Peter Rojas - SVP @ Mozilla

Welcome to another Founder Wisdom podcast. We have Peter Rojas for round 3, uh, with us today. Uh, Peter is a founder at Gizmodo and Engadget. He was previously working at Meta and recently started a new role at Mozilla as the SVP of new products. I'm eager to discuss the new products that Mozilla is launching. Peter, welcome to the pod again. Tell us a little more about yourself and what you've been up to.

Yeah, thanks for having me back. It's always nice to chat, and I appreciate you having me on 3 times. I don't know if that's a record for you yet in terms of guests, but, uh, yes, probably. Nice to, nice to, uh, nice to be welcome back. Um, yeah, uh, so, you know, I, uh, I, I'm— I've been at Mozilla for, uh, since February of last year, uh, and, um, I've really been focusing on, um, you know, ways that we can, um, kind of extend the history that Mozilla has of really centering users and building products that have a lot of trust, transparency, very private, and have focused on really thinking a lot about data sovereignty and the needs of especially enterprise users, whether relating to AI or not, obviously very hot topic. I have a number of products in the portfolio that we are either building or have shipped, We have a product, I have a team in Europe building an enterprise, secure enterprise version of Firefox, we call it Firefox Enterprise. And so this is something where, you know, because of how much work now happens in the browser, this is a trend that's been going on for a long time. You know, what we found is that there are companies that they love Firefox, for example, but it doesn't really suit their needs in terms of being able to be more secure for them to be able to set policies around what, you know, for example, like what sites their employees are using to have more secure data policies. But also, frankly, that, you know, that there are a lot of— and the reason why we're building this in Europe is a lot of companies and institutions in Europe, you know, are increasingly focused on their data sovereignty and, you know, who has access to their data, especially their browsing data, and looking for, you know, solutions that don't necessarily— that can get them off of you know, Google, Microsoft, you know, and really give them a little bit more control over, you know, how their data is being used. So we're, you know, that we have our first customers in France right now, and we'll be really launching fully later this year and moving aggressively to, you know, focus on France and Germany to start, and then we'll come into the US next year and other markets. We have a security for AI company called Odin. And Odin is really interesting. They have a community of 1,700 researchers all around the world which are looking for vulnerabilities in large language models and agents and, you know, things like OpenClaw, and helping companies. And we just closing one of our biggest customers right now. It's a big name. I can't say who it is, but it's a company you know. And essentially like helping them secure their chatbots, agents, stuff like that from, you know, flaws and You think of it as almost like an automated red teaming type of product powered by, again, this community of researchers that are humans that are finding these things, not just synthetic. We have synthetic vulnerabilities as well, but I think the community and tapping into that wisdom is really powerful. We have an AI website builder called Solo, which has over 100,000 sites on it right now, built on it, which is pretty exciting. We have something called TabStack, which is essentially like a programmable API for the web. If you want to connect your, uh, agent— agentic experience and have it, um, be able to, uh, sort of tap into the web. Um, we have a product there, uh, and we have a handful of other things, uh, in the pipeline. Uh, and so, you know, pretty excited about the work that we're doing. Um, you know, it's, uh, uh, I think more important than ever, like, the work that Mozilla is doing to really try to make sure that the web stays a place that everybody can build, you know, build on and participate in.

Fascinating. I've never studied Mozilla, though. There is their history. They're a nonprofit, right? Like, can you educate me a bit more about their mission and what, what they've built? Obviously I use Firefox like a bunch of people. Yeah, tell me about like a quick crash history course of Mozilla and also where you guys are going in the AI revolution.

