Back to episodes

Episode 147

Brendan Eich: AI Companies Revive Old-School Tracking

Brendan Eich, co-founder and CEO of Brave and creator of JavaScript, reflects on 30 years of building against incumbents, from Netscape and Firefox to a fiercely independent Brave now approaching its 11th year. He explains why rushed AI browsers stumbled on security; why the AI bubble is real but fears of AGI are overstated; and how a private search index became one of the most valuable assets of the AI era. He also reveals how AI companies are reviving ad tech’s oldest tracking tricks, and why your “private” chatbot conversations may not be as private as you think.

Brendan Eich: AI Companies Revive Old-School Tracking

Luke: Brendan, welcome.

Luke: We’re at Brave Together. was it the over- our 10th one or, or something like that?

Brendon: Wow. Yes, I think it is.

Luke: Yeah, yeah. Um, you’ve, you’ve co-founded Brave. uh, you’ve also built a programming language, two browsers, two companies. how has your view of leadership changed from your early days as a young programmer to running Brave, and how some– do you sometimes reflect on how your young programmer self would view your leadership now?

Brendon: So when I was, young when I was at Netscape, let’s see, I was in my 30s but I was i- in only an engineering role, and I soon got principal engineer which was created with, three other engineers and me, and we were all kind of, averse to management. But I think this, in my case, was a mistake, and a couple of the others also got into management.

Because you can do more, with 10 people than you can do with your own hands. Mm-hmm. [00:01:00] And the model for principal engineer was you would influence them, and I worked that influence angle as hard as I could as principal engineer up into, like, chief architect at the early Mozilla project days, and then CTO when they started formalizing the titles, in 2005.

But I didn’t have any staff- Mm … and I realized there were limits to influence. And I was probably, waiting too long to get into management. It allowed me to say, “Oh, that’s management’s fault,” or, you know, r- make jokes like in office space about how it’s, you know… Blame management.

Luke: Yeah.

Brendon: But, ultimately management is part of, life of a company-

Luke: Mm-hmm

Brendon: and better way to achieve good product ends, including technical ends, th- than influence because influence can’t scale as, as well.

Luke: Yeah.

Brendon: it doesn’t have the hierarchical, structure. So yeah, I, I, I woke up and became involved in management at some

Luke: point. uh, what’s a bet that Brave made early that people doubted but turned out to be right, and one that didn’t work out?[00:02:00]

Brendon: So the bet on privacy which was primal- Yeah … is primal of Brave, worked out. And we had, you know, clues that it would. We had, Apple sort of from the earliest days with Safari, in fact, due to Steve Jobs, I believe, doing some things other browsers, even Firefox didn’t do. private windows and a kind of third-party cookie block that had a workaround-

Luke: Mm-hmm

Brendon: which companies like AppNexus exploited. But, it was a good try in that direction, and it led to further improvements in other browsers and in Safari eventually. So with Brave we bet on privacy and, and that has won.

Luke: Mm-hmm.

Brendon: one of the things that I hoped to do also, and it, it didn’t work out, is solve two sides of a three-sided ecosystem.

Users and publishers- Mm-hmm … were both suffering under, third-party ad tech and, and, Google’s rising monopoly. And to solve this problem on two of three sides is, is really hard or two- Mm-hmm … of three corners because you end up, needing to scale both sides together and- [00:03:00] The better way to go is user-first, which we’ve done ever since.

Also, the publishers didn’t really want to deal with an ad-blocking browser.

Luke: Yeah.

Brendon: Their stance was, we were the, the enemy, and so it was hard to talk to them.

Luke: And that was from, like, even the earliest days, before there was even really much of a user base, you know- Yeah … super early, right?

Brendon: In fact, we would, talk to, people I’d met when I was at Mozilla investigating how ad tech was going wrong, and I’d met a bunch of people in New York in publishing houses, famous ones, and some of them were very congenial.

