🎧🍌 Building Infrastructure for AI Browsers with Paul Klein, Founder + CEO of Browserbase 🅱️
Unlocking better AI agents, how the best infrastructure companies become product companies, why you shouldn't start a company, the playbook that raised $27m in nine months, + liking your own tweets
Paul Klein is building the infrastructure for AI browsers. This conversation gets into the future of software and AI agents.
We talk why authentication is a huge problem in AI, how the best infrastructure companies become product companies, and the memo he wrote that convinced him to start Browserbase despite not wanting to build another company.
A year ago, Paul was a relatively unknown commodity, and definitely did not want to raise venture capital again. He shares the playbook he used to go from zero to raising $27 million in nine months “as a non-famous person” (his words).
He also opens up, sharing all his lessons learned in the arena as he’s processing them, like what he thinks will unlock better AI agents, why you should like your own tweets, and how Browserbase competes with incumbents.
Timestamps to jump in:
2:39 How LLMs unlock automation online
8:34 The future of software (AI agents)
11:21 Why AI agents need better authentication
12:59 Lessons from Twilio on building an infrastructure company
17:27 Learnings from his first startup
19:56 Bubbles, and how they drive innovation
20:37 Reasons this moment in AI is special
29:58 Why technical founders love post-PMF
31:55 The memo that started Browserbase
34:09 Why a startup should be a means of last resort
36:53 Being a solo founder
42:24 Importance of in-person culture
45:56 The best place to find engineers
48:34 How Paul hired a contractor army to build Browserbase
50:16 Why you can’t hire mercenaries
54:28 The power of emojis in marketing
57:39 Browserbase early growth playbook (3 videos)
1:04:00 Benefits of sharing an office with other startups
1:06:00 Sales lessons from his parents
1:08:07 Why startups are like video games
1:13:43 Successful founders work the hardest and are shameless
1:18:44 Customer support is a startups greatest differentiator
1:22:06 Paul’s playbook that raised $27m in nine months as a non-famous person
1:29:03 How investors make decisions
1:33:10 Risks help startups avoid competition
1:36:37 Great infrastructure needs its own frameworks
1:39:05 Long-term thinking in LLMs will enable mass AI agents
1:42:21 Avoiding tech debt with AI moving so fast
1:43:48 Infrastructure companies need to become product companies
1:46:54 The Sine Wave philosophy to startups
Referenced:
Paul’s Memos:
The three Browserbase marketing videos:
Find Paul on X / Twitter and LinkedIn
👉 Find on Apple, Spotify, and YouTube
Transcript - (read on Rev)
Find transcripts of all prior episodes here.
Turner Novak:
Paul, welcome to the show.
Paul Klein:
Thanks, man. I'm super happy to be here.
Turner Novak:
I'm excited to have you too. Really quick before we really get going, what is Browserbase for people who just don't know?
Paul Klein:
Sure. Yeah, Browserbase is an infrastructure platform. We make it really easy to run web browsers like the one that runs on our laptop, on a server in the cloud. So we run thousands of web browsers that many AI applications use to go out and go to websites on your behalf, click buttons, fill in forms, and take actions.
Turner Novak:
Why is that a big deal?
Paul Klein:
Yeah. For the longest time, if you wanted to automate something online, you probably had to write a script that was like this one-off script. You wrote one script for one website. I remember one of the first things I did as a programmer was go out and try and automate a thing in my life.
Turner Novak:
Do you know what that was?
Paul Klein:
You know what? It was something super dumb. It was like check my school website for my sports schedule. When's practice this week because I have to go check my website on my own and see, so I try to build something-
Turner Novak:
And does this happen a lot? Just if I'm not a programmer and I have no idea how this works under the hood, are a lot of people automating a lot of things on the internet?
Paul Klein:
I think the web is one of the most important resources of all information out there. And oftentimes people are using APIs but there are websites that don't have APIs. You have to go to the website and actually parse the web data and take action on it. And it's a super hard thing to do because one, you write these scripts that work for one single website, but then if the website changes, your script breaks. You say click the calendar button and they maybe change the button from calendar to say schedule. Well, your codes looking for the calendar button, it doesn't work anymore.
Turner Novak:
It's not really even automation because you have to keep fixing things. You can't really automate this stuff.
Paul Klein:
It's super deterministic, rewriting deterministic code for a non-deterministic internet. And kind of like why is it important now? Well, all of a sudden, LLMs make it possible to generate code on the fly and you can ask an LLM like, "Hey, on this website, where is the schedule? I'm trying to figure out when I have practice." And it can help you find what your support schedule is or it can help you find the item that you want to add to your cart. It can help you look for the flight or see an available slot for your yoga class. And this is all because LLMs can generate code on the fly while looking at a website.
Turner Novak:
So you can truly make it automated because you can just say, "Oh, there's this little thing that changed." Instead of Paul having to spend 20 minutes fixing it once a day, the LLM just does it.
Paul Klein:
And you can write applications. You build applications that actually can work on websites that you've never seen before.
Turner Novak:
Oh, interesting. How could that play out?
Paul Klein:
Well, in the past let's say you wanted to write an application that booked a demo call on a website. Let's say you're a VC and you're trying to book a demo call on all these websites. You-
Turner Novak:
[inaudible 00:05:27] Browserbase book a demo with every company that has announced a series.
Paul Klein:
Exactly, right. You'd have to write a script for every one of those websites. But with AI you could say, here's a hundred URLs, go to these websites, book a demo call please. And that's just way powerful. You have to write one script versus 100. And what Browserbase is we're the runtime for these automation scripts. We are the browser that's actually being acted upon by the LLM to accomplish these things on your behalf.
Turner Novak:
So why is it a good time to do this right now?
Paul Klein:
I've actually been working on this specific problem, headless browser infrastructures for quite a while now. I know a lot about the browser. The browser is super hard to run in a server environment. Imagine running when you run Chrome on your laptop, it basically nukes your laptop. Now, imagine running that on a server which is not as powerful as a MacBook Pro. It's often much less. It's like a Lambda-
Turner Novak:
Really?
Paul Klein:
Yeah, most serverless functions or maybe one or two CPUs. Your M3 is a beefy thing. It's pretty intense. And what's really hard is running these browsers in cloud is actually a very complex infrastructural problem. Developers have to spend a bunch of time on this, but it's not really helping them build their application. It's just something they have to do. They rather would spend more time building the features, not making the infrastructure work. So that's kind of where we come in. But really the rise of large language models and the ability for us to now generate code has made the need for infrastructure like Browserbase more than ever because there's new capabilities.
You can now generate automations of these websites on the fly, and you're no longer limited by how many scripts a developer can write. You're limited by how much code AI can generate. And that really means for us being an infrastructure company, there's just a lot more possibilities that have been unlocked and we're really excited to service them.
Turner Novak:
So what is an example of a way that someone might be using it today? If I'm a browser-based developer, any case studies or maybe favorite podcasts can make this exciting for people? What are some really interesting things you are doing with it right now, just to give people an idea of what's kind of going on?
Paul Klein:
Yeah, I think what I think about the use cases of Browserbase, there's all the obvious ones that come to mind. Like go on my LinkedIn and send a message to somebody or, "Hey, can you book a flight for me?" But where I get more excited is all of the super esoteric use cases. Like, "Hey, I want you to go upload this receipt on this legacy website because they don't have an API." There's all these use cases in enterprise where, "Hey, I need you to go to this court website and pull the court schedule for the day and tell me if this judge has a case like this." These are all tasks that people have had to do in the past.
Turner Novak:
So these are a lot of manual one-off, maybe non-recurring, maybe they're recurring type of tasks that a human would probably be doing on their own manually cannot be automated as of today.
Paul Klein:
Yeah, the future of software is software doing work on your behalf. And if you think about the work that we do, we're doing this work in a web browser. I go to this website, I check the information, I do something with it. In the future, software is a piece of software you're going to use that's going to have a button. You're going to press that button and it's going to go out and use other software for you. It's agentic. You hear a lot of talk about AI agents and stuff, but what that really is just smarter software that can use other pieces of software.
Turner Novak:
Is that a definition of AI agent?
Paul Klein:
I would say my definition of AI agent, I'm not even going to say the word AI, it's so controversial. My definition of agents is software that uses other software, that's agentic. And if software is using other software, guess what? A lot of software lives within a web browser, on websites. So the thing about Browserbase is there are so many different use cases. Literally anything you can do on the internet, you can automate with Browserbase because it's a web browser that's controlled by an LLM that's going to click buttons, go to pages, fill out forms just like you would.
Turner Novak:
So what are some ways that the internet is probably going to change that? Maybe some of these are obvious, maybe some ways where it's like, oh, that actually is not that obvious?
Paul Klein:
If AI is going out and using the internet, what does that mean for the internet? Is the internet dead? It's like are people going to pay all their content? I think a lot of people have talked about AI web crawling and how that's being fed into things like perplexity or open AI. I think we're in this kind of weird period where right now the capabilities of AI has really kind of outpaced the internet that it's kind of using. The internet hasn't caught up with the way that AI's going to go out and use it. Yeah listen, there's still a lot of things that need to improve for AI to be effective to replace people and to do work, but there's already a lot of promising stuff that's happening.
Turner Novak:
And what's the biggest example of just the internet needs to get better to upgrade for AI?
Paul Klein:
Yeah, I think for me right now we need to think about authentication. How can AI log into a website on your behalf? Right now it's pretty junky. Right now there's a lot of credential sharing. The most secure way is you share your login cookie with your AI agent. Then it authenticates us with... As you we're lucky to have Okta Ventures, Alan from Clerk, Reid from Stitch. All of these great identity people are investors in Browserbase. And I've spent a lot of time with them talking about how can we make it so AI agents can log into websites on your behalf in a secure way. And I think that's the next big thing that's going to change because right now you giving your AI agent your Netflix password probably is the most secure thing.
It probably needs to be able to log in and say, "Hey, Paul, can I log into your Netflix?" Needs a push notification and then you say, "Yeah, go ahead, AI agent. You can go check my list and let me know what's leaving Netflix soon." Just like authentication is the next pillar to fall to make AI agents a reality in this world.
Turner Novak:
And you are not a newcomer to authentication. You've been working on authentication for a while?
Paul Klein:
Yeah.
Turner Novak:
When did you first get started with it?
Paul Klein:
How'd you find that out? You must have looked up my LinkedIn or something, crazy. I've been in the authentication space for a while. I started my career actually on the login team at Twilio. So I was in charge of Twilio, which is a text API company, an infrastructure company. I was in charge of their login, sign up flows. And I've really kind of been on the other side of this, which is preventing bots from logging into websites which is what I did at Twilio. Yeah, exactly. So I've seen that and my background and identity in access management, and really my time at Twilio was really formative for the work we're doing now at Browserbase having been at an infrastructure company in the early days.
Turner Novak:
Plus you probably knew how to navigate people not wanting you to access.
Paul Klein:
Yeah, absolutely. I think that when you look into anti-bot, what it really is is does this web browser look like a real web browser? And I think here's the cheat code, use a real web browser. Don't do any weird tricks. Don't say something you're not. Our Browsebase browsers are very recognizable. We're not saying we're a Windows computer. We're saying we're running on Amazon, we're in a cloud environment. And just by not lying, you end up being more trustworthy. I think it's just a good strategy in life as well as in anti-bot, coincidentally enough. Just be as transparent as possible.
Turner Novak:
And so what happened on your first day at Twilio?
Paul Klein:
Yeah, so I joined Twilio in 2016. I was actually an intern, and I come in to our all hands meeting. And Jeff Lawson, the CEO calls up all of the interns and we do a fun fact. It's like a normal thing, hey, where are you from? What team are you on? What's your fun fact? And immediately after our fun fact, we sit down, I'm like, holy crap. I just presented to the whole company. That's so scary. Jeff comes out and he's like, "Oh yeah, and one more thing. Here's our S1. We're going to go IPO in three weeks." And I'm like, what? This is crazy. And everyone's celebrating and everyone's getting rich in this room and interns don't have any equity, which is fine. But I was so stoked. I was like, wow, this is happening. And I really got to stick around at Twilio. And yeah, I'll never forget I had just turned 21 and that was my first open bar ever, was at the Twilio IPO party. So it was really a crazy way to start off a career in technology.
