[cold open] Hank: This is a David Goliath moment for Browserbase where they've got a sling, they've got to shoot their shot, and they've got to take a stand against Goliath because if there's a Goliath in AI, it's OpenAI. 

[intro music]

Gonto: Hi everybody, we have a lot of topics for today. 

Hank: When OpenAI launched, oh shoot, I forgot what OpenAI's thing is called.

Gonto: Operator. 

Hank: That's Operator. Yeah. And then what is Paul's thing called? 

Gonto: Open Operator. 

Hank: Open operator. Okay. That's why I was confused. So he did a fast followup and he had this great copy. I'm gonna actually read it so I don't mess it up. The copy was “you don't need to pay $200 for AI” and then he did the launch.

It's just so beautiful. It says… it's basically like hey, here's the thing that they're offering for free and then it becomes about how did we do this? You know, blah, blah, blah. It's based on Browserbase and it was awesome. I loved it. 

Gonto: I loved it too, but not because of that. I loved it because of how fast it was. Like it was 24 hours after OpenAI shipped Operator, which is actually like a competitor for Browserbase eventually. They launched these to show like you can do it for cheaper and it just works. I tried it and it's good. 

To me the main questions about Browserbase that they didn't do it in this launch, but they need to do it soon is, Operator will launch an API. Once Operator has an API, it's basically the same as BrowserBase. So the question that I would have as a potential customer for either OpenAI or BrowserBase is, why is BrowserBase better? And my gut feeling, without knowing shit, Is that OpenAI is better just because OpenAI said that they did a custom model for Operator, so they retrained with rl their model to actually become better at using websites and using the browser.

I don't know if that's exactly what Browserbase is doing, or maybe they're just doing like the basic model and using it differently, but diving deeper into why Browserbase is different or better or differentiated. I think it's something they need to do probably in the next two weeks, I would argue, before people go to Operator.

Hank: Yeah, you're saying they did this good launch. They did the fast follow. They rode the wave. I loved the copy and the tweet. RavageG tacked on there too, because it was built with, you know, Vercel and the AI SDK. But then they're not tying it closely enough to, hey, what is Browserbase? How can we get you hooked on that before OpenAI creates their API and creates a true competitive product?

Gonto: Exactly. And they said they will ship it. So they will ship it. And before that they need to connect why Browserbase is better, how is differentiated and talk about that, which they did not do. Like if I try now Open Operator, which I have an Operator, they look almost the same. So as a user of those, I don't know why you want it better. They need to tell me. 

Hank: Yeah. And it has to be more, I, I think the copy is great for getting the initial set of attention. The, “you don't need to spend $200,” but that's not sustainable. 

Gonto: Exactly.

Hank: Because ultimately if you're building with it, you'll pay your VCs money to get the better thing. And it's actually….

Gonto: Pricing should never be the mode.

Hank: Pricing is not the mode. Yeah, nor is openness or any of that junk. 

Gonto: Exactly.

Hank: And this actually goes back to our David and Goliath discussion from a couple episodes ago. This is a David Goliath moment for Browserbase where they've got a sling, they've got to shoot their shot, and they've got to take a stand against Goliath because if there's a Goliath in AI, It's OpenAI, despite the Browserbases and the DeepSeeks, which we're not really talking about today, et cetera, et cetera.

Gonto: I agree and I think this links… I'm talking about how you should talk about differentiation and why it's better. However, you found the paper that I really liked that talks about Hhw you should understand it less, like, uh, tell me more about that. 

Hank: Okay, so this is a study on AI adoption. And what they found was the more you understand AI and how it works, the less likely you are to adopt it.

The less you understand AI and how it works, the more likely you are to think, wow, this is magic and start using it widely. And that is such an interesting insight and I wonder how much it translates to technical products aside, you know AI or not, should you be explaining to people how does your CDN work?

How does your logging like actually work on the fundamental layers? Like are your engineering blogs going deep in a way that suddenly your customers who understand, or your would be customers, the looky loos, suddenly understand, oh, maybe I could just build this? And suddenly they have these other thoughts, and you've created a new competition for yourself, versus just getting people to that first moment of, wow, it works, and then letting them adopt the technology more directly.

Gonto: Yeah, I think about it in two different ways. On one side, Like in growth, there's this term of the aha moment or the wow moment or that moment where the person is like, holy shit, this is really what I needed. And for that, I think if you don't understand it, it's better because why? Because if you don't understand it, it seems like magic.

I really like the phrase I don’t remember, like that, that it says about how really advanced technology looks like magic. And I think that's great for an aha moment or something like that. But then I think it is a bit different for engineers because engineers really like to understand how things work inside.

Because if you're gonna use something, you need to understand how it works. As an example, I don't know, React. React uses a virtual DOM. If you're a coder for React, and you don't know that it uses a virtual DOM, maybe you're fucking up when you're coding it. Because you don't know exactly what's going on, in what order, or what happens.

So as an engineer, I think you have to deeply understand the technology that you use, so you don't fuck it up. However, I think you can combine both. Like, to me, the best of this is, you first experience the wow and aha moment. Without actually understanding how it works and when you're like wow then I think you should explain to them how it works either with a blog post or something so they get it. And then after that I don't think it will as much influence how it is.