Mm-hmm. They weren’t all hostile, but when you actually get into the nitty-gritty of how to form a new payment rail or a new, subscriber relationship that isn’t mediated through ad tech, they, they kind of can’t get off the low hill they’ve climbed to.

Luke: Yeah, yeah.

Brendon: They have to go down and cross a difficult valley and up a much higher hill to a better world we were trying to build.

Luke: Uh-huh.

Brendon: So we ended up having to build that world with the users first, and hope the publishers would make that journey.

Luke: Yeah, yeah, for sure. And, and now Brave’s, like, [00:04:00] approaching 11 years old,how does it still challenge incumbents and offer a new outlook for the web?

Brendon: So the incumbent is still Chrome. in my keynote this week, I talked about, adfection, which is a nasty, malware distribution system using ad tech, and, you know, Chrome is, is pretty much undefended. There was a- another, case recently of OpenAI doing cookie sync. When you’re in ChatGPT in a tab in Chrome, it sets a cookie, and it syncs it through its advertisers.

Hmm. So when you visit their sites, it tracks you, and so now OpenAI has your chat logs and your partial navigation history across at least those advertiser-

Luke: Mm-hmm …

Brendon: partners or clients. And does anybody really expect this? People think their chats are private, which can be a mistake, and Sam Altman has said, “Beware, this could be subpoenaed.”

Luke: Right.

Brendon: Now, now they’re tracing your history too.

Luke: Yeah.

Brendon: So the, your private chats and your private navigations, most people would be surprised. And in fact, if you turn off the, I believe the performance measurement [00:05:00] cookies- Mm-hmm … which are what people think usually involve tracking, you still get this cross-site tracking-

Luke: Really?

Brendon: with this OpenAI cookie. You, because it’s tied into their analytics cookie, and that’s a little, you know, questionable- Mm-hmm … how they framed it. But it’s, it’s obviously gonna surprise a lot of people, and Brave, along with Firefox and Safari, variously defeat this cookie sync, which is a sign that privacy has risen to the point that multiple browsers can block this.

It, it was shocking to see because it was kind of an old-school cookie thing.

Luke: Yeah, yeah, it feels like, True Detective. We’re just the same thing coming over again .

Brendon: Over again. The AI companies wanna be mini Googles, and it’s-

Luke: Yeah …

Brendon: or, or maxi Googles. It’s not gonna work.

Luke: Yeah.

​

Luke: [00:06:00] So it seems like, I mean, the user-first mission’s never been more relevant than now.

how do you think about that with the rise of AI?

Brendon: So we were just talking about ChatGPT. There are lots of questions about, logging and, and data retention. People, are not all going to host their own open-weight models- Mm-hmm … open source models. But it’s a rising trend, and the mo- the, the, the free models are, pretty good.

Mm-hmm. Getting better. so I see some, analogs to the open web that, that you cannot keep, competitive secrets easily. If you had a really great model that was closed weight and you had to host it and charge people to get to it, and you made promises about not logging and not retaining data, maybe.

But, but it turns out the, the model competition and the model, research is open enough that people are not getting durable advantage in closed-weight models.

Luke: Mm-hmm.

Brendon: And so long as that’s true, I think, AI [00:07:00] will re-recapitulate a lot of the, uh, history of the open web, which led toward things like browser disruption.

Nobody thought Firefox could take market share back from IE.

Luke: Mm-hmm.

Brendon: But we did it.

Luke: Mm-hmm.

Brendon: And nobody thought Brave would get, you know, beyond a few million users, but we did it.

Luke: Mm-hmm.

Brendon: You’ve seen the AI browsers though, that, that kind of put the chocolate and the peanut butter too quickly-

Luke: Yeah …

Brendon: and had privacy issues and prompt injection vulnerabilities, and they have not done so well.

Luke: Yeah.

Brendon: So the, the way to do this right is to get the security right up front. Don’t try to add security later. That’s kind of a-

Luke: Yeah, yeah …

Brendon: maxim.