And for me, I told myself that wouldn't be the last one. Ever since then, I've been thinking about how can I build my own company and really be a category defining company like Twilio has been for the telephony space.
Turner Novak:
Yeah, it's interesting the IPO should be, I don't want to say it's the ending because it's a long journey, but it should be maybe closer to the not beginning of someone's journey and first day. Interesting to kind of experience that right away.
Paul Klein:
Yeah. And I think what was especially cool was to see just how proud the team was. And the team that took Twilio to IPO had grown a lot alongside the company. And to me, the most important thing with building Browserbase is that I want to build a company that enables people to grow more than they could anywhere else. I want someone to choose to work here like, oh, this company it's going to be hard. They seem like they're going to succeed, but I know for sure I'm going to grow more here than anywhere else. And as long as that's the reason why people are working for us, I think we're going to be very successful.
Turner Novak:
Any other big lessons from Twilio?
Paul Klein:
Yeah, I think Jeff... It's like Twilio and Stripe were really the two companies that I've looked up to the most in the infrastructure world. And specifically with Twilio having been there, I thought that they did an amazing job of making every all-hands super... Like you were going to get your MBA working at Twilio. You'd go to the all hands meetings and you knew what net revenue retention was. You knew about dollar-based expansion. You cared a lot about diversity of revenue. And I never took that for granted. And leaving that experience, I knew a lot about the economics of cloud infrastructure businesses and why? Hey, you probably want to have a diversified customer base because if one person churns like WhatsApp did or Uber did, it's going to take the stock price. So the power laws of distribution, of usage of cloud infrastructure it's super interesting.
And being able to learn that firsthand in a very transparent environment like Twilio all hands, I thought that was foundational for me. And I keep that up within our team where I share really as much as I can, pretty much everything about the business. And try and teach everyone why the things we're doing are working and why there's things that aren't working and how we improve them.
Turner Novak:
Is there ever a risk of over sharing, like you share too much? And maybe an example things might be going well, really easy to be like, "Yeah, we're crushing it." What if it's a bad day? It's like revenue dropped this month, or our biggest customer is not happy, how do you navigate maybe sharing the not so good things?
Paul Klein:
Yeah, we've encountered that and I think you just share it. When we had had a big customer churn and that really sucked, and it was a wake-up call. I was like, "Hey, everyone, this customer churns." And the data is all there. Everyone can see all the charts of the usage. We have that published weekly if not monthly in Slack.
Turner Novak:
Oh, you actually publish every customer?
Paul Klein:
Yeah, everyone knows every customer.
Turner Novak:
Interesting.
Paul Klein:
It's all there. And when this big customer churned, it was a really important lesson for us to kind of look back and say, "Well, why did they churn?" Okay, well, our product wasn't really servicing them this well in this way. Do we want to service people like this? Should we have a different product strategy? And it was a combination of execution and strategy that we really kind of took as a lesson and eventually have one similar customer to them at larger volumes since then. But every churn is a lesson. If you just ignore it, you're missing out on a ton of valuable insights.
Turner Novak:
After Twilio, you did start a company. It was Stream Club.
Paul Klein:
Stream Club.
Turner Novak:
Yeah, how did that come about?
Paul Klein:
I remember it very fondly, I just had a really tough day at work. And I came home and I got a very convenient Instagram DM from my future co-founder. He is like, "Hey, what are you doing this weekend?" I was like, "Nothing much." Kind of upset at work, he's like, "Well, do you want to hack on something with me?" And I was like, "Sure, let's do it." So me, Lan and June all hacked on this thing which was a virtual events' software.
Turner Novak:
Virtual what?
Paul Klein:
Virtual events' software.
Turner Novak:
Oh, events. Okay.
Paul Klein:
Yeah, it was right early in COVID. We were like let's make it easy for yoga teachers to charge for their online Zoom yoga classes. Classic April 2020 startup idea. All right, we were pretty early. All right, so we hacked on this thing. We launched it, got to the top three on Product Hunt. And I was like, oh my God, this is awesome. We built something that works. And I think I had been feeling a lot like I was a small cog in a big tech machine, and this was the first experience where I was like, wow, I actually have agency over what can happen. If I work on my own company, I can see what I'm truly capable of. And with Lan and June, I knew I had the compliments of an insanely good designer, an insanely good growth person. And I could bring all the engineering side, let's all work together and do this thing. And that's when I left my job to start Stream Club and we were off to the races.
Turner Novak:
Did you do it for the right reasons? What are the right reasons to start a company?
Paul Klein:
Yeah, I think, well, once again, I wanted to start the company because I was like, oh my God, I don't like my job and I feel like I wasn't finding my meaning. But I don't think we started Stream Club because it was the best idea or the best market. I think we were able to come to some conclusions about why this is a good opportunity. Someone is going to build great virtual events' software in COVID, and there's going to be a shift to this creator economy and the future of work is online. I think a lot of Twitter was talking about the creator economy back then too.
Turner Novak:
There's a lot of... I feel like it's sort of coming back.
Paul Klein:
It's coming back?
Turner Novak:
Yeah.
Paul Klein:
I'm still bullish on the creator economy. I think we got a little crazy there in 2021. All right.
Turner Novak:
Yeah. I think my general way I approach these is I think usually the discourse of the things people are talking about, the context of it is generally correct but the timing is often not correct. So you can just look at the dotcom bubble. We were like, oh, we're going to get 15 minute grocery delivery in 1998 or whatever. We did get it but it was 20 years later. Crypto, everyone's going to be buying NFTs and making all their purchase on NFTs in a couple of months or whatever. Maybe NFTs is a good example. That one's ever coming.
Paul Klein:
Yeah, Tulips were really valuable and they were going to be even more valuable in 20 years, I promise.
Turner Novak:
Well, it's kind of one of those things where I think all the logic that goes into these things and the technology is probably going to benefit society in some way, but the timing of when it happens and how you actually make money from it, I don't think we get that right very often. But I think it's helpful.
Paul Klein:
Do you think that's the case with AI then?
Turner Novak:
I actually think AI is tricky because it's probably the most use cases. It's very clear today, it's actually doing things. I think probably the tricky part was it was really expensive originally, so there's a lot of capital momentum that got burned. That would probably be my one sentence takeaway. Probably built the wrong stuff.
Paul Klein:
Well, there was an old wave of neural net company, startups that were like, oh my God, AI is happening five to 10 years ago. And this is not the first AI wave. So I would say that actually aligns with your thesis, which is there's a big spike and then sometime later there's actually true innovation, which I think you're right. You've seen with the creator economy, I think there's a lot of really interesting stuff happening now too.
Turner Novak:
Yeah. And I think even if you go back to just through human history, I only have been to maybe 100 or 200 years, but there was a point where, I forget the exact number, but it was something like 10% of UK's GDP was invested in railroads in a certain year or something. It was double digits of their entire GDP that they invested to build railroads. And again, I forget the exact number but the returns were atrocious, but railroads significantly changed the economy. And you could say objectively made the world more efficient, better, saved lives, things happened faster. So again, it's one of those things is AI like railroads, where it's very clear going to make the world a better place? But how much and where is the money going to be made? I don't know.
And I think probably you'd say today it's probably a good time to be investing in AI versus maybe the peak. I don't know, I would not say I'm pessimistic but I'm a very cautious person where when I see there being the most exciting hyped up category, I kind of give this allergic little bit of a, maybe we should pause and just make sure we're accurately assessing and not moving too quickly into things. But there's usually a reason for it.
Paul Klein:
Yeah, it's also for my business, there needs to be healthy competition at the AI lab level. And I think I'm surprised at how that's actually panned out. And there really is... if you asked me back in 2022 or 2023 if OpenAI would be the only lab, I would've said yes. I was like, oh, who's going to keep up with them? But now we have healthy competition from OpenAI, Anthropic, DeepSeek just came out. We have all these interesting open source plays, Gemini. There is a lot of competition at the model level, which means that developers are switching between models as they get better.
Turner Novak:
It's amazing if you're using them.
Paul Klein:
And it's amazing if you're an infrastructure company because as a developer you want to have infrastructure that comes along with your prompt choice. It's really easy to switch out a model, but are you going to bring your whole database, port it over to, okay, if I switch from the OpenAI model to the Anthropic model, am I porting my embeddings over? No, I'm going to have my embeddings in a separate database and just switch the model out underneath, which I think it shows that you can have healthy model level competition. Introduces the opportunity for infrastructure providers to decouple themselves from the models and really play a separate game, or how do we enable these models to be more powerful and do more things? Well, with us we give them browsers that they can use to control the web. You can use 3-5 Sonnet or 01. Either of them work, they have their pros and cons but they both need browsers.
Turner Novak:
Yeah, I think it's probably, is there a customer or a use case that needs to be solved that isn't being solved yet by someone who has a built product yet? So it's like if you're building the eighth competitor in a space or funding the ninth or 10th competitor in a space, and it doesn't do anything unique, maybe not a good idea. But if it's like, there's this clear use case that's not being solved. It's going to save a bunch of people, time, people are going to pay you for it, et cetera, maybe there's an opportunity there.
I feel like there's some interesting analogies maybe when you talk about what's my opinion on the hype-ness and the sustainability of maybe AI is like a wave or cycle. I just kind of think if you compare it with crypto, so I feel like with crypto there was almost like a... I don't know how similar, but there was a sense of we need to be doing this. So if you just take a corporate entity, let's just say Facebook, let's say the Facebook board or even you're a mid-level tech company, you're like a five billion public tech company and you're like, "Crypto is a thing. We need to have an NFT strategy." I think there's a pretty rational case if just as a CEO or management team you can talk to the board, talk to the investors, maybe you say there's a case for it. There's probably a ton of cases you can say like, "This actually doesn't make sense for us. We don't need a crypto strategy. Here's why it doesn't make sense."
But I feel like with AI, it really is just a part of all software. So I think it's actually harder to go to your board or your team and say, "We don't need to do anything with generative AI or large language models. It's not going to affect us at all." So I feel like it's just much more applicable. So I think it probably comes down to a timing. That's how I would think about it, as the cost and then the effectiveness. Does it actually work yet? Or then a timing of in nine months or in 18 months, will it be good enough or is it maybe... We actually don't know yet. It might be five years still. I feel like that's where it kind of breaks, but you still hit a point where you got to be working in it. So it's like the excitement level I think is justified because it still does work.
Paul Klein:
Well, I think there is some magical moments here where you can build something using AI. And I think with our customers, what I see is especially somebody who isn't like a software company, but they come us because they maybe hear about us via one of our investors or something and say, "Hey, we have this workflow that people do all the time. Can you help us build something to automate that?" And we say, "Well, we won't build it for you, but here's all the tools that we have. We have Stagehand, which is our open source browsing framework." And what they do is with a few simple prompts, all of a sudden they have this little AI application that's going to a website filling in a form and then clicking submit. And they're like, wow, I had to do that on my own.
The moment of magic is so tangible with these little AI web automation applications. And I think that's what's so special about this moment is that people can tangibly see the results right away. And sure, do we have a desire for it to be better in the future and do more? Absolutely. I think that's what's so funny about technologies. When you see something or do something badly, you're like, oh, well, I needed to do a really good thing. You see a shitty robot and you're like, well, it's a robot. But it's running around and jumping on blocks. You're like, well, it's not doing surgery yet. Yeah, it's not doing back flips yet. I think it's the same thing with AI, we've taken for granted that we have a chat box that we can ask things and it gets a pretty good job of responding. And that's why there's actually tangible utility within the first couple of years, which is super rare for technology.
So I think you may be going back to the question way earlier, what are people using Browsebase for? I think the same answer with what are people using LLMS for, there's just so much stuff you can do. It's really mind-blowing. And what we try and do and we want you to do a better job of this, is showing people the primitives and letting them use those primitives to kind of create powerful applications that help them do something in their day-to-day life.