Yeah, so, um, you know, so Mozilla was started, um, over 20 years ago, um, essentially to, to, um, almost to like be a home for Firefox, right? And Firefox was started because there was a risk, and it's hard to remember, this is a long time ago, but there was a risk that Microsoft with Internet Explorer was going to essentially dictate the web and have a lot of control over what counted as a website. It was incredibly important to have an independent open source counterweight to that. Something that wasn't a business, so to speak. It's a business, but it's a nonprofit business. Our goal is, always to prioritize the needs of the internet and the web and the users of the web. That's our mission and that's, you know, non-negotiable for us, which is really great. And so, you know, over the past 20-plus years, Firefox is really, you know, has it lost market share? Absolutely, right? But there's no denying that. But it is an important, you know, seat, a voice, and we have a seat at the table when it comes to defining the standards for what counts as the web. And we are the only independent browser engine out there. It's incredibly important, I think, to make sure that a couple of big companies can't decide what counts as a website. You don't want to be in a situation where Google— I'm a Google user, love Google, use their products every day. This isn't about being anti-Google exactly or anti-Microsoft or anti-anybody. It's just that it's not healthy to have one or two companies being able to decide the rules of the game. And we've seen this with what's happened with apps, right? Where, um, you know, Tim Sweeney has fought a multi-year battle against Google and Apple to be able to crack open the, um, you know, the duopoly of the App Store, right? And, you know, as a VC, um, you know, I remember it was— I was at Betaworks for 6 years. You'd often talk to founders that would say, I have an app, I can't get into the App Store because it falls afoul of some rule, right? or, uh, you know, and they're perfectly good rules for the app stores, right? Um, but again, it's not necessarily healthy that we let two companies decide what apps can be on people's phones, right? I think that it's— it's— and so I have a lot of respect for Tim Sweeney for the, you know, the battle that he's waged to open this up. So this is really about the—

you can't compete with me. That's not exactly right.

That's, that's part of it, right, is, is, um we're always going to preference our own apps. We've seen that with Apple and Google over and over and over again. This is really what Mozilla is about. Now as we enter the age of AI, we're trying to have a similar position, which is let's try to have as much of the— and this isn't just like a Firefox thing, this is just across Mozilla, which is multiple entities. I work for Mozilla Corporation where we have Firefox and the new products organization. There's also Mozilla Ventures, There's Mozilla.ai, which is a startup focusing on building open-source AI tools. There's Mozilla Data Collective, which is a marketplace for datasets. There's Mozilla Foundation, which does our philanthropy and advocacy. And so, you know, at the end of the day, the thing that ties us all together, you know, when it comes to AI is that like we don't want to end up in a similar situation where a handful of companies have complete control. We know we support open-source AI, but we also support an open-source stack for developing on top of AI. So it's not just about the models themselves, it's about all the tools tools that go into the stack and making sure the developers have as much choice and control as possible over what, you know, they build and how they want to build it.

Any comments on the nonprofit structure? Is that actually helpful? We've seen how it have driven OpenAI, uh, to weird places. You feel something similar could happen with Mozilla?

You know, I think we're a little bit different, um, you know, in that, uh, I don't think there's like pressure to You know, take Mozilla Corporation, which is the for-profit entity with— that's owned by the nonprofit. I don't think we're— we have— we have that pressure. I don't think we're— we're not ever going to be a trillion-dollar company in the way that OpenAI is. But, you know, I would say that, you know, it is a tricky situation. And, you know, on the one hand, I understand why OpenAI felt like it needed to make some changes in order to be able to, you know, access the capital, the money it needed to be able to train models and deploy. It's— they ended up needing more money than you could raise for a nonprofit, right? You needed 10—

because Elon dropped them, for context, right? Like, Elon was out, and then Sam sort of panicked. It was like, we need massive revenue very quick. So that's why they launched ChatGPT.

Well, no, that's not exactly what happened. ChatGPT was not launched to become a profit business. Like, I mean, I know that I've talked with people there and it was launched actually to help get training data and it became a hit. And then they, I think they've more fully realized the potential of it as a business. But even before that, I think it was, the concern was that they were not gonna be able to, I mean, OpenAI was started because they didn't want Google to dominate AI, right?

Yeah, that was Elon specifically didn't want.

Yeah, Elon and Sam and others, right? So, um, you know, on one hand, I am— I am— I have sympathy for Elon's position that he put money into a nonprofit, um, and that, you know, it was converted into a for— essentially it's a for-profit company now, right?

Yeah, but no, I don't have zero sympathy for Elon. He wanted the— he wanted the power seat. He wanted the CEO role. He wanted power.

Well, that's what I— that was the thing I was going to follow up with. But he also clearly wanted it for himself, right?

Yes.

And so, you know, it's, it's, it's, it's, I think that notionally, I think the idea that a nonprofit, you know, was converted into a for-profit, I think is something that we should be really careful about. But what's a nonprofit?