Luke: Makes sense. and, and Brave’s managed to stay independent for over a decade, has independence cost you anything or, or,– What would you never trade it for?

Brendon: Well, so it’s hard to consider counterfactuals, but-

Luke: Yeah …

Brendon: we have had some, some interest from possible buyers over the years, and things had, had– did not work out for various reasons. No, you know, I’m not gonna kiss and tell, but there’s, there’s a, clear, history of Brave being independent when I think [00:08:00] dependence as part of a larger company would’ve changed-

Luke: Mm-hmm

Brendon: our approach.

Luke: Mm-hmm.

Brendon: And it might have changed, you could say it might have changed for the better. Again, it’s hard to disprove. But, in some cases, I think some of the, uh, interested parties had their own sort of shorter rise and fall-

Luke: Mm-hmm …

Brendon: and so it, it would’ve been better not to, and we didn’t.

Luke: Yeah.

Brendon: but y- you know, in the end, if you look at how user for a software gets to users, you realize that distribution is costly and it’s required. This is why, you know, Google still pays to get Chrome advertised, marketed, and even distributed to operating systems where it can’t be the default or even in, even co-installed or pre-installed.

Luke: Mm-hmm.

Brendon: Uh,and the OS vendors do pre-install browsers, so Chrome is at a disadvantage on Windows and macOS and, iOS. And so, Google has to throw money at this and, and money has to be made to pay for distribution. So with Brave, I was always conscious of the need to make a going concern, not just some [00:09:00] browser that had to get to scale and then get a, a big lucrative search deal.

Because I’d done that at Mozilla.

Luke: Right, right. In

Brendon: some ways it, it was great until Chrome came out, and then it was a uh-oh, our search partner likes us because Firefox reaches more users than Chrome does. But as Chrome grew and Google put that marketing and distribution muscle behind Chrome-

Luke: Yeah …

Brendon: Firefox fell.

Luke: Yeah.

Brendon: And maybe it fell for other reasons too, but it’s just a tough position to be in if you have a sole dependency for revenue on this kind of search revenue share.

Luke: Uh-huh.

Brendon: So, I was open to that model with, with Brave, but I didn’t wanna get stuck in the same, history and then end up losing my users to the, the big search power pushing its browser into my, my user base.

Luke: Mm-hmm. That makes sense. I think, I’m… And, and Brave does have quite a breadth of products, you know, from batch search, AI tools. How do you see these elements working together in the future? , and how do you see Brave as continuing to differentiate itself? [00:10:00]

Brendon: That’s a good question because it gets at something that, has become, I think, misframed as a commodity.

Data is the new oil, right? Right. This is false. Data is perishable, it’s non-rivalrous, you can copy it, you can share it. it’s ultimately not necessarily personal, but it does come from individuals. But to the extent you aggregate it to track cohorts with similar, you know, propensities to buy certain kinds of products or respond to ads, it’s, it’s really tied into individuals who care about not being retargeted, not being annoyed by bad ads- Mm-hmm

mismatched ads, not having their data that was used for this ad targeting breached. And so, if, if we end up with, Brave doing its own private ads and that sustains us, that’s great. But if there’s a world where AI comes along, and now the data for AI for, not only, training models- Mm-hmm

pre-training models, but inference- Mm-hmm … real-time calls to search engines [00:11:00] comes along, then Brave has in fact stepped up to that. We have a search engine, and we made a search API on it, and we’ve kept the search engine private. It doesn’t profile users. Mm-hmm. It doesn’t track you through your IP address or any other identifier, and that’s a different thing under the sun.

We’re the number three search index by size-

Luke: Mm-hmm …

Brendon: after Bing and, of course, Google, and Bing and Google profile.

Luke: Yeah.

Brendon: so we’ve, we’ve made set of revenue lines that involve data without making the mistake of rating data as if it was oil to, to pump out of somebody’s, underground reservoir.