Turner Novak:
I think there's also, just from an investment standpoint there's a question of does it make more sense in terms of early stage startups? Should the funding be going and solving problems or whatever, very early stage? Is it better to do it more in the growth stage or are you going to make more money in the public markets? I think that is an interesting part of the equation you should probably think about as an investor. So I think a lot of technology you can work across the board. But you can also say like, man, Facebook's business model is you hook up a credit card and it just buys ads if Facebook can guarantee people will click on them. What is generative AI going to do to Facebook's ad system? Well, it's just going to run all the ads automatically and Facebook's going to make a trillion dollars in market cap because of it. So it's like, can you do anything related to what Facebook is going to capture the value in? Probably not. Probably don't do something related to generative AI and ads unless you're capturing a new consumer demand, like consumer eyeballs and attention span. So that's another way I think about it is just what is too far in the strike zone? I feel like this is a common thing people say is if your product's on OpenAI's roadmap, might not be the best thing. Try to find things that they're not going to do. I don't think that's always the right case.
Paul Klein:
And it totally depends on ... I think if you're talking to an investor, but if you're talking to a founder, I think a founder should be taking a lot of shots on goal right now. There is a lot of asymmetric upside to starting a company. It's actually funny. The thing I lose candidates to the most if I'm trying to hire someone at Browserbase, my number one competitor, they start their own company. It's so brutal. I'll meet an amazing engineer. I'm like, "Oh yeah, you should come work here." And then by the time we're done with it, they got into YC and I'm like the YC four batches a year has screwed my hiring pipeline because I used to ... I'd only have to worry about it two times a year. Now there's always a YC app I'm competing with.
Turner Novak:
Yep. Every three months.
Paul Klein:
Yeah. It's brutal. And then I have to wait for them to go through their company and if they figure it out, hopefully I may have put a check in and I can get the upside with an angel investment, but if not, I have to wait a year or two until they run out of capital and come back. I love working with second time or former founders because I think they end up having a much better attitude towards startups. Now that Browserbase has product market fit, you can really just come in and execute. There's so much low hanging fruit that we need to pick. We need people who are ready to come in and just execute and go hustle. Especially when you're a technical founder at a startup, you want to execute so badly but early stage startups are so much more like product market fit to be discovered. And you can't really execute your way into PMF. You have to really have an earned secret as my buddy Zack Hargett loves to say. You have to know something about a space that nobody else knows and then be able to execute on the smallest version of that to validate it and then get into the right hands of people and do it in a repeatable way. And that's super hard.
So these founders who can't figure that out or they can't figure it out in time when they come to Browserbase, they end up being just amazing team members because they have so much expertise and an appreciation for like, "Oh my God, product market fit is amazing. Let's go take advantage of that." And they're ready to come in and just execute, which is fantastic.
Turner Novak:
Yeah. Well, it's funny you told me that you actually didn't really want to start a company in the first place.
Paul Klein:
Oh, yeah.
Turner Novak:
How did it come about where you went from exited the last company starting Browserbase?
Paul Klein:
There are several text messages that will eventually be leaked where I say quote me on this, I will never start another startup. There's actually another one from earlier this year like, "I will never raise a round of venture capital." I'm like, "Shoot. What happened?" But I think I couldn't even have predicted just how right we were going to be about our early theses around Browserbase. I knew so much about headless browsers from being CTO of Stream Club, the startup I talked about. And I spent some time really digging into this. And in late 2023, I kept getting hit up by people saying, "Hey, you know a lot about headless browsers. Can you help us with something?" I was like, "Yeah, sure. I got nothing going on. Let's just chat." And by the 50th conversation, it became abundantly clear that there was a need for headless browser infrastructure.
Turner Novak:
Was this other founders?
Paul Klein:
Other founders, yeah. People building early stage AI product. It was the bleeding edge. People were like, "Well, what if we had AI control of the web browser? That'd be cool." And they came to me and talked to me about it. For what it's worth, I was planning to leave my job and I was going to travel the world for nine months before my wedding. My wife and I had it all planned out. It was being super amazing. And one night I sat down and I wrote this 3000 word essay about why there needs to be a company that exists like this. It's called an internet browser for AI.
Turner Novak:
We'll link it in the show notes if people want to read it.
Paul Klein:
Please. I lay out everything about our plan. And shared with my wife and I said, "What do you think of this?" And she's like, "I think you need to go do this thing." And that's when it became real that this is the only company I could ever start now. I'm done after this one. I'm a one trick pony and that trick is headless browsers.But it just became so clear to me that there needs to be this thing that's solved and I could actually do a job of it.
Turner Novak:
So why did you not want to do it?
Paul Klein:
I think that going back to the question earlier about, well, was I starting a startup for the wrong reasons for the last one, I think that you should do a startup as a last resort, as a means of last resort. Because startups are so hard. They're probably the worst way to make money. You end up getting all these weird health issues most of the time. You lose your hair, you get stressed out. It's generally a net negative on your life unless it works really well. However, it is still the thing that has caused the most amount of growth for me in my life because it puts you in front of challenges where you really have to rise to the occasion.
And the only reason to take on all of this effort is if you have hopefully a somewhat higher likelihood of success. And that means really thinking through the category you're entering into. Are you building something that's directly on OpenAI's roadmap? Yeah. You're right. Do you want to waste two of that of your life on that? I don't know if that's the best use of your time. The opportunity cost of your time is so high you really have to make sure you're doing a startup for the right reasons. I think a lot of engineers, they want to go start companies, especially after working in big tech because they want to see what they're capable of. And I admire that. I was that engineer at my first company.
Turner Novak:
So are you maybe not careful enough or you're just a little bit too gung-ho about just doing anything versus the right thing?
Paul Klein:
I've never talked somebody out of starting a company. So I was successfully. And I think that's because being naive is actually what makes first time founders once in a while insanely successful because no one can talk them out of doing the crazy thing that actually works. And I think the second time ... Sometimes I see this thing, I'm a fourth time founder and I'm like damn, good for you. But you really only have one shot per decade I think to build a great company
Turner Novak:
Really? Why?
Paul Klein:
I guess decade for me being from 20 to 30, 30 to 40. You can't brute force it. I think sometimes I see things where people are like, "Oh, I'm going to do a whiteboard full of startup ideas and pick one." I don't think that's how you can build a successful company. I think the way you build a successful company is having deep experience with a problem, deep insight about a industry, and then really doing something to push forward that industry in a way that's valuable and defensible as a business, which is like good startup ideas are objectively rare. And if you're lucky enough to find one, you got to jump on that as hard as you can.
Turner Novak:
So it sounds like in maybe 20 years, let's just say Browserbase is ... Maybe you're still involved or whatever, but we're talking about headless browsers, browser infrastructure, there's another idea. You might actually start another company then if there's another opportunity.
Paul Klein:
In 20 years, there's going to be like AI clone. My AI clone will start a company. I'm good.
Turner Novak:
Good way to cover that. You dodged that. You don't have a co-founder. How did you navigate starting a company? Did you think about I want to bring on a co-founder, I'm going to hire some people. How did that all come about?
Paul Klein:
Yeah. As much as I said earlier that I didn't want to do this and I wouldn't raise venture in the end, there was a lot of intention around, okay, well if I do this, here's how I would do it. And for me, I'm really lucky that I have an engineering background and I've been a co-founder of a startup before. I felt capable enough to start the company and get everything together on my own. And I felt like I knew enough people who'd probably want to come join the team that I could get away with doing it on my own. And honestly, this time around, I'm married. Being married and having a startup is great because my wife is like ... She's my rock. I can share everything with her. My first company I was single and I had co-founders, you say you're married to your co-founders. If you don't want a co-founder just get married. She shares some upside in the equity, right? We're married so.
Turner Novak:
That's true. I guess technically she might own like 50%.
Paul Klein:
We'll say we own the whole thing. Family business. But no, when I started it, being a solo founder was a choice because it allowed us to move faster. When you have co-founders, you constantly need to get in alignment. There's the co-founder meeting where you talk to each other and then you have the team meeting where you talk to everybody. And it's really important that co-founders present as a unified front because at some companies they'll be like, "Oh, everyone has their favorite co=founder." And like, "Oh, I got to get something done. I don't want to talk to Steve. I'll talk to Bob and then Bob will help me convince Steve to do the thing." I just think that dynamic actually slows companies down.
For us I felt like ... And maybe it's for me, but I felt like having one person be able to make these shot calls without having getting buy-in from a co-founder team allows us to move a lot faster. And maybe now that I have a board of directors, I have two people on my board, Bucky and Reid, amazing guys. I have to keep them informed with what's going on, but in the end I get to call the shots and for that reason we can move really, really fast.
Turner Novak:
So what do you think is the most important thing when hiring, you have to bring people on, obviously if you're a solo founder. How do you just generally approach and think about that?
Paul Klein:
Yeah. Hiring the first person was the hardest challenge I think I faced as a startup yet. You're a solo founder working on a company and you have to hire the first non-founder person. Who in their right mind would do that? But I think-
Turner Novak:
Why do you say that?
Paul Klein:
I think it's crazy. If you're a person wanting to join a company and of all the options you choose, there's a company of one and you're the first non-founder employee, I think that's just a really hard dynamic and you have to be a very special person who's low ego, who sees the value and the business, believes in it deeply. And for us, Dom was someone who's actually I'd known for about a year and he just genuinely believed in what we were doing. He was so excited to work at a startup. He wanted to be employee one. And I think there's very few people who want to be employee one and will wear that as a badge of pride. And that's Dom. And I'm like, could not be happier that he was the first person. Because that was the thing that really got the ball rolling. Because once you have an employee one and Dom is there, it's a much more convincing sell. The time from zero to employee one a few months, and then from employee one to two was a few weeks. So the ball really started rolling on hiring after we found the person who said, "I'm ready to go do this."
Turner Novak:
So how does it change? Is it social pressure or do you free up time?
Paul Klein:
Well, I think the most important thing as a solo founders become multi-threaded as soon as possible. And what was nice is that I had a lot of contractors that I had hired before Dom. So Dom was full-time equity owning employee one, but several people had worked for Browserbase beforehand. So when I started the company, the first thing I did was I raised a small pre-seed round 1.9 million or so. Alana Goyal from Basecase led the round. She's amazing. Some great investors as well that joined. And I did that so I can go hire contractors and become multi-threaded. Because if you're the solo founder and you're building the product, then also selling it, then also recruiting, you just get nothing done. And I still did recruit, I still did sell, but at least if I could have the product continue to be built while I was recruiting, move things forward. And I'm lucky that I met someone like Jason Hellman's, who was our first contractor, understood the problem, loved it, and he was able to keep working on the dashboard while I was out talking to people. And as a result of that, we moved really quickly even before hiring our first real employee.
Turner Novak:
So when you were fundraising, did you say, "I'm going to hire contractors."? Was that part of the pitch or was it like, "Don't worry about this. I'm hiring some people."?
Paul Klein:
I was hire a team is what I said in the pre-seed. And then we raised the seed a few months later and I still hadn't yet hired a full-time equity owning employee. But I had a ... I called it a contractor army and we had a bunch of people and we had a product and we had some people using it.
Turner Novak:
How many people?
Paul Klein:
How many contractors?
Turner Novak:
Yeah. How big was the contractor team?
Paul Klein:
I think it was like nine at it's biggest.
Turner Novak:
Oh, wow. And you sad never ... Did you do that on purpose? Were you thinking-
Paul Klein:
Yeah. So I always wanted to build an in-person company in San Francisco and it's really hard to do that. I think especially it's hard to hire employee one and you add all these extra constraints. They're a first employee, they must go to an office five days a week and work in person and they have to have this really specialized skill set of understanding great frameworks and developer tools.I think I narrowed that down to five people in San Francisco and I found one of my five.
But in the meanwhile, I had a bunch of people on Twitter that I just really loved their work and I said, "Hey, do you want to come do some work for us? I can pay you money." And they were down. And I think balancing working with consultants or contractors versus building a full-time team, you have to always think about, okay, what do I need right now? And I think that we've had contractors that started with us in the very beginning, they're still working with us to this day and they're amazing. And as soon as they move to San Francisco, I would gladly switch them to a full-time employee and they come work in our office. But you have to also have your values. And for us it's five days a week in office is how we build our full-time team. Because that's really the long-term mission of Browserbase.