So on this I think it's a bit different for engineers. I think for the masses if they think AI is magic, that's fine. I'm not sure it applies for engineers. 

Hank: I don't know. Yeah, I'm curious. It's a, it's a really, if, if I had the capability right now, I would run some sort of marketing experiments on this front, you know, maybe others have, but it's interesting.

And I mean, I see right now, it's popular for DevTool influencers to speak against cursor and other AI stuff. And that's really interesting to me. Because it's still useful and you know, they have all these thoughts and it feels very similar to like, I don't know, “everything new will kill you” is what people always say.

And it's usually the people preaching against adoption of a new technology that actually understand it the most. And so I see a lot of that here and I see a lot of that in DevTools with engineers too. Some engineer who's smart is like, no, no, no, you can't move to the cloud, like. You need to control your servers, you gotta have it all this way, blah, blah, blah.

And now people are saying, no, no, no, you can't move to, like, managed infrastructure, you gotta control your cloud, blah, blah, blah, blah, blah. And, I don't know, it's a pattern that repeats itself, and so I found this study really interesting. I guess I don't have too much more to say about it. 

Gonto: Yeah, I agree. I don't know. I'd love to hear other people's thoughts. So if you have thoughts and you want to tweet at us, that would be great. 

One last topic for today. I saw a blog post from Matt, the founder of Netlify, that I really liked. And he talks about agent experience. Like, we've heard about user experience. We also heard about developer experience.

And what he talks about is that what's coming is agent experience. And he talks about in the future, everything will be an agent, like agent will work at companies. They will make potentially buying decisions for you. They will use services. So starting to think about how agents interact with your web, with your product, with everything is a must for anything that you're building or shipping now in the future.

What's your take on this? 

Hank: Currently, we're seeing AI browsers and you know, they're effectively super advanced scrapers. They're becoming more and more important. And you need to consider them in how you're building your site out because that's a new form of SEO. And as your traffic from organic decreases, I've, I had multiple people ask me in the last month, Hey, is traffic just decreasing for everybody else? Or is it just me? And I'm like, I don't know. I think for everybody, maybe? I don't know. I'd have to ask, you know, Marcel or somebody. 

But the other thing to think about here is, as you said, agents are going to become employees. Effectively, we're going to have AI coworkers. Whether you can like see that or not, I believe it wholeheartedly.

I don't know if it's a year from now when things like Devin or Lindy or whatever are going to be good enough to be considered coworkers, but definitely f5 or 10 years from now. You'll have people in Slack mingling with bots in Slack and boy, those bots are going to be your favorite coworkers. They're going to be faster, better, and everything. 

And these AI coworkers are going to gravitate toward the sites built for them and the tools built for them. So it's a whole interesting world of tooling and thinking and, you know, optimization. You know, I didn't want to call it SEO because it's not search engine, but it’s, it's something, right?

I don't know. 

Gonto: It's different. I've seen, for example, some products now have an LLM.txt which they use for the docs. So they build docs for LMs where it's just markdown and they actually maybe get most of the stuff from docs and put it in one file that it's a markdown where they can learn how to do it.

And I think the same will happen for other things. Matt, in his blog post, talks mostly about products, but I think it's the same for websites. And as I think about it, there's three things that I think are interesting. Number one, if agents are going to start using products, APIs start becoming a lot more important than your dashboard and the UX and how it works. Because if you have a good API that is well explained and you use it, It makes a huge difference. 

My advice to companies that are starting to think about agent experiences, use your own API in your dashboard. So you have your dashboard, instead of just building things for the dashboard, just build an API and use an API from your dashboard.

Because if you can do things from the dashboard for the user using that API, that means that an agent will be able to use that API. It's an easy way to test it out. 

Similarly, I think websites will eventually become an API as well. I mean, humans will continue to watch and see websites. But I think agents will need to be able to ask websites like what's the home, who are the customers that you have, what are the main use cases that you're selling, what features do you have?

And all of that, I think in pure text is better right now, at least for agents. So I think finding a way to have an API for your website or starting to have some TXTs for that, I think will make a difference. 

Last thing to mention is AIs read very fast, a lot faster than humans. If you look at GPT, for example, it reads between 12 and 15 articles for every search. Perplexity between 15 and 20 approximately. So that means that before what matter was for every keyword, I need to be first position because humans click the first or maybe the second. AI's do not work like that. So for AI's, it's a lot better to actually be on the top 15 results of multiple keywords rather than be first results in just a few. Just because if you're in the top 15, they're gonna find you. So don't care about being first. Rather care about being in the top 15, 20, which are the ones that AI will read. 

Hank: Yeah. Very interesting. And, you know, there's, there's so many other things about what kind of company AIs will talk to each other and sell each other stuff, you know, and buy from each other.

I don't know. It's a wild new world. 

Gonto: Thank you for listening to us today. Um, and yeah, we'll, we'll see you soon.  

Hank: We'll see you soon. Bye.