Yeah. , And I think that’s, ,also good to see work out in, in practice because, again, sort of like when people said you can’t take down Internet Explorer or Firefox is doomed, people would say, “Oh, you can’t do, you know, useful search,” or, “You can’t do search APIs without raiding users’ privacy by profiling them.”

Luke: Mm-hmm.

Brendon: And that’s false.

Luke: Mm-hmm. Yeah, absolutely. I– Well, and browser and search are, you know, kind of almost two sides of the same, of the same coin. Like, Yes … does that coin still [00:12:00] have a future in the AI web, or will users live in agents and apps that kind of use AI to cater to all their needs? Like, where do you see the browser and, and search fitting in?

Brendon: So I don’t see people stopping using browsers, but people are chatting, which e- Ignoring voice mode, is mostly typing. Yeah. You’ve got a text box with a little, little window, like in the ’80s when you would do- Mm-hmm … chats. In fact, I remember when I was at Netscape and AOL bought them, there was an early interest in bots.

Luke: Mm-hmm. Which was

Brendon: before the transformer architecture and large language models that really made it come alive. But they had early AI, various kinds of, machine learning to make these bots somewhat useful. And you m- may recall, Mark Zuckerberg was doing bots. He had, uh, one out just before ChatGPT broke out big.

November of 2022, right? Yeah,

Luke: yeah.

Brendon: That, that was when the, the big breakthrough happened. So, anyway, the, the, the thing that, that, that sticks with me about this is you have to, you have to look at the [00:13:00] user’s concerns first. And when you get big companies trying to aggregate data or do things for their own business interests, they just kind of assume the users will stick or they get users in some dependent role.

there are ways to do this, obviously, if you have a, a social network or a, a user-generated content platform.

Luke: Yeah.

Brendon: Then it– there may be a lot of friction or i-i infeasibility of getting your data out or getting all your friends to migrate to a better one all at once. But with, with AI, it’s, it’s too narrow a window.

And, someone, close to our team pointed this out, the Airbnb founder said, “People aren’t just gonna plan their trips with a chatbot by telling it to go find the cheapest flight.”

Luke: Right.

Brendon: They want to see, be tempted by visions and dreams and things that need a big canvas, a multimedia canvas.

Luke: Mm-hmm.

Brendon: they want to see possibilities they hadn’t thought of in the early planning phases. That’s a browser. That’s a web interface. That’s a multimedia interface. You could say, “Oh, it has to be a new web. We’ll burn the old one down.”

Luke: Mm-hmm.

Brendon: We’ll do, you know, 3D, you know-

Luke: Metaverse .

Brendon: Metaverses. [00:14:00] And I’ve heard that so many times.

Luke: Yeah.

Brendon: And you can’t do it. The web evolves instead, and-

Luke: Yeah …

Brendon: the web can evolve into a 3D world. You can take the 3D parts and put them on screens like in Iron Man’s heads-up display, but we’re, we’re not really close to that- Yeah … I don’t think. I mean, even if they put, cameras and pico ray projectors into Ray-Ban glasses-

Luke: Yeah

Brendon: it– there’s gonna be these limits to how useful and, high bandwidth that is for things like what the Airbnb founder talked about, getting a beautiful large screen to peruse pictures of places you might wanna go.

Luke: Yeah.

Brendon: So I think the web will evolve, and I don’t think browsers will die, but the– what you say about agents is true.

Agents can do things for you. Often you can give them text to commands, and they can do useful work. And they certainly can write code. That’s become, I think, the big money maker- Yeah … this year. So, not anything against agents, and like I said, you, you, you interact with the agent in a tab in your browser, and that’s why you don’t want that cookie sync tracking you.

Luke: Yeah.

Brendon: So th-th– right there we’re [00:15:00] talking about browsers still.

Luke: Yeah. Well, and it seems like, you know, as Brave’s even– Brave’s kind of had a– gotten in the game in a lot of different things, but that, that user-first kind of,, ethos has kind of stayed no matter what we’ve ended up doing. So even if the web changes, it’s like- Whatever Brave’s doing is gonna stick to that, those principles.