Turner Novak:
The long-term mission is have an in-person.
Paul Klein:
Long-term cultural thing is I want to build an in-person company.
Turner Novak:
So that was really the tipping point on not specifically having full-time employees and having more contractors because you could not get in person in San Francisco.
Paul Klein:
Yeah. It took time. Okay. I didn't want to compromise my value of I want to build an in-person company because that's so, so important to me. I just think it's going to work really well and we now have 15 people in person in SF and it's fantastic. The vibes are so good. I didn't want to compromise on that. And I also wanted to accomplish our goals. The best way to recruit people is to have a great product. So I think there's a hedge there. And it's not me saying I choose to be in person, not because I think being remote doesn't work or being remote is bad. It's just my personal choice of what I want to build for a company. I think a company that does remote super well, it's this company called Railway. They're an infrastructure company. They're fully remote. They work I think completely in Discord and it's an amazing team. They do an amazing job. So I don't think it's one versus the other, remote versus an office. It's really just like, what culture do you want? We had two new people start on Monday and the energy in that room was like, you could feel it. It was the first day of school vibes. It was like everyone's like, "Whoa. Haven't seen you in a little while. How you been?" It was really fun.
Turner Novak:
It was like back to school. My kids went back to school on Monday.
Paul Klein:
Yeah. Same thing with us in our office. We went back to office because we took some time off. Being in person makes the features that we ship feel more real because I can see when Garrett has been grinding on something for two weeks. When I see Garrett is the last person to leave and the first one in building this feature and then all of a sudden it ships it just feels like such a bigger accomplishment versus if you're in a fully remote company and there's an all hands meeting and Oh, we shipped cheaper pricing. Everyone's like, yay, clap, reacts. It doesn't really hit as hard as I saw Garrett pouring his soul into this feature and now it's live and it's working. Let's go. It energizes you.
Turner Novak:
So it seems like summarizing what you said, you probably as a founder, founding team, what do you prefer? What do you get the most energy from? How do you work better and whether that's remote in-person, hybrid in San Francisco, in New York, in Bali, some people work in Bali. Yeah.
Paul Klein:
Wherever. Yeah.
Turner Novak:
Okay. Yeah, that's a good way to think about it. And then you've hired people from Hacker News before. Do you DM them or is it you posting? How do you do that?
Paul Klein:
I love weird sourcing ways. I love trying to find people in all these different areas. Hacker News ... And I'm going to out myself here. It is the best place to find great engineers. And you could do this in two ways. One, they have regular hiring threads where they say you post like, "Hey, I'm hiring." Sometimes the opposite side of that, who wants to be hired? And there's a lot of diamonds in the rough there. Yeah. I'll spend hours scrolling through and reading each person's profile and going to their personal website. There were just some fantastic engineers who are just throwing their name out there.
Turner Novak:
What is the thinking usually of one of those people? Is it a big tech engineer who's underutilized and they're just like, "I want to try something new."? Is it somebody who hates their job? Is it college kids? What does the pool of talent usually look like that participate in those?
Paul Klein:
The funny thing, it's all over the place. There's people who've worked in big tech. There's people who are career technologists. There's people who go to Burning Man a lot. There's people who never leave their house. You got a full spectrum of people who are really in the end passionate about technology. And the other way I'll do things is I'll try and find Hacker News conversations that are relevant to our space and I'm going to try and find the people who have the spiciest takes and I'm going to email them and say, "Hey, I saw your comment on this. I agree or I disagree. Do you want to have a chat? What do you do? Where do you work? Do you want to come work for us?"
Turner Novak:
Would you ever hire someone that disagreed with Browserbase? They're like, "This is not going to work. I don't like the architecture. Bad idea."
Paul Klein:
No. I wouldn't hire someone who disagrees with the vision of the company. I would hire someone who disagrees with me. It happens all the time. I constantly get in disagreements with the team and that's how we have good decisions. We're not like a culture of we all agree with each other all the time. So I actually like people who have strong opinions. Ideally they're loosely held, and if we can disagree and commit and build something together, it's good. But if someone does not believe in Browserbase being a company that will be successful, then I don't want you to come work for us. If you really want to work here, it'd be such a weird thing. I hate you, but I want to work for you. Might be toxic. I don't know.
Turner Novak:
I feel like you get some of that with some of the really big tech companies when you're crossing thousands of employees where they don't like the company.
Paul Klein:
Well yeah, we don't pay well enough, I guess, to get those problems. Maybe if we're paying millions of dollars, it'd be like, fine, I'll do it. But no, thankfully we are still a humble small business as they call it.
Turner Novak:
As they call it. And so I want to actually touch base on the contractors again. It seems like how do you get good quality out of contractors and maybe contractors is a scary word. Part time. I don't know the best way to say it that's not derogative. Or you maybe think lower quality work. Can you get good work from it?
Paul Klein:
Yeah. So maybe defining what is a contractor? A contractor is somebody who I pay generally on an hourly basis, sometimes on a weekly or monthly basis who gets to choose their own hours. They get to work when they want and they can leave at any time and we can part ways at any time. How is that different from a full-time employee? Well, I think it's just a different relationship where they can say, "Hey Paul, I'm going to go to Ecuador for six weeks." I'm like, "Sick. Sounds good. I'll take care of it. Don't worry about it." An employee, they need to let me know. We need a plan. Is this more of a permanent relationship there? Now how do you hire good ones? Well, that position in the description is pretty high trust. So ideally you're only working with people as a consultant or a contractor that you trust and you trust them to go off on their own and build something. And generally, you'll know within the first few weeks of working with someone who's a contractor that they're super high trust, high integrity and care. So I think that more often than not, people end up hiring contractors. We've hired people who are contractors full-time before. It's a great strategy. And I think it's just something that you have to take some risk. Ideally you're hiring people to be contractors or people you've heard good things about.
Turner Novak:
Yeah. And you've said you don't really like hiring mercenaries.
Paul Klein:
Yeah. Yeah. Going back to what I hired the person who doesn't believe in Browserbase who's really good, no. I think that to do good work at our company, not only do you have to be highly skilled, but you have to believe in what we're doing. We just had our monthly all hands on Monday, and the thing I left everyone with was, for us optimism is the force multiplier. If you believe in what we're doing at Browserbase, it's going to enable the work you do to be more successful. If you've ever done something before and you're like half a believer in it, the quality just goes way down. So we have to have everyone on our team really believe that we're doing this thing and it's going to work, otherwise we're not going to be able to endure all the hard parts of building a startup.
Turner Novak:
Yeah. So you're saying there's an unlock of productivity or skill or quality where just if you're really into it, you really enjoy it overall, do a better job.
Paul Klein:
The placebo effect. If you took a pill and it said you're going to be better at your job and you don't know if it does or not but you believe in it and you actually end up doing better. I think same thing. If you're optimistic about the outcome of this thing, you're going to end up taking little micro actions that will result in you making the company more successful. Maybe you'll push a little bit harder, or maybe you'll say something you think is risky, but you think it's really important. Those are the conversations that define a company.
Turner Novak:
Yeah. That's true. You might have a really, I don't know, different or difficult to voice or share opinion that's pretty important but you still want to keep making 200 grand a year and you don't want to have to look for a new job. So you might just be like, "I don't know. I'm not going to say anything about this." But if you really care about it and you're really important or it's really important to you, and maybe you not only get the courage to say it, but actually, because you're so into it, you can phrase it better and you just put more into how you pitch people on it or something like that, or share the opinion. I can see that too.
Paul Klein:
We are an emotionally vulnerable company. And what that means is that we talk about the hard things. I think when you're in person, you can pick up when someone's having a bad day, you can just sense that, you notice that, their body language, the way they handle themselves. And it's on us to really try and one dare to trust each other. It's risky to trust somebody. If you trust somebody, they let you down ... You have to do this bold thing of trusting somebody. And then you have to actually share vulnerably and own your mistakes. If someone trusts you and you make a mistake, you better earn that trust back by admitting, Hey, here's what I messed up. And we really focus on how do we make sure we're communicating and creating an environment where people can say these hard things like, "Hey, I actually maybe disagree with the direction of this product right now. Let's have a conversation about it." And I think having everyone talk about it will actually result in better outcomes. We do this thing where whenever we are discussing a new big feature, we do a request for comment. It's called an RFC, and everyone comes in and comments on it. And listen, we pepper this thing with comments. We're really getting into the weeds about this thing.
Turner Novak:
Good, bad-
Paul Klein:
Yeah. Why would you name this thing this way? It's your job to remove your ego from the work and also look at this thing and really criticize it. And that's how you create great work. But that's super hard. People put a lot of effort into something and for others to come in and say what's bad about it. Calling your baby ugly. You have to have a level of trust that these people behind the scenes really deeply care about you and trust you to be able to take criticism. So we start with building that baseline belief that, Hey, Garrett actually really cares about me. He likes me. He's going to rip my work to shreds, but that's because he wants us to be better.
Turner Novak:
Garret just loves me.
Paul Klein:
Yeah. He cares. It is like if the coach stops coaching you at practices because given up on you, that's not something you want. You want someone who actually cares enough about yourself to really get into it.
Turner Novak:
Okay. So question then, people caring about your stuff. How do you get people to care about Browserbase and maybe developer tools in general? What have you found works best there?
Paul Klein:
Yeah. I'm really lucky that we've gotten to a place with our brand that I'm starting to see the results of people recognizing it outside of me putting it in their timeline.
Turner Novak:
And it's the bee, the bee emoji.
Paul Klein:
The bee emoji. That was the first growth decision of what's your logo going to be? And I think when I was talking with Zack Hargett, who's like I said, one of my buddies who helped me on the brand, we were looking like, what are the brands that we love right now? And LangChain, their original logo was the chain emoji and a parrot. Did you know that?
Turner Novak:
I didn't know that was the original. It is still their logo.
Paul Klein:
Well hey, it still looks like the bee emoji. You start with a simple thing. Well, let's just make an emoji for our logo. We got to keep building. And that's what we needed in the very beginning. But I think ... And I forget who it was, but someone talked about owning the emoji. If you can own an emoji, it ends up being a really powerful growth thing because your logo is in everyone's pocket.
Turner Novak:
I've heard Austin Rief talk about that on Morning Brew with coffee mug. People have said that I've done that with banana. I thought about it for sure. I was like, "Oh, this is interesting." It wasn't the reason, but it works. Having an emoji.
Paul Klein:
I think owning an emoji is very powerful. I think for me, I liked the name Browserbase because it was easy to search. We could get the .com, and I thought the SEO would be good.
Turner Novak:
Really? What was the SEO like?
Paul Klein:
So when I was at Mux, the company before Browserbase that I was working at, I was helping out with SEO and I always would lose to a company. Mux is a video API and they help developers store videos and play them back. And I would always be losing an SEO to this startup called api.video. And no matter what I did, I could not beat them because they had-
Turner Novak:
That is incredible.
Paul Klein:
Because their name was video in the domain. The company name was freaking video. So I was like, "I'm Browserbase and I'm going to win because I'm going to have browser my name and I can beat you now."
Turner Novak:
Understood.
Paul Klein:
Yeah. That's proven to work because it's a base for your browser. It's a very simple name. Maybe we'll have to rebrand in five years, whatever. But for now, the way that we've gotten distribution is really by trying to communicate this esoteric thing, which is a headless web browser. These are things that people don't think about. It's like a very unloved tech stack. But I wanted to become the browser guy, the headless browser guy of San Francisco. If everyone in San Francisco knew that Paul was the person to talk to, if you had problems with headless browsers, I believe that we would win 2024. And exit into the year, I feel like we've really accomplished that. And we did that through a series of really great landing pages, good launch videos, like compelling tweets and compelling demos that really all came together to really show, hey, it's possible for your AI to use a web browser to do things on your behalf. Browserbase is the first thing you want to use when you do that. When you make that decision. So it all comes down to focusing on geography, focusing on a problem, and just really owning something. Building a reputation for owning something.
