love to get your take. What they are thinking about is most agents will browse like a separate internet. They think that most companies will have to build like a parallel website that is mostly markdown or has MCPs or is prepared for an agent to read it. And their take is 10% of the viewership will be humans and it will be your regular website, but for all of the rest, it will be mostly agents browsing and you have like a special design website that is specially done for those agents. If that will happen, I think it's a problem for you guys. Like, do you agree with that vision? What do you think about it? Or do you think like that will never happen? Sounds like a

lot of work. You want to rebuild the entire internet just for agents and the models are going to keep getting better, such that they'll probably browse better than us in the future. I think that that take has a lot of anchoring in the current model capabilities, but in very soon time, models will be able to browse the internet in the current form better than you and I will. And I think that you have to imagine the model capabilities get better and better. I'll use an analogy. We have all these roads in San Francisco and we have a self-driving car. We have Waymo driving on these roads. Now it would certainly be more efficient for us to rebuild all the roads of San Francisco for Waymo's or maybe have a Waymo lane, but I don't think we're ever going to do that. If Waymo can navigate the roads that we currently have just as well, if not better than the people who are driving them, you don't have to rebuild the roads. You can continue to have AI use the infrastructure that already exists. And sure, like AI might do the majority of the browsing. Guess what?

We're going to be the slow ones on the internet. I think also like if you ask a developer, Hey, you have to have the mobile app version and you've got to have the thing and now you've got to build the AI version. Well, guess what they're going to use to build the AI version. They're going to use AI that has to be able to look at the dang page to make the AI version. So then you don't even need it anymore. I think it was just interesting.

It's a super common take. I get emails about it all the time. And I think those people like they are approaching the now and they're going to build something valuable. There will be AI first primitives for, you know, MCP is a great example of this. You know, if you're booking a flight, you should just use the MCP server from Delta. You don't have to use the website if you don't need to. But if you're going to a lumber manufacturer in Mexico, they're not going to have an MCP server. They're not going to have a first party AI integration. And I just genuinely believe that people will still have to use the internet for a long time and they're not going to unlearn, you know, that behavior. You know, it's going to take a generation or two before people browsing the internet goes away. And if people are still browsing the internet, you still need websites for people. And if AI is going to keep getting smarter, it can browse the web the same way we do. So I don't think it's worth investing all the effort to rewrite the whole internet. I think it's worth building great MCP servers for high volume use cases. And hopefully that protocol keeps taking off. I think we'd use it in conjunction quite well. But at Browserbase, we're tackling this very, very long tail. We think it's a long, fat tail of things that can be automated online. And I'm always surprised at how many international customers sign up to use Browserbase because the websites that they automate in their countries are not up to spec. They're still in PHP. They're still pretty painful. PHP is great. Yeah. PHP is great. Sorry. Yeah. Fantastic. The hottest stack in most modern, for sure. But if you go get a, I have to go travel to India soon. I have to go get an Indian visa. That website is a great example of a website that can be automated by AI without rewriting anything. It's super painful to use.

How much is your team using AI in their day to day and what are they using it for? Is it more for support? Is it more for marketing? Is it more for the engineering team? Is it for all? And how much do you encourage the team to use it daily as well?

We hardly use it for marketing. I really hate AI copy. So for the most part, I really discourage use of us having AI write our copy on our website. I think it's just so easy to smell that out. And when I read AI copy, I just don't like it. Yeah. The em dash is giving away. We use it quite a bit on engineering. Our engineering team is all using AI to code gen and also review code. I think it's funny because we have a lot of engineers who are kind of like converts. Like previous like anti AI coding and then like they fall in love with cursor and now they're like, what was I doing before this? So that makes me very happy. And then on support, like what I actually, I built a tool recently where I use Claude Code to call our, you know, our database, our support messages to try and help triage things and like create reproducible scripts of what's going wrong. So I think we're using AI to generate internal tools that help our customers better. That's been quite cool. So it's mostly an engineering world. I think I'm very wary of using AI on the marketing side. I think I would use it for video editing. I don't think I would use it for generating copy. I would definitely use it for images too. So on the engineering

side, besides cursor, what else do you use? Like, do you use any of the reviewing platforms like CodeRabbit or something like that? Any specific things? Yeah, we're using cloud

code. We have a few more things, but mostly just the code gen tools. Like we're not super fancy there yet.