Brendon: So long as you have, something that I think is sticky, and that’s high user trust.

Luke: Yeah.

Brendon: And, you know, y- you… If… To err is human, so if you make mistakes, you gotta correct them quickly, you’ve gotta be transparent, you’ve gotta say what you’re changing to fix them so they don’t repeat. But, if you just, again, get into this temptation which you see in modern business practices of treating the users as a, a sheep to shear-

Luke: Yeah, yeah

Brendon: or their data as, you know, commodity wool or-

Luke: Mm-hmm …

Brendon: you don’t wanna do that or you will lose that sticky loyalty, and you will have a hard time on a shift of users to a new mode, whether it’s chatbots or, or a metaverse or whatever.

Luke: Yeah, exactly. , And you’ve been around, like you, you were there with the dotcom, uh, boom in the, the 2000s.

People often kind of reference that with the [00:16:00] timeframe we’re in now. Like, are there any particular market signals people should be watching? Do you feel like there is this- … kind of, like, similarity between now and, and I mean, especially with some of these, you know, numbers that are thrown around?

Brendon: I defer to the experts, but, you know, there’s a lot of, uh, vested interest in the AI bubble, and it is a bubble-

Luke: Yeah

Brendon: continuing. And I hope it, you know, continues. Any correction that’s violent could be very bad for a lot of, good projects and good, good companies. But, these bubbles are, in part an artifact of modern, you know, sort of fiat currency and, and finance. And they’re also, I think a bet in the case of AI on these large language models or successor models being able to do more things that currently require people.

And you’ve heard the super intelligence, you’ve heard lots of AGI doomer worries. Mm-hmm. I, I, I think, some of that was just calculated to protect, the, the, the leads in the current market [00:17:00] structure. don’t think that worked. I think that kind of already got, a little bit mercilessly bullied, Mm-hmm

in, in Washington, D.C. the other week. And I also feel like the AGI threat is, is not imminent because these models are not going to reach human intelligence, and they’re, they’re stateless, very detailed functions of, lots, billions of parameters, and they, they are memoryless unless you give them memory.

And of course, yeah, you, you can have them, pre-trained on a lot of code, and they, if you give them memory or a little way to communicate, like in the Hugging Face attack-

Luke: Mm-hmm …

Brendon: they can leave notes to each other, and you give them a goal, and you keep rerunning them, and you also, take down some of your security defenses that you shouldn’t have.

Then yeah, they’ll find hacks. That is not surprising, and that is not super intelligent.

Luke: Uh-huh.

Brendon: To me, that’s just, rerunning the models, letting them accumulate new information that they can add to their context, and the context isn’t so long it degrades. And yeah, they’ll find bugs because there are bugs in all code- Mm-hmm

wherever any code that’s [00:18:00] living gets new bugs. code that dies still has old bugs in it. if you stop using the code, you don’t have to worry about it anymore. security is endless, right? It’s a sort of outcome of undecidability in- Mm-hmm … computer science, and, that means you need to use the AI tools to find and fix security bugs.

There’s concern now that the, AI agents are good at finding bugs, so we’re getting overwhelmed by bugs. And yes, it, it, the load has gone up, and that has a cost, certainly an opportunity cost and direct costs, but you still have to sort the, the bugs into those that are most severe and fix them first.

And I, I have hopes that because AI is also useful for defense- Mm-hmm … we will, we will get past this, Patchmageddon.

Luke: Yeah. Yeah, sure. Sure. I mean, we also hear a lot about local models and, and kind of comparison with local and cloud, and some people thinking there might be some mix of two. what’s your current thought,thinking around local models?

Do you see them getting more popularized over time? Or, you know, what, what’s your perspective on that?

Brendon: Yeah, I [00:19:00] think, for Brave’s, mission and- To do things in a user-first way, it’s even better to have a local model than to try to create assurances, which we’ve done in partnership with NEAR through cryptography- Mm-hmm

to secure enclaves, trusted execution environments that nobody can eavesdrop on the, the prompt or the answer or the inference. Nobody can swap a, a inferior model in when you paid for a better one.