Turner Novak:
So when you talk about, you did a couple landing pages, videos, any examples that stand out of the greatest hits? Maybe the things that got the first customers, the thing that got you the biggest, just move the needle the most.
Paul Klein:
There's three videos that basically have made Browserbase what it's today. The first one was one I shot in Solaris, which is a coworking space in San Francisco. And this video was hilarious. It was like me talking at a camera. It was the first time I ever did a startup launch video, and I just did my pre-seed, and behind me there's a whiteboard that says, "Secret plan." I got an American flag on my right. There's a lot of little Easter eggs.
Turner Novak:
The American dynamism?
Paul Klein:
Yeah, it was right when all the Gundo boys were coming out, right. I was like, "All right."
Turner Novak:
Oh, you got their likes and their retweets.
Paul Klein:
Exactly. I was like, "Hey, we're engineers." Engineering is blue collar. We're engineering some stuff.
Turner Novak:
We're working in the code.
Paul Klein:
Yeah, exactly. Threw up the American flag and had the secret plan behind me, and I just kind of talked about the problem to a camera. In hindsight, the video was kind of cringe, but I think it channeled the right elements of it was slightly produced and edited. My buddy Austin shot it and edited it for us, and it looked good. The messaging was clear, which is like, "Hey, I'm planting my flag and building this thing. I'm calling my shot. If you want to come join me, come join me."
Then we hired somebody off of that video, and we got a bunch of customers off of that video. It got 100K views on Twitter, which I thought, I was pretty happy about. Then all of a sudden, I realized every launch that's major we should do, should have a video. The second video was kind of like a montage of short clips. I think Antimetal did this first really well when they did this launch video where they had all these clips.
Another company recently did it too, but it's like this theme where you see people doing launch videos that are TikTok edited, where it's like clips from TV shows that are all kind of stacking on top of each other.
Turner Novak:
Moves really fast-paced. There's a narration over top kind of thing.
Paul Klein:
Exactly. Yeah, yeah, yeah.
Turner Novak:
I know what you're talking about.
Paul Klein:
That was our seed announcement where the opening scene was a scene from iRobot and it's like, "Why has AI not lived up to its hype?" Then the second scene is Roombas crashing into each other, and it's like, "AI has let us down," and there's an Alexa, "I'm sorry, I cannot do that." It was a very meme-y video. That one got 250K views announcing our seed.
Then the Series A video, which I think was our best one, we actually went all out for. We had someone produce it, we hired an actor who's this AI [inaudible 01:00:04] guy we found to do there. He's sitting there, and you can hear Elon Musk in the background like, "AI is going to be the most powerful thing that happens in this decade."
He's building a startup with Browserbase, and the startup goes viral. You see the story of like, "What will you build? The next billion-dollar company will be powered by Browserbase." I think we've gotten a lot better at storytelling for a category that's not necessarily known for storytelling.
Turner Novak:
Yep, browsers or infrastructure.
Paul Klein:
Dev tools and infra. However, I'll kind of contradict myself. I think Vercel and Guillermo is fantastic at storytelling, and I think developers are choosing tools because the story resonates with them. For me, I wanted people who were starting their own companies to use Browserbase as a core part of their infrastructure. That's what we made that last video.
It's really been about how do we story tell for our category and our problem? It's not something that people really know about. I think we've done a good job of really representing it.
Turner Novak:
These are all on Twitter. Was Twitter pretty important? What's the power of Twitter for developer tools or for a founder in general?
Paul Klein:
Yeah. It's the only platform that I've been able to crack so far.
Turner Novak:
Have you tried other platforms?
Paul Klein:
I'm so bad at posting on Hacker News. I don't know if I'm shadow-banned or what happened, but I've never cracked Hacker News in terms of posting. I've cracked it for hiring, but not for posting.
Turner Novak:
Talk to Joseph at Roboflow who introduced us.
Paul Klein:
He's amazing. He's amazing at Hacker News. I'm super jealous. I think, I don't know what it is. That's something I haven't cracked. Because we're a dev tool, I haven't found that short form video on contents like Reels, YouTube Shorts, or TikTok is worthwhile yet. I think some people have really cracked that as a channel.
I think you have to pick a channel and just own it for a while. If you try and go multi-distribution channel too early, you end up kind of being bad at all of them, I think.
Turner Novak:
Yeah, I can see that.
Paul Klein:
If you pick one and crush it, then I think you really win. That's just kind of been X for us.
Turner Novak:
Is there a risk to being too over-reliant on it, you think?
Paul Klein:
I think that we haven't tapped out the... I don't think we've tapped out the TAM of X developers yet. I think once again, we're selling to startup, we're selling to the decision makers, right? Who's picking, who's deciding to buy Browserbase? It's probably the CTO, the CEO, head of engineering, or an early stage founder. Oftentimes, people in the startup ecosystem are on X.
Now, if I go sell to JP Morgan, I don't think Jamie Dimon is going to see my X launch video. I'm going to have to go on a Madison Square Garden ad on the side of the building or something, right?
Turner Novak:
I bet he has a burner.
Paul Klein:
Maybe, right?
Turner Novak:
That's the thing I've found is the most unexpected people will follow me or reach out to me, email me.
Paul Klein:
I think we just haven't tapped out that category yet. I think at one point, I'm excited for us to start doing more video content on YouTube, and more like how to build this thing. One of our partners, Theo Brown, he does this amazing job of these videos that walks through the pros and cons of technology.
We've been lucky to sponsor some of those videos. I'd love for us to have our own in-house video team producing stuff for developers, because video content's just super great to consume on a technical level.
Turner Novak:
Yeah, Joseph, again, that was one of the first things I think they did. They do a couple a week, couple usually videos a week, I think, for Roboflow. Another...
Paul Klein:
Hey, they're a series B company. Cut me a break. All right. We're a level behind them.
Turner Novak:
He was doing it years ago.
Paul Klein:
No, you're right. No, Joseph is a much better founder than me. I won't deny that.
Turner Novak:
He also connected us. Just we'll give him a shout out for-
Paul Klein:
Shout out Joseph.
Turner Novak:
When I had him on a couple of weeks ago, I was just like, I ask most people, I'm just like, "Anyone else you think I should have on?" He mentioned you, so I was like, "Holy shit."
Paul Klein:
It's amazing that I think one of the biggest hacks that we did in the early days of Browserbase is actually sharing offices with other great companies. Joseph knows this, but I always wanted to share an office with him. To me, he's one layer ahead of us, one level ahead of us.
Being able to see how that team operates as they just announced their series B, like it's just really fantastic to see great founders and try and learn from them. I surround myself with great founders all the time, as much as I can. That's why I learn most about how to do this job.
Turner Novak:
That to me sounds like an interesting hack of solo founder, share an office with maybe other solo founders.
Paul Klein:
Ideally someone ahead of you, a little bit further along. I think sometimes, especially in the early days, you'll get trapped in... You only perform as well as your expectations. If you see another startup crushing it, you're going to want to do better. This is actually a great story. When we were doing Stream Club, we had an office, it's actually funny, this is now Jack Altman's new office.
It was my old office for some reason, and it's in The Mission and it was us. Then we sublet to another company called Hightouch. You've heard of Hightouch maybe?
Turner Novak:
Oh, yeah.
Paul Klein:
Yeah. Well, back then, we were about the same size. We were both kind of early stage pre-CEC companies. Something happened where Hightouch started multiplying. They took up another desk, then they took up another desk. Then all of a sudden, there was more Hightouch people than Stream Club people. Then all of a sudden, it was the Hightouch office and Stream Club was subleasing from them.
I was just showing up to work every day, I was like, "Why are we not growing like this?" I think working with other talented founders and seeing how it happens up close, Tejas and Kash are still great friends of mine. They've crushed it since then, but it was a big motivator for us to work harder. Seeing when things are working, it is really clear, clearly different from when things aren't working. You want to see things are working with your company relative to others as much as you can.
Turner Novak:
This was when you talked about Alana at Base Case, one thing she actually told me to ask you, did you learn anything from your parents in terms of sales?
Paul Klein:
Yeah. You know what I realized? You're the Nardwuar of Tech podcasts.
Turner Novak:
I actually don't even know who he is.
Paul Klein:
You don't know Nardwuar?
Turner Novak:
I've heard people mention him before. I've never watched him.
Paul Klein:
He does these rap interviews and he finds these really deep cut questions that require inside information.
Turner Novak:
We got to build unique content. If I've learned anything about media companies or anything media related, you need unique.
Paul Klein:
You kind of look like him a little bit too, actually.
Turner Novak:
Okay, I got to figure out what he looks like.
Paul Klein:
Yeah, I got to look it up. You asked-
Turner Novak:
What did you learn from your parents about sales?
Paul Klein:
Exactly. I'm, as a founder, as a CEO, you're always selling, right? You're selling customers, you're also selling candidates. I'm lucky that both my parents are fantastic salespeople. My mom has a background in pharmaceutical sales. My dad, he did semiconductor sales, so-
Turner Novak:
He at NVIDIA?
Paul Klein:
It was not NVIDIA. Aero Electronics.
Turner Novak:
Oh, okay.
Paul Klein:
Space.
Turner Novak:
Aero Electronics. Interesting, never heard of them.
Paul Klein:
Yeah, yeah, yeah. They sell semiconductors. Both of them, I kind of grew up being around salespeople. I think I probably was being sold to my whole life, like, "Oh, my God, you're going to love the Olive Garden we're going to, it's the best breadsticks you'll ever have," right?
Turner Novak:
Such a great family restaurant.
Paul Klein:
Subconsciously got used to being sold things and then reusing that. I think that it wasn't as much active learning, but I certainly imagined that I personally loved computers. I was very introverted. I just wanted to play video games all day long.
Having two extroverted sales manager parents probably rubbed off and created me, who I am today, which is extroverted, love to talk to people, but I really deeply love technology. I can talk to you all day about headless browsers, so it's a weird...
Turner Novak:
Deadly combination.
Paul Klein:
I know, weird hybrid. I think that's what made me really happy to be in the position I am, which is being able to run this company, and balance my love for technology, but also the need to share that story and really rally people around it.
Turner Novak:
What video games did you play growing up?
Paul Klein:
Oh, everything.
Turner Novak:
Like top, most time spent?
Paul Klein:
Most time spent, StarCraft, I got really good at Starcraft II, I rated in WOW. I was actually running a guild for a little bit. RuneScape, I did so much RuneScape. I did some RuneScape botting, actually, which I actually think taught me a lot about botting.
Turner Novak:
How is that different or the same? Is there scraping and stuff like that, or is it more automation?
Paul Klein:
Very similar. I think back then, RuneScape botting, you had to download a sketchy EXE file and run it on your computer. I think I got so good at computers because I definitely infected my family computer like 50 times, and learned everything about how to fix when you're...
Turner Novak:
A RuneScape virus.
Paul Klein:
Yeah, we have a root kit. You learn a lot about the command line, because you're like, "Mom's coming home in 30 minutes and I got to remove this keylogger." All the pop-ups that are popping up on the computer, you have your RuneScape virus. I think I've always loved games where you're setting up systems and then you're seeing systems work, so like Civilization. Yeah, it's been a big part of me. I haven't gained that much. I really want to get a PS 5. That's my... I haven't had a console in years, so watch out.
Turner Novak:
I played a lot. I played a lot of Civilization. Was when Civilization 3, I think that's the one it is, I beat it on the hardest difficulty level. I don't know if you've ever played it on DyP.
Paul Klein:
Never.
Turner Novak:
It's impossible. All the sliders for the computer tweaks all the way over, they just destroy you. I've beaten it on that level before. It was the most fulfilling accomplishment I've ever had. Then I played a lot of Halo 3 and then Call of Duty 4. I went to a Halo 3 tournament, actually, when I was in college, it was after my freshman year. I got to the highest level in the hardest playlist in Halo 3. There's this thing called Major League gaming, MLG, this professional league back in the 2000s.