We could be more fancy though. You can get AI to write copy like a human. You have to pick which human you got to put a lot of work into it, but I'm starting to have success there and it's still, still takes a little editing, but if you work at it, you can get it to be a ghost writer, but probably not worth it.

Is that why your emails have been sounding off to me recently? I thought you were mad at me, but maybe it's just your

I've been using what he does for most of my content. I do Whisper Flow. Yeah. We deep research sometimes, and then I do upload. So I created like a doc. They said 20 page doc with good examples on how I stand by examples of how I don't want to sound and sections. And I think that plus Whisper Flow, plus deep research has been pretty good. I've actually

changed how I write because I was at pretty early on after GPT. I started getting that questions cause I I'm a very structured, I'm not like a formal writer, but I have a very structured like concise way of writing and I use lots of bullets and I got asked enough that I was like, I need to change it up a little bit so that people stop asking. So I have more, more run on sentences, more tactical typos to throw people off so that they know

it's me. You have to speak in rhymes a little bit. Do you guys want a quick preview of our first out of home campaign? It's just a, what they call we papering where they put posters up and we're doing it in New York city this week in one of the key messages that we're going to try here. I love your feedback on it. It's a red poster. It says, Browserbase on the bottom and it says, ask your AI. Now you may get the reference, right? And I think that what's cool is like, I hope somebody says to their AI, what is Browserbased? Cause when I've been testing that prompt, you get just the most amazing sales pitch and I'm really curious to see how well that goes. Of course, it'll be accompanied by three or four other posters that are more like headless browsing infrastructure for developers or Browserbase, you know, automate the web. But I, I really am like playing with the ask your AI. I think we want to bring that back because if you're asking your AI about Browserbase, it's going to stay really amazing. It's a very good callback. I love the ask your AI.

My question is, will it work in New York? Like in here in SF for sure I think it will work. But in SF and New York, I don't know how many people know of that. But you don't

have to know about it. This is a reference to Twilio's ask your developer campaign, which was great. It was also 10 years ago, but to me this works because it doesn't need that context.

Yeah, we'll see. It's we're putting it outside of an event in New York, a developer event that we're going to try and we paste the outside. So we'll see how it goes. I think we should follow up in a couple of weeks after the thing goes out and see,

oh, this thing's good. This thing's going to take a few weeks to see that campaign will have been live.

All right, we'll do it. We'll do a quote tweet with this and I'll tell you the attribution

from it. One thing I did in New York, maybe useful for you that is, was very cheap is that if you want to do it in a conference in a certain section, you can actually pay for ads on the taxis only when they are in a specific section of the city. So what I did was for all of the taxis that were moving around like three blocks away from the conference, as soon as they arrived there, it would show my banners and my ads. So like they were moving around all of the space. Like that's a really good way also to get people to see it as it's

moving. Okay. Yeah. This is another great hack because if you spend enough of your funding, you'll have to do another fundraise, which will lead to another launch.

And those have gone really well for you. Yeah. Surprisingly, you know, we raised a lot of money, but pretty efficient business. Haven't touched most of it. I mean, the interest rate from our most recent round just paid for our Hawaii offsite for the whole team. So, you know, the easiest way to go make a million dollars is to raise 20 and then collect your interest. We'll close

it for today. Thank you for coming and thank you for answering one of our questions. Yeah.

Thanks for having me. And I hope everyone got to enjoy more about the story of Browserbase.

Thank you.