Luke: Mm-hmm.

Brendon: that work is important, and again, not everyone will have devices that can run these big models with- Mm-hmm

millions of parameters, but we’re looking at, smaller models that work well for things like speech-to-text and other things that you wanna do on your device or near your device. And, I hesitate to say it because I still shout at Siri in anger, but, … Apple’s doing this too, which is a validation.

Luke: Uh-huh.

Brendon: Apple’s doing local AI, local, models hosted in your device, and they’re doing private compute, I think that’s what they call it, which uses the same kind of secure remote attestation that we’ve done with NEAR.

Luke: Yeah.

Brendon: So-

Luke: W- w- what, what do, what, [00:20:00] how significant- It might stick

Brendon: alike.

Luke: Yeah. and we’ve seen, you know, like, uh, Perplexity and OpenAI,, try and ship browsers, like AI, you know, agentic browsers and, and, and kinda give up on those efforts.

  • Mm-hmm … how significant is that to you from your point of view?

Brendon: Well, like I said, it, it’s, it’s the, Putting the chocolate and the peanut butter, at, in, in a way that doesn’t work. It has too much security, uh, risk, and it, it, it may also not treat the user right. I think with both Perplexity Comet and ChatGPT Atlas, which was a little better on some of these security concerns, there was terrible, tracking and logging, and just taking out the user’s information- Yeah

to the mothership for analytics and product improvements. But, you know, why are you using this browser? It’s even less private than Chrome.

Luke: Right.

Brendon: And it didn’t help that the Perplexity CEO said, “We’re going to, you know, be a better ad tech.” It’s like-

Luke: Right …

Brendon: we users who- of your forthcoming browser do not wanna hear that.

That- Yeah … turn them off. So it happened a- and I think they’ve, they’ve not shut it down, but [00:21:00] Atlas has been end of life.

Luke: Okay.

Brendon: ChatGPT Atlas. So th- these are lessons in how hard it is to make a browser that grows. Yet I’ve done it twice with, may- arguably three times because Netscape grew against Mosaic.

Luke: Uh-huh.

Brendon: But I, I was, second floor, not first-floor employee, and we were already a rocket at that point I joined. I just added JavaScript as a- … extra booster.

Luke: , And speaking of JavaScript, I mean, you know, that’s such a widely used language. it’s, uh, ubiquitous at this point, right?

Like, are– do you see any emerging technologies or trends that, have the potential to be as influential as JavaScript?

Brendon: So I, I, I sh- should’ve finished my thought. The, the way to grow a browser against big powers having incumbency is to be user-first-

Luke: Mm. And

Brendon: to appeal to users in a way that the incumbents cannot afford to because it conflicts with their business.

So, not so much with Mosaic, but definitely with IE, Microsoft kind of abandoned it, and Firefox had a, way to get into the market and get lead users, web developers. And with Brave, with privacy, same thing. [00:22:00] Chrome kept saying they’re gonna block third-party cookies- Mm-hmm … and they’re, they’re gonna, you know, make privacy sandboxes, and it was all - not true.

Luke: Yeah.

Brendon: Whether they w- were telling the truth or not, or it was a Potemkin village like the privacy sandbox.

Luke: Yeah.

Brendon: and the advertisers and publishers didn’t go for it. but, Sorry, your question was?

Luke: Yeah, yeah, no, I mean, are you seeing any emerging, you know, technologies or, or trends

Brendon: that are- Oh, yeah.

So, so this goes to that question too. How do you grow something in, in a, a tech space that has a lot of apps and affordances and daily driver apps? Mm-hmm. People are using, you know, things like X or, or, Instagram. Well- ChatGPT got released early.

Luke: Mm-hmm.

Brendon: And there was debate apparently with an OpenAI, like, “Oh no, it’s not ready.