Yeah, they did these in-person tournaments and I got to the highest level in the MLG playlist in Halo 3, and I was like, "Cool, I beat the game. I should probably focus on college," because it was freshman year. Then a kid in one of my classes was like, "Oh, you're wearing this MLG hoodie." We started playing together and like, "Dude, you're so good. We got to go to some of these tournaments."
It was kind of like maybe you leave retirement kind of a thing. It was maybe like what you did, where you were like, "I'm never playing again." We went to this tournament and we just got absolutely destroyed. It was double elimination bracket. We just lost every match, got knocked out the first day. Then most recently, I used to play NCAA Football with my brothers a bunch. Did you ever play that one?
Paul Klein:
I did, yeah.
Turner Novak:
The last one they made was 2014 and they made a '25. I don't know if you-
Paul Klein:
The '25 just came out. I know.
Turner Novak:
It just came out over the summer.
Paul Klein:
It's been fantastic. I don't play it, but I've heard good things.
Turner Novak:
Yeah, so I actually, I got it. It's been the first video game I've really played much in like a decade. It's very nostalgic, anyways. That one, I don't know. I don't know if it's a systems thing. It kind of is, though. I don't know.
Paul Klein:
Yeah, you're running plays, you're making decisions that have outcomes, and then making better decisions in the future. There's some mechanics of course, but in the end, you have to make the right calls.
Turner Novak:
I do the dynasty mode too, where you're recruiting the team. I usually pick the worst, I did Ball State and I took them and was winning the national championships.
Paul Klein:
Very VC of you. Yeah.
Turner Novak:
Heck yeah.
Paul Klein:
You're on the board of your NCAA team. LPs, check out his team. If he didn't win the championship, I would invest personally. It makes it, similarly, I think people who love video games, to me, the startup is a video game in a lot of ways. It's a system, and we kind of constantly get to look at how we improve. There's some skill here. You have to get good at writing code.
Turner Novak:
You have to grind too.
Paul Klein:
Yeah. Oh, oh.
Turner Novak:
You just do the same thing over and over again.
Paul Klein:
Yeah, and I think for us, the grind is definitely part of it. I had this rule, I would never ask someone to do something that I wouldn't do. I think the founder always has to be the hardest working person in the company, and I try and set that bar. I work a lot, but I love that, and I think the grind is very fulfilling for me because you get to kind of see incremental progress.
On Friday, I spent all this time investigating this little tiny bug that was impacting this one customer on this one specific website. I found out what it was, and that was worth the whole day of finding that bug and be like, "Hey, I found the bug." I felt like a hero. I think if you don't enjoy those little moments, a startup is going to be such a hard experience, because you have to take joy in the little tiny grind achievements. It was like I got level five in Fletching in RuneScape and I was like, "Let's go."
Turner Novak:
I just had a really interesting recent example, my friend, his name's Antoine Martin, he has a company called Zenly. He sold it to Snap, started a new one. Same thing, it's like a social map. It's like Find My Friends, but way better. I showed him this one thing. I was like, "I'm getting this bug. Here's a screenshot," constantly texting me about it and I'm like, "I feel so bad." I am not just retesting the app and seeing if it's working, and he's sending me the Slack screenshots of everyone debugging it, and their logs, and they're trying to figure out what's going wrong with it.
I take five hours to get back with him. I was like, "Yeah, I tried it. It still didn't work." Just constantly giving me the feedback on how to get through. Then I think it took us basically two days to fix it. It was probably because it was always for me taking a while to redownload that app from the app store and try it again. I feel like you got to just really enjoy doing that stuff.
Paul Klein:
You have to have no shame as a founder too. When we launched Stream Club on Product Hunt, I texted pretty much everybody I knew. I went on Facebook and it was back when Facebook Messenger would show green dots next to people who are online.
Turner Novak:
Oh, no.
Paul Klein:
Everyone who had a green dot. I messaged them on Facebook saying, "Hey, we're on Product Hunt right now. Would you mind upvoting us? I know you're online, because I see your green dot. It would just be super helpful." I texted high school ex-girlfriends, everybody. Everybody that would maybe try and do it.
You have to be so shameless with your social capital as a founder. I think if you can do that, then you can actually hustle to get your startup out there and get people to talk about it.
Turner Novak:
How do you get by or get past... That's really cringe.
Paul Klein:
So cringe.
Turner Novak:
How do you go from, man, this just feels so cringey to actually doing all that stuff, putting yourself out there? I'm sure that first video you shot with the Secret Master plan, there's probably people that saw that and thought, "What a loser," or something. How do you get past the putting all this stuff out there to now, you're at the point where it's worked, it's awesome. I'm doing it again. I'm being super shameless.
That's a common question I get about Twitter is I'll say something that's just dumb and just, or I'll see the wrong thing, or people will make fun of you, or they'll judge you or whatever. Is there a barrier, and how do you get past that?
Paul Klein:
I think it's deeply psychological. You have to accept, one, you are an imperfect person. I'm an imperfect person. I'm going to say something wrong. I'm sure I've said something wrong in this podcast, but I make mistakes and I own that. At the same time, I try and be as authentic to myself as possible so that if I say something, as long as I'm being truthful, always, I am consistent.
I think it's when you try and act as someone you're not publicly, it becomes a lot of pressure to keep that up. I think you have to actually like yourself as you are in an authentic way to be able to put yourself out there on a stage and say, "Here's who I am." I made a joke on Twitter about, I tweeted a picture of an In-N Out Burger and I was like, "Having my daily dose of microplastics." Then people were like-
Turner Novak:
I saw that, yeah.
Paul Klein:
I was like, "I slayed the burger, it was fantastic." They're like, "Oh, my God, the seed oils, what are you doing?" Just people on the internet are crazy, and as long as you like the stuff that you're doing, everything I post is for me. I genuinely like my own tweets. I don't think I actually clicked the button. I might be able to start doing that.
Turner Novak:
You actually should. No one can see them.
Paul Klein:
Great.
Turner Novak:
Just get the social proof of there's already one like.
Paul Klein:
I'm going to start doing that because I won't tweet something that I don't like. That's actually a way that you can be comfortable. If somebody says, "This is bad," and you're like, "Well, I don't care what you think as much, because I like it." It removes the pressure of doing things for others. Now, it's tricky, because as a founder, you have to produce content that resonates with others, and you have to actually have outcomes from the things that you do.
If I put out a launch video, it needs to have views and stuff. I think in the end, that requires you to take feedback from the environment, feedback from the people who consume your content, and then take a lesson. If I do a music video for Browserbase and people say, "Oh, my God, this is so cringe," and I love it, well, I don't know if a music video is landing in terms of something that I want to do to promote our content, and I should maybe reconsider that medium next time.
Turner Novak:
Yeah. One thing that resonates with me when you talk about the joke about the microplastics in the burger, that might've been a joke you would've made in the group chat that 10 people would've seen, but you just put it on Twitter, and I don't know, maybe a couple thousand views, whatever, depending on how I could get a million views, and you could get 10 new customers, or someone who wants to work at Browserbase from that.
It's hard to quantify the benefits, but also if you're doing something that you maybe already are doing anyways, or you already enjoy doing and you get fulfillment out of it, it's not that difficult, I guess. You have to train your psyche, and your mind, and workflow around it.
Paul Klein:
Yeah, and there's hacks and techniques. I did this thread about like, "Here's all the things." I had a list of learnings I had from the last year, and at the end of it, every investor update, I share a lesson I learned. I put that list on X. Yeah, I kind of formatted it.
You have to click the show more button. I know the algorithm would prioritize that, because people spend more time looking at that tweet, and therefore, it would spread-
Turner Novak:
Yeah, it's two minutes. If people spend at least two minutes on a piece of content, it tips it in the algo.
Paul Klein:
Exactly. Yes, there is some strategy to it, but I think I also was just like, "Oh, I'm really happy I wrote all these things down, I'll share them." I think there's a blend of authenticity and execution that need to come together to create good messaging that people actually resonate with.
Turner Novak:
I guess cycling a little bit, but maybe related to just the internet being always on, how important is customer support and customer service, especially as a startup?
Paul Klein:
Yeah, especially as an infrastructure company too, our customers build features that depend on our thing being online, and that's a high bar. Being an online, doors open 100% of the time, even if you're sleeping, we really need to make sure our stuff works well. It's also complex, because our customers use Browserbase to go browse the entire internet.
Websites on the internet also may be less reliable than us. If someone goes to a website online and they're like, "Hey, how come this random website's not working?" Is it Browserbase or is it the random website? We've had instance where the random website is down. There are some health insurance portals that just stop working in late at night.
Turner Novak:
Really?
Paul Klein:
Yeah.
Turner Novak:
Like 90% uptime.
Paul Klein:
They just shut it down for maintenance. If you want to go file an insurance claim for Aetna at like 3:00 AM Pacific Time, I'm sorry, dog, it's down for maintenance.
Turner Novak:
Oh, man.
Paul Klein:
People will wonder if it's our fault or their fault. What I've learned from the great founders out there is that customer support is the biggest differentiator you can have. People have chosen to work with us because they have my cell phone number and can get on a call, and they can come talk to me. I think I've really tried to put that into our team.
I read this interesting thing about Super Base recently, it's a database company. At the top of every offer letter, they say, "Customer support is your number one responsibility."
Turner Novak:
Really?
Paul Klein:
Every single offer letter. Guillermo from Vercel, he says, "Customer support is our biggest differentiator." I think it's not a mistake that each of these great infrastructure companies make that a priority, because developers are counting on you. It's similar. We all know how bad airline customer support is, and we all feel so screwed because we have to get home on that plane. We have no other options.
If your feature's not working because your provider is down, you feel like things are out of your hands as a developer. I think that's why we try and be as transparent as possible. We have a status page where we post, "Hey, if we go down, here's why we went down." Remember when OpenAI, they had a big outage recently, and I think there's a lot of stuff they can improve from that, but the postmortem that they posted, they said, "Here's exactly what went wrong with our system."
That helped them earn more trust back from the engineering community. We've all been in a place where, in this case, OpenAI, they had a runaway Kubernetes cluster, and this configuration change broke it. Many engineers feel empathy, because we've had that happen for us too. Above all else, it's trying to be transparent, trying to be available, set expectations accordingly. We closed our biggest sales deal, which involved me supporting a customer the day before my wedding this year.
I got a call from Sophie on our team, like, "Hey Paul, can you take a look at this? I know you're about to get married tomorrow, but it's kind of important." I'm like, "Absolutely. This is super exciting." Sometimes we have to make sacrifices to meet our customers where they are. Just because stuff is happening in our life doesn't mean that their lives are stopping. We should do our best to prioritize them, because they're taking a big bet on us. I'm just super grateful for them.
Turner Novak:
Yeah. I think another interesting topic, if I Google you or shoot you on YouTube, I look at some of these other podcast conversations you've done, it seems like a lot of people are really excited about, "This guy's a great fundraiser." Zero to 27 million raised in nine months, I think is the headline.
I think you kind of described it to me before as fundraising as a non-famous person. I think those was your phrasing. How do you fundraise as a non-famous person?
Paul Klein:
Yeah, I am not [inaudible 01:22:26], I'm not Elon Musk.
Turner Novak:
Maybe you're famous now because of those headlines.
Paul Klein:
Yeah, I think it's definitely a good click-baity title. For what it's worth, we've raised a lot of venture capital in a very short amount of time. We raised three rounds across nine months, 27.5 million in a 2 million pre-seed, 12 and a half million seed, and 21 million series A. You don't do that that fast without learning how to fundraise. I think the lesson I've learned with fundraising is actually build relationships before you fundraise.
I've been lucky to have had, I'm a one-time failed entrepreneur, so second time founder, one time failed. Since the first time, I've gotten to know a lot of investors. I've spent time talking to them about my ideas. I was talking to a founder recently and he said, "Oh, you're supposed to never talk to VCs. Talking to VCs is like talking to the cops."
Turner Novak:
Talking to the cops.
Paul Klein:
He's like, "They're going to pull out their little notepad, and everything you say is going in the record." I was like-
Turner Novak:
Yeah, I can see that, I guess.
Paul Klein:
I was like, "I have a rap sheet, I guess, because I'll talk to people."