Like, what's the intention really is like, we're not. And why would someone not want to profit in a capitalistic system in an AI arms race scenario? Is it, is it like nonprofits and unhealthy relationship with money?

I'm not sure I follow that.

Well, the nonprofit, I mean, if you want, in my opinion, if you want to help the world and create value in the world, right, like you need to make profits, you need to actually go for money necessarily, else you just can't scale, that else the competition will go for-profit and will not play in the same rules would advance quicker, in my opinion, would help the world faster than your own quote-unquote nonprofit. I understand that nonprofits want to do good, they want to help the world, but it seems the greatest model nowadays is, um, social startups, right? Like, uh, you go, you go for profits, you do good at the same time, and you, you can just advance faster. Of course, there's tax advantages to nonprofits, but I just feel that you need to play the capitalistic game to move as fast as possible.

Well, I think with respect to if you want to make AI models, I think it's hard to do as a nonprofit unless you were able to get government-scale backing. If the EU, for example, said, hey, we're going to commit $15 billion a year to having our own open, great, that would work. I think this is beyond the scale of what you could expect from philanthropic efforts from donors. I do think you have a point there. I think that there's a broader question about what's the right structure for things that are open source projects like Firefox. You could argue that— and Mozilla Corporation, just to be clear, is a for-profit entity. It's just wholly owned by a nonprofit. We do generate profits. We make money and it's a— we just let the profits from our work either go back into, you know, go back into the work that we do at Mozilla Corporation to fund new things like a lot of the work I'm doing, or we use that for philanthropic and advocacy roles. Now it helps that the amount of money we're able to make from Firefox covers the cost of developing Firefox, right? Um, you know, in theory you could get to a place where, um, you know, one of two things happens, right? Either like we make so much more money than it requires to develop that, um, you know, you start to think about like, you know, what is the right structure for something like this in terms of, um, you know, uh, generating lots and lots of profit and making sure that we're responsibly using those profits in alignment with our nonprofit mission. Or you could say that the cost of developing the engine exceeds the amount of revenue we can make. And then you have to sort of think about like, you know, how do you solve that problem, right? And, um, you know, so I, I think the structure we have right now, it works. It's worked for 20 years. I think it'll continue to work. Um, but, uh, you know, I can understand why in the case of OpenAI there was, you know, the other path was for them to not be able to build, um, you know, ChatGPT-5, right? They would not have been able to do that, right?

Do you agree that the latest attacks on the internet have been social media-based, aka Zuck like modifying the social fabric, and then Cambridge Analytica, and as of late Elon acquiring X, uh, to veer the elections in some way, and/or Trump starting The Truth, which is pretty ridiculous.

Yeah, I mean, that's— I think that what we found is that, you know, that social media is so key to shaping people's perceptions of what is going on and what is important that, you know, that there's really no way— there's not like a— we don't live in a world where there is, um, you know, just sort of like a regular media anymore, right? It's just everything is, is filtered through the experience of TikTok and X. Forgot to mention TikTok. TikTok. And, um, and I think that that's how people now experience and filter— it's how— I just say it's how people structure their understanding of the world around them now, right? It's all, it's all mediated through these social media. And it's funny because my master's degree is in critical theory. And if you go back and read people like Guy Debord from the '60s, Jean Baudrillard from the '80s, they were talking about this stuff decades ago. They were saying that we live actually in a world where our understanding is actually mediated through images, right, through media. I mean, I was sort of mediated through media, but like that, and it really structures how we think about things and how we think about what is real and what is important. And it's not just an AI, and I don't mean it like in an AI deepfake kind of way, because it's not about AI or not AI in this, AI maybe makes some things a little bit worse, maybe make some other things better, but this isn't about like, about fake versus— this isn't about like creating fake things exactly. It's that the nature of social media is so much about harnessing and aggregating attention. And we know what drives attention. We know what the hacks for driving attention are, right? It's about sensationalism. It's about controversy. It's about, you know, hatred, right? And so it's about inflaming emotions, right? If you go back to Marshall McLuhan, he talks about how as we start to like our social relations start to flow through media like television, now the internet, we default into being a very reactive emotional place. That's hard to work with. It's hard to, it's hard to, you know, uh, it's hard politically, I think, to create consensus when everyone is coming from a place of being, you know, emotionally inflamed.