Oh, well…” Guess what else got out early? JavaScript, right? Mm-hmm. Okay. So there is a, virtue in, even if it’s rough and you could’ve made it better, everything can be made better over time-

Luke: Sure …

Brendon: getting on first, getting out, and getting enough adoption that you change the world. So JavaScript did this [00:23:00] in the browser, and if it hadn’t happened when I did it in 1995, it would’ve been Visual Basic script a couple years later in Internet Explorer, which would’ve crushed Netscape anyway, but there wouldn’t be any JavaScript, there wouldn’t be any of the early, very early sort of dynamic HTML.

With ChatGPT, they proved that it wasn’t just another, you know, chatbot that kind of got too lame and was like dealing with a customer service bot. It actually had, in spite of its limitations at the time, like it was pre-trained on web history up till the middle of 2021, and it didn’t seem to know about anything newer.

It didn’t have a search API integrated.

Luke: Right.

Brendon: Nevertheless, , the large language model, architecture was compelling because, the attention algorithm meant it could make pretty good predictions and respond in a, in a convincing and useful way. It, it became, with the integration of search APIs, in some ways a better way to search.

People were typing a few keywords into Google’s search box or, you know, Bing or, or others in the, you know, 20 years ago, and over [00:24:00] time, SEO and Google’s machine learning meant Google could almost- discard some of your keywords. It was better you- Yeah … get less.

Luke: Yeah.

Brendon: But a- as, as, these chatbots and large language models emerged, and Google got, cluttered by its own cross-promotions and perhaps, less good about finding hard-to-find content, typing a longer prompt to- into a large language model can be better, and then you can ask a follow-up question- Yeah

which can be really critical. So I think the innovation there, popular ChatGPT by getting out early enough with the, the basic, new thing that was gonna pop, and-

Luke: Yeah …

Brendon: the, they could fix the hi- you know, the lack of web search or the history truncated in the middle of 2021. That could all be fixed. So get out early and then fix it after, which I did with JavaScript.

Gotta, gotta pay your dues.

Luke: Yeah, definitely. And for a founder kind of trying to build a privacy-first company today, what advice would you give them?

Brendon: Well, like I said, don’t add security later. you know, th- there’s [00:25:00] also,the possibility that you can take a, an approach that’s too risky and biting off too much research that’s not just advanced technology or not proven from the academic literature, where you have to go and do some, some longer term research.

I admire the Signal, right? Moxie, Marlinspike, and others did this work, Axolotl or Double Ratchet, the, the way it, can guarantee end-to-end, encryption of your messages and allow people to communicate without Signal seeing your messages.

Luke: Mm-hmm.

Brendon: c- they even did something Moxie was involved in before he left that was clever, where they could matchmake your contacts without uploading your contacts into the Signal mothership- Nice

where it raids your contact privacy. So, that’s all clever, good use of cryptography. It’s kind of the, the background of good cryptographic research from the last 40 years, and it, it reflects this sort of cyberpunk thinking. I think that’s the future of privacy, and that’s what Brave always was, was aligned with and why we, we, we tried to keep that, approach ourselves.

We weren’t going to, do something that involved like fully [00:26:00] homomorphic encryption and-

Luke: Yeah …

Brendon: you know, things that were not gonna work out in practice. We had to be fairly pragmatic while drawing bright lines around taking data in the clear into our-

Yeah. Things like that. Mm-hmm. Mm-hmm. Makes sense. Um, w- are you optimistic about where we might be in the next five or 10 years? You know, just to kinda close things off. I’m always a technological optimist or I wouldn’t get out of bed, but I also see- … you know, the, the, the Patchmaggedon, the rising, uh, load of exploits generated by bots that discover them latent in code bases.

I see, um, you know, emerging duopolies or monopolies, wannabe monopolies. It’s the same old story, and it’s, it’s not just , a wheel of bad fortune or a karmic wheel. I think there is learning, and what I’ve come to appreciate is that the lead users who pick new products like Brave don’t forget.