Turner Novak:
You got a whole Palantir database.
Paul Klein:
Yeah, on Paul's opinions is there in the rap sheet. No, I completely disagree with that. It's not like the cops.
Turner Novak:
Isn't it a waste of time? Don't people say it's a waste of time?
Paul Klein:
I feel like people ask me interesting questions that helps me think better. When I wrote the original memo for Browserbase, I sent it to a few investors that I trusted, said like, "Hey, tell me everything that you would criticize about this."
Turner Novak:
It was sort of also a request for startups, almost?
Paul Klein:
Yeah, the original memo was like, "Someone should build this." That someone was very clearly me, and I think I was employed when I published that, so I didn't want to say, "I'm starting this company," and call my boss the next day like, "Hey, by the way, I just published this." I think I still had to come to terms with it.
Yeah, when I published it, it was like a request for startups. I think I had to convince myself that I was the right person. In the end, I think investors, just like founders, are people too. You should not be afraid to talk to them. Oftentimes, they will ask you interesting questions that make you think about things.
Turner Novak:
How did you, let's say, I'm assuming the very first round you raised, there probably wasn't a whole lot yet. The second one, I think it was 12 million was the number you said?
Paul Klein:
Four and a half for the second one. There also still wasn't a lot. We had a product, we had a lot. We'd validated the initial hypotheses, so well, maybe I'll tell it step by step. The pre-seed was really reputation-based. I had gotten to know Atlanta. She really understood deeply my vision for this company. She'd seen me execute at the current job I was at, which I'd done a good job at my job. I think she really believed in what I wanted to do, and said, "Hey, I see your vision here, go do it."
Then for the seed round, Bucky Moore, who's on our board now at Kleiner Perkins, he really got to see, "Hey, when I talked to Atlanta in January, here's what I said I wanted to figure out. Here's all the evidence that this is actually happening that we've seen in the last four months. We need to double down. Let's go do it." He was like, "All right." Thankfully, I had met Bucky for a random walk two years before.
We'd gotten a chance to get to know each other. Then when I did the series A, Reid Christian, who's from CRV, who's also now on our board, I'd had a chance to hang out with him a year and a half before, because I just said, "Hey, I like your stuff on Twitter. You're investing in Vercel? I love Vercel. Can we talk?"
Turner Novak:
It was that simple.
Paul Klein:
I think for all of these things, there's a combination of you're not trying to get someone to invest in the first time you've met them. There's some time between that. You've actually built a relationship with them, and you also have somewhat of a good idea that's related to a thesis that they have. All these investors have invested in great [inaudible 01:26:29] and infrastructure companies.
I'm building a category-defining infrastructure company, and I've taken the time to get them to trust me because I've tried to get to know them. I wrote about dev tools for a year on my own personal blog, and I sent it to them before I even started the company.
Turner Novak:
To all these people, you had sent the thesis to?
Paul Klein:
I had sent several. I had a whole sub stack that I was doing. I was publishing every other week for a year before I started Browserbase. A lot of people had seen some of them. I had a couple of memos that really popped off. I had one called... It was "Next.js vs the world" that talks about how components are the new API. I had one about the rise of product engineers and how important they are for building good companies.
Turner Novak:
That one sounds familiar.
Paul Klein:
Yeah, I mean that one got a bunch of views. And I had one called "Death to the backend" talking about how I think people writing backend code will just be using infrastructure in the future.
Turner Novak:
Backend code is probably going to get just like-
Paul Klein:
Yeah, like Convex or a Supabase or Clerk, all these backend-as-a-service companies or companies that provide infrastructural components, I think. You don't need a user's table anymore, you just use Clerk.
Turner Novak:
And the engineers or PMs or other people on the team can just use their products instead of having a backend engineer be like one person that manages the products or something.
Paul Klein:
If you can generate your frontend with AI and you can just plug that into a backend that's run by some third party company, you don't have to maintain a lot of stuff anymore, which I think is quite cool. And I mean some people, we can go into this super deeply, but maybe the point of all this was I was writing about technology and about my industry, which is infrastructure, because I cared about it a lot and people had... someone says a lot like timestamped thoughts, you know?
Turner Novak:
Yeah. I've heard that before.
Paul Klein:
Yeah. These were all timestamped thoughts that I've had and it became a lot easier to kind of diligence me and say like, "Wow, Paul's really thought about this for a long time." When I raised the pre-seed for Browserbase. The second slide of the deck is a link to a YouTube video of me talking about headless browsers in 2021, like two years before. So you have kind of clear evidence that I've thought about this for a long time and a pretty valid direction to go in. And I think that's all investors look for. Tell me if I'm wrong here as you are a investor, but from my perspective they want to know why now, why this, why you? And if you can very clearly articulate why is now the right time for this thing, why are you building this thing, and why are you the right person to build it? I think you have a compelling case for building a good company.
Turner Novak:
Yeah, I think that's right. Another thing I think about is just will you make money? If I give you money, will you make me more money? And that's the simplest form of answering what you just said.
Paul Klein:
Yeah.
Turner Novak:
So it's like, I mean one of my friends, she's like, her framing is, do I think this founder is going to be a billionaire? I'm like, okay, that answers the question because I get a part of your company, you make a shit-ton of money, and I also make money because I invest it. But that comes down to what's the product, what's the market, is it the right team? I think one thing I've changed a little bit is less and less on me being super analytical or things that I think I'm smart about with the marketer or the product and more so about just the founder and the team and just ability to execute. I think actually one company I'm looking at investing and I know nothing about the market at all, I think the founders are really good. So.
Paul Klein:
Yeah, I mean for early stage, I do a little bit of angel investing and the biggest miss I've ever had, the founder came to me and said, "Hey Paul, I really want you to invest. I think you'll..." And I just didn't like the company. I was like, "Oh, this company doesn't seem very good, but you're awesome. So good luck." And then they pivoted of course because the company wasn't very good and their new company's fucking awesome and they're crushing it and I'm like, "Oh my god, how did I miss out on that?" And since then, if there's a founder that's really, really good, even if they don't have the right idea, I'm still going to... I'm going to tell them it's not the right idea, but I'm going to tell them I'm betting on them and hopefully they figure it out.
Turner Novak:
Yeah, I actually have done that before where I've said, "I am not sold on this, but I'm going to give you money because I believe in you."
Paul Klein:
Right.
Turner Novak:
I think, I say this a couple times on the podcast, but there's been a lot of times where somebody comes to you with a pitch deck and says all this stuff, whatever. What is exactly in the pitch deck is never the thing that works and creates enterprise value or whatever. Maybe they got the general customer right, maybe they got the general market right, maybe all those things. I had one, pivoted from the most B2C possible thing you could imagine to the most B2B possible thing, complete 180, I would call it a 540. Literally full 360 and back another 180. It was just a complete change in the product.
Paul Klein:
They're dizzy after that. Yeah.
Turner Novak:
But when I was investing I was like, I really like this idea. I just don't even think it's going to work because I think it's impossible, but I think he might be able to figure it out. But if he pivots to B2B, do I just trust him with the money and do I think that they're going to build stuff with it and make a company with it? So I don't know, that's one way I've been kind of just framing it in my head is just like, do I think that this team is going to make a company that does something?
Paul Klein:
Yeah, I think the best founders that I've met are relentless. They just will not give up. And as much as they... Startups are unpredictable, if things are going the way of the pitch deck and it's working, that's super special. I think we're lucky that at Browserbase it's been pretty much to plan the whole year and that's been just super lucky.
Turner Novak:
Does that scare you at all that it's going exactly how you thought or?
Paul Klein:
I think it's going better than I thought. I thought that hitting a million dollar run rate in our first year was the stretch goal and we crushed that. I think I underestimated how well it could work, like the upside. And I think that's why if you go back to like, well, Paul said he wasn't going to raise venture, Paul said he wasn't going to start a company. Why would I trust what this guy says? I think I try and be conservative on how successful things can be and I focus more about what could go wrong and making sure I really understand the things that go wrong super well. I'd be happy if I misread the upside and we're up a lot bigger than we actually are, but I would be less happy if I misread the downside and actually there's some structural problems with this business that we just invested a bunch of time and money into.
Turner Novak:
You don't feel like there's structural issues with the business model or products or anything?
Paul Klein:
At least from my perspective, I'm okay with any of the ones that exist. I think that it's really hard for a hyperscaler company to build a company like Browserbase. I think if Amazon launched a headless browser infrastructure product, I don't think they would win against us. I think that it would be, one, there's some problems that they have to overcome with dealing with anti-bot. That's really difficult for a company like Amazon to do.
Turner Novak:
They're like the biggest company in the world. Don't they have the resources?
Paul Klein:
Yeah, but I also think that they have a lot of risk. Imagine this New York Times headline, "Amazon Web Services Solves CAPTCHAs". It kind of looks bad. "Browserbase Solves CAPTCHAs". They're like, well, "Who's Browserbase?" It doesn't matter. There's no risk. Yeah, it's a terrible headline, but "Andy Jassy The Bot Master". I could come with 50 of these that would cost Amazon more in market cap of a risk for the stock then it's worth it to build that business. But for us, we can take risks. We're a small startup and I think that we're doing something that's on this new edge where people are still figuring out the rules of the internet, they're still figuring out how AI can go online and CAPTCHAs exist because they're kind of one-size-fits-all. It's like a CAPTCHA exists to prove that this is a bot, but now there are good bots and bad bots.
Turner Novak:
Yeah, there's lots of bots on the internet.
Paul Klein:
There's different types and CAPTCHAs are kind of like, oh, it's a bot. I don't know if it's good or bad. And I would love to make Browserbase the arbiter of good bots one year, one day. We're like, if a Browserbase browser comes in, you know it's a good bot and you want to let it in because it's going to go buy something, it's going to come help your customer do something, it's going to actually add more usage of the downstream product." But I think we have a ways to go before the internet catches up to what's happening with AI. And yeah, it's going to be interesting to see, but I think it's really hard for a company like Amazon or Cloudflare or really compete in that space because it's so high-risk.
Turner Novak:
And there's a couple other, doesn't Microsoft have one?
Paul Klein:
They have a framework. They have a framework called Playwright and that's the open source framework. Google has one too, it's called Puppeteer. And those are more control planes. So you use Playwright to write code that controls a browser. The browser in this instance would be like Google Chrome. So Playwright controls Google Chrome. So Playwright's the framework, the browser is the browser and the Playwright control the framework.
Turner Novak:
And then if you were to say use Browserbase instead because it can do this?
Paul Klein:
So you would use Playwright to control a browser that's running within Browserbase. So Browserbase is the infrastructure. It's kind of like, think about a database company. You write SQL code to talk to a database and there's many different types of databases. Postgres, mySQL, MongoDB. Playwright is to SQL as the browser is to a database. So Playwright controls the browser, but if you want to run like 50 or 100 browsers, it's actually quite hard to do that. And Browserbase makes it so you could spin up 1,000 browsers like this, super easy. And they have all this cool features, like we show you what happened in the browser, we have a recording, we have all those logs and stuff. So Playwright can talk to many browsers within Browserbase and we handle all those infrastructural challenges.
Turner Novak:
So are there any big product decisions that you think you got right or that you specifically thought about over the past year or two, three years?
Paul Klein:
Yeah, I mean for me, if you're building great infrastructure, you also need to have a great framework. And I think what Vercel did, had this kind of concept of framework defined infrastructure where you build a great framework and you build the infrastructure around it. The Browserbase infrastructure exists really to power our own framework, which is comparable to Playwright. It's called Stagehand, like an actor because there's Playwrights, there's Puppeteers, and there's Stagehands.
Turner Novak:
Okay, I got you.
Paul Klein:
The fact that you have to explain the name means it's a bad name, but we're stuck with it. So it's called Stagehand, stagehand.dev, and it is a AI web browsing framework. So instead of, in the old world, you'd have to say, "Hey, this specific button, click this specific button." It probably has a bunch of like, it's the fifth div on the page, it's two links below or two links to the right of the other link. It's very specific to the page. With Stagehand you can just say, "Hey, click the checkout button," and we will code gen the right code to find out where that checkout button is and click it for you. So Stagehand is this product decision we made to build a framework and own a framework that we actually think is the right framework to use when building applications that can interact with the internet.