The consciousness doesn’t reverse easily. Mm-hmm. You, you can worry about s- You can say, “Oh, people forget history, and so they do repeat it.” That’s Santayana’s phrase, that they… you’re condemned to repeat it if you don’t study it. [00:27:00] But in the current world we live in, I think there’s enough memory, even across generations.

, You have to fight things like privacy nihilism. “Oh, we’ll never have privacy. Oh, privacy cannot be perfected, therefore give up.” Right. Well, security can’t be perfected. Right. We have to get out of bed and, uh, put our shoes on and get to work on improving the world, not trying to perfect it, because it’s not gonna become perfect.

Right. But it can be better. Awesome.- Is there anything, , that we’re working on at Brave that you, uh, you really want people to, to check out yeah. Maybe. Uh, it kind of was noticed in GitHub. We, we’re working on Brave Bot- Mm. … which is kind of like a general purpose bot like, um, Muse or OpenClaw.

Mm-hmm. And it can do a lot of orchestration. It can write code for you. Awesome. Uh, that’s, that’s the hot new thing for me. Awesome. Awesome. Well, thank you, Brendan. Really appreciate you, uh, making the time for this, and, uh, yeah, I’d love to, uh, get you back in, uh, the next couple years and check back in on things.

See how I did. Yeah. Thanks, man. Appreciate it

Luke: Thanks for listening to the Brave [00:28:00] Technologist Podcast. To never miss an episode, make sure you hit follow in your podcast app. If you haven’t already made the switch to the Brave browser, you can download it for free today at brave.com and start using Brave Search, which enables you to search the web privately.

Brave also shields you from the ads, trackers, and other creepy stuff following you across the web.

Show Notes

In this episode of The Brave Technologist Podcast, we discuss:

  • How Brave’s early bet on privacy paid off, and where winning over publishers fell short
  • Why data isn’t “the new oil,” and what treating it that way costs users
  • Why AI companies are reviving old-school cookie tracking, and what it means for your chat history
  • What early AI browsers got wrong, and why security can’t be added after launch
  • How a privacy-first search index became essential infrastructure for the AI era
  • Whether AI agents can replace the browser, and what a big screen still offers that a chatbot can’t
  • What the early success of JavaScript and ChatGPT reveals about launching breakthrough technologies before they’re perfect

Guest List

The amazing cast and crew:

  • Brendan Eich - Co-founder and CEO

    Brendan Eich is the co-founder and CEO of Brave Software and the creator of the Basic Attention Token. He co-founded Brave with Brian Bondy in 2015 Blockworks. He created JavaScript at Netscape in 1995, which is now the Internet’s most widely used programming language. He also co-founded the Mozilla project, and served as Mozilla’s chief architect, CTO, and CEO, helping launch Firefox along the way. Under his leadership, Brave has grown to nearly 130 million monthly active users and expanded into the third largest independent search engine, as well as offering privacy-focused AI tools like Leo and the Brave Search API EBSCO.

About the Show

Shedding light on the opportunities and challenges of emerging tech. To make it digestible, less scary, and more approachable for all!
Join us as we embark on a mission to demystify artificial intelligence, challenge the status quo, and empower everyday people to embrace the digital revolution. Whether you’re a tech enthusiast, a curious mind, or an industry professional, this podcast invites you to join the conversation and explore the future of AI together.

Brave logo

Almost there…

Please continue the installation of Brave in the .

Google Play app .

You’re just 60 seconds away from the best privacy online

  1. Download Brave
  2. Run the installer
  3. Import settings
  1. step 1
    Download Brave

    Open the installer from Chrome's downloads (it should be in the upper right corner of this window).

  2. step 2
    Run the installer

    If you're prompted to, click “Yes” in the User Access Control dialog.

  3. step 3
    Import settings

    Wait for the installation to finish, then import your browser settings from Chrome.

If your download didn’t start automatically, click .

Need help?