Turner Novak:
And that makes it easier for people to adopt like Browserbase infrastructure?
Paul Klein:
It's much easier to build workflows now that go out and automate something online because you actually have a framework that's designed for that. These other frameworks, Selenium, Playwright, Puppeteer, these were all built for testing. They're all built to test your web app. As a developer you would use Playwright because you want to have it go through the login flow 10 times and make sure it doesn't break. But now there's a framework, Stagehand, which is built for as a developer, I want to build some software that's going to go to the IRS website and find the tax code for some random state and investigate whether or not farmers should get taxed for the corn they throw away. I just made that up, but I mean it's an interesting thing that you have to write that for every single state, that's 50 different scripts. What if you were to write one script that uses AI to say, okay, this customer is in California, I will scroll on this list until I find California and click into that? And no, California doesn't tax the farmers for the corn, obviously.
Turner Novak:
Are there any things it doesn't work for yet or it's not good at yet? And maybe the things you're excited about, "Yeah, I think next year we're going to get there?"
Paul Klein:
I think that the biggest thing holding us back for building reliable web agents, a web agent is like a AI that can navigate the web on its own, is actually step planning and long-term thinking. So for Workflows that take less than 50 steps, I think it's actually pretty... it's getting there, but if you have a pretty esoteric, long workflow, it's going to be really hard. If you needed to navigate the... What would be especially hard here?
Turner Novak:
What about with the tax revenue with the corn, like doing something with that then also?
Paul Klein:
I think if you wanted it to go through your TurboTax and complete all of the tax steps for you, that's probably too hard for a web agent right now. It's just so many steps. And if you think about how an agent works, you often give it a high level goal, like do my taxes, and then what does it do next? Well then it has a planning step where it's like, okay, what do I have to do to do your taxes? Step one, go to turbotax.com, and you hand that off to your web agent, which will go to turbotax.com.
Step two, click start taxes, and you just kind of repeat this over and over and over again that the planning step becomes so long, it'll often get caught in a loop where it's like, "Oh, okay, I need to see if I'm married or not. Go check my marriage certificate." It's like at one point when you're building these agents that are long-running, because there's so many steps the the prompt becomes less effective. It's because it's too much context in the prompt for it to be really reliable.
Turner Novak:
Interesting. How do you think it's going to get better?
Paul Klein:
I mean, we've already seen how post-training is making models more powerful. We kind of saw these benchmarks, o3 from OpenAI. I think that step planning and really thinking, like being interesting thinking to large language models, is going to result in them being able to handle more and more complex tasks. You're going to be able to ask it, "Do my taxes," and it actually can think about all the steps that it has to do to really reliably do your taxes.
Now, does this happen next year or the year after? I don't know. But all I'm saying is that it seems like we've really started to master the, "Click this button on this website" workflows and we're starting to see more progress on the, "Do this complex workflow of multiple actions." It seems inevitable that soon you can give AI a high-level task and they can just iterate over and over again until it finds out what to do.
If you've tried using Devin yet, the AI programmer that you kind of use, it's not great, but you can see it doing this planning step where it plans how it's going to attack a problem, it does that, it then reflects on its plan, it does it over again. It gets stuck sometimes and it takes hours sometimes, but that agent loop is something we're going to see more and more of in the next year.
Turner Novak:
Interesting. Yeah, probably one of my, I say favorite, but it's also really frustrating when you're using ChatGPT or something, like, "Hey, did you finish this yet?" It's like, "No, I'm not. I'm still working on it." And you're like, "It's been like an hour. What are you doing?"
Paul Klein:
That's the problem, I mean imagine because models right now, they may have such a small context window that an hour's worth of reflection might not fit into the context. They have to chunk it up and that means you're losing information, larger context, more accurate models that are trained on more data will result in more powerful agents and it's just going to take some time.
Turner Novak:
So if I'm building a tech company that's in AI, it's changing all the time, is there a lot of, I don't know, tech debt kind of where you basically built something and then the models all get better and this context changes and you had all this stuff built for this world of shorter context windows. Is this something you've kind of been running into or?
Paul Klein:
I think things changed so fast, especially for our customers that it's hard to keep up sometimes. You may build a state-of-the-art AI feature and then all of a sudden three months later like, oh my god, you aren't re-ranking your chunks? What a boomer. And I think that the techniques are changing so, so much that things are constantly changing. And hey, maybe AI will help us keep up to date with the best way to use AI. What if Devin was self-improving where it's like, "Oh, I learned how to actually improve my code gen abilities." And it's self-improving. I think it'll be a while until we see that.
However, it is hard to keep up with things and that's why we've really invested in great evals for our AI features at Browserbase so we know, is something performing well. If it's performing well enough, we don't necessarily have to change it that much. That's what we're trying to cost optimize. So starting with the eval and then working backwards to the feature I think has allowed us to build AI functionality that tends to work well and last.
Turner Novak:
Another question, maybe related a little bit to what we talked about earlier, being an infrastructure company versus a product company. There's some that only do one of each, some that started, I feel like you probably started at infrastructure and then also moved to product. Is that a good idea, is an infrastructure company in terms of defensibility of business model, is it too complicated? How do you think about that?
Paul Klein:
Yeah, I had a really interesting conversation recently with one of our engineers. They were like, "What do you think of v0?" Which is Vercel's AI website generator. And I was like, "I think it's one of the smartest strategic moves that we've seen in infra yet."
Turner Novak:
Really?
Paul Klein:
Yeah, here's why. So what is Vercel? Vercel, they host your website, your Next.js application, Next.js is the framework they built. So they're the hosting platform for this framework they built. Well, what's happening, more and more code is being generated by AI. If you are the framework company, the hosting company, you want that code to be generated using your framework. So what do you do? You build the best damn website-generating website, v0. So v0 only generates Next.js code, and Next.js code runs best on Vercel. And I think that's just so smart that as an infrastructure company, you're building the best product that generates products that use your infrastructure. That's mind-blowing to me.
So I'm coming back to that and saying, how can I build the thing for Browserbase that generates the Stagehand code that runs on Browserbase? And another analogy here is what ChatGPT did for OpenAI, you know? Like, ChatGPT was the application layer and then OpenAI's APIs were the infrastructure. And I think that's just a really, really smart way to go launch something, and it really ensures that your infrastructure can be enduring. If someone builds the best company in the world on top of you and you don't capture their revenue because they just switch off of you, I think a lot of infrastructure builders are thinking about how do they have their own v0s right now? So it's super insightful to see a new strategy.
Turner Novak:
And you look at what are some of the most valuable technology companies, they're basically both, like Amazon, AWS and the website, the commerce portion. Maybe that's not quite the same, but Google owns the infrastructure and the distribution, Microsoft, the infrastructure, they also have the hardware, also all the applications and the operating system. It feels like the more fuller stack you go, it's probably like order of operations, right? Of like, what do you do at what point? But they're all trillion-dollar companies.
Paul Klein:
Yeah, I mean it's certainly order of operations. Microsoft didn't touch a lot of the other stuff until the operating system was really good. Amazon had to make e-commerce great before they even consider AWS. And there's so many e-commerce saying-
Turner Novak:
Products moving back to infrastructure.
Paul Klein:
It's decades too, right? I don't think about that too much right now because we just got to get to the point where we're Amazon, the great e-commerce website, or before we even think about that-
Turner Novak:
Best bookstore.
Paul Klein:
Yeah, exactly.
Turner Novak:
Maybe you're the best bookstore online. Maybe you're starting to be like, "All right, maybe we add CDs or something." I think one last question I want to ask you was you have this concept called the sine wave concept. What is that?
Paul Klein:
You got to stop talking to my employees. I don't know what you're doing over there. But yeah, internally I think about the effort that we put in at work as a bit of a sine wave, where we will have moments that we're peaking and we have a lot of high intensity, and after that you have to follow with some low intensity. It's the only way that I see work happening sustainably. And those high energy moments where we're working really, really hard, that's when great work happens. You do need to have those moments where we're pushing before a big launch, we're supporting a customer where we have all hands on deck and we're working extremely hard. But if you don't also have those moments of recovery where people are taking time off, people are working a little bit less hard, you can't reach the peaks that are necessary to really do hard things.
Like, if you're always at the top, you burn out. If you never push hard, you never do great things. And if you kind of sit in the middle, you're kind of in a place of just average. So I think what I've discovered is that, and it's actually crazy, if you look at our GitHub commit charts, you can literally see the sine wave happening. You can see like, "Oh my god, we really pushed hard." And now even our fundraising, we pushed really hard for our first self-service launch in June, and then we raised in September, we launched that then. So I think that's proved to work really well for us. And it's just all about being strategic about when you push and when you recover.
Turner Novak:
What's a good way to know when to push and when to recover?
Paul Klein:
Well, I think you don't really know. If you're at a company that's product market fit, you won't get to know when to push, you just have to push. The pressures will result in you being pushed. And then oftentimes you have to make sure you're being aware of how hard the team is pushing and for how long so you can come up for air. An example for us is like, hey, we have to hit a really big ambitious goal this quarter. We need to work really hard to get this feature out the door. Or we've had a customer come in and say, "I want this thing." And we as a team say, "We want to get this customer. Let's push really hard to make this feature work."
So I think those have been moments that external factors have resulted in a push. And then after that, like we just took our first team holiday where between Christmas and New Year's, everyone was off. And I stayed online, I'm on call, I was handling customer questions. I'm going to take my holiday in a few weeks, take some time for myself. But that was great because everyone just got to relax and not look at their Slack for a week. And then we all came back energized. And I think that's going to lead for us having a really solid January as we push really hard again.
Turner Novak:
So it sounds like you don't get to choose when you push, but when that happens, you then have to be like, "All right, we need to make sure we ease back a little bit and recover and take advantage of-"
Paul Klein:
Of and not even a little. It has to feel noticeably different.
Turner Novak:
Oh, really? Okay.
Paul Klein:
Yeah. I think you have to tell everyone to go home, they can go chill for an afternoon. Or like, hey, let's make sure we have a really great team dinner and spend some money on it in these recovery weeks. It's important. It's like your body, if you push at the gym every single day, you get injured. You got to have a rest day. I think the same is true for startups.
Turner Novak:
Yeah. Speaking of the gym, I did leg day two days in a row the past two days and getting up from the plane sitting here for five hours, I was like, "Oh my god, why did I do that to myself?" Top of my-
Paul Klein:
I wish I could go to the gym. Damn, must be nice, VCs.
Turner Novak:
You got to weave that in. You can't say that because there's tons of founders, like the Gundo guys are like benching in their office.
Paul Klein:
I've never seen a founder that has PMF and raises a series A with a six-pack.
Turner Novak:
Yeah, that's right. I definitely do not have a six-pack.
Paul Klein:
I've heard secondhand that a very famous VC has told his investors, "If someone raises their series A and doesn't get fat afterwards, then they're not going to make it."
Turner Novak:
What?
Paul Klein:
I'm not going to quote that, but it's something I heard because at that point, you're so stressed that you have to go all in on your company, and I've gained 20 pounds in the first year of Browserbase. I'm not proud of that. And I'm getting back after the gym actually. So it's going to happen. But I think at one point you do only have so much willpower per week, and I just wanted to maintain my fitness, but I think if I was going, spending an hour per day, five days a week in the gym, five hours doesn't seem like much but the amount of time that cuts into, it's a lot. And I think it's super, super hard. And I've done an Ironman, I've done three marathons, I did 1000 Pound Club. I'm a fit guy. And even I couldn't hang onto the gym while I was building a company that has PMF. I just think it's like... It's a hard expectation to put on yourself to do both.
Turner Novak:
Yeah, man. That's fair. Well, this is a really interesting conversation. A lot of fun.
Paul Klein:
Thanks, man.
Turner Novak:
Thanks for doing it.
Paul Klein:
Yeah, appreciate it. Thanks for having me.
Stream the full episode on Apple, Spotify, and YouTube.
Find transcripts of all other episodes here.