How does one fix that then? That— can Mozilla like take a part in fixing that? Because governments are not about to come up with laws to, uh, fix that problem. I don't— I think it's pretty draconian as well to— for government to try to control the internet and tell people what to watch, right? Like, is there an actual solution to that and how could that get worse with AI?

Yeah, I wish I knew. I wish that there was some— there are easy things to do because if I knew what to do, I would have built it. Not just necessarily for Mozilla, I would have just already done it. I think that— so I don't have good answers. I think you're right that Government regulation, there's some things that might help. But I think at the end of the day, we don't necessarily— making governments the arbiter of how people can participate or what they can consume or share or create, I think it's also a really difficult place to be. I don't have answers for that. I wish that there was more that we could do. I think that the part that we can play is that by making sure that at least the web is a place that you know, is participatory and, and is not, you know, dominated by a handful of big companies, that there's still a place for people to be able to create their own websites, their own media, right? Um, you know, that it's not just about algorithmic algorithms, like, you know, determining like what you see, um, and what you consume. But it's a— that's a— that's one of the— you're talking about probably maybe the biggest question facing us in the 21st century, and You know, with respect to AI, um, you can already see that the ability to manipulate at scale will get worse because the cost of, of creating fake, um, comments, fake content, you know, um, you know, fake influencer campaign— campaigns— that cost is only going to drop and drop and drop. And, you know, people are so cynical now they don't believe, they don't trust it very much anymore. Sorry, you— I think you went on mute.

You said Mozilla vs Google and Apple, but didn't most of Mozilla's revenue come from Google?

It does, yeah. And that's part of what my job is to do, is to diversify our revenue so that we can start to get away from having most of our revenue come from Google. So that's, that's job number one for me.

Interesting. And the next products we've briefly talked about, like GenAI cybersecurity type of product. You've heard about Claude Mithos not so long ago.

We actually did. We were one of their partners for that rollout.

Oh, nice, man.

And tell me a bit more about that.

That was groundbreaking, right? We talked about the open Claude moment, which also we can talk about that moment, but I feel the Claude Mithos moment is also like a a historic moment in which we're like, holy fuck, we're potentially in trouble here.

Yeah, yeah. You know, it's, you know, Anthropic, I think, has been trying to be responsible about how they've, you know, introduced this. We had access to Mythos early and before it was announced. And we, you know, we were able to partner with them to identify, you know, a few hundred, a couple hundred vulnerabilities, you know, bugs in Firefox and we spent a lot of time patching those up and fixing those over the past couple of months. This is something that is going to be an issue going forward. It's going to be, can AI help patch the bugs? Yeah, I think it can. But I think that we're going to be in a place where there's going to be a lot of— sooner or later, there's probably going to be some AI-identified vulnerability that gets exploited. And a lot of people are going to, you know, people might lose a lot of money. There might be, you know, something really bad will probably happen out of it, right? That's probably an inevitability. And I think that's why, you know, Anthropic and, you know, at least is, you know, trying to be really responsible what they do. I think the challenge is that we all know, you know, a year and a half from now, 2 years from now, like, you know, the best models will be probably much more powerful than Mythos is right now. And so, um, you know, I, I don't know, um, you know, it's already illegal to hack, you know what I mean? So it's like, um, uh, you know, so I think that like there'll be some real questions about how do we, uh, you know, how do we intelligently regulate use of AI so that we don't end up in a place where, you know, you wake up one day and your bank account's empty.

Yeah, yeah, that in my opinion, that will happen anyway with the AI revolution, the agentic revolution. I do think a bunch of people are not going to be able to keep a job in the agentic revolution. Do you feel the same way?

You know, I'm kind of two minds because I think that, um, I think, you know, I, I do think that that there's a case that while the nature of jobs may shift, that the work— there will still be new things. And that, you know, I think we've seen examples, generally speaking, that when we can make people more productive, you know, it ends up creating like net positive effects for the economy. And, you know, people tend to, you know, require— we tend to hire more overall in the aggregate. Now, could I predict exactly what, like I have kids, I have a 17-year-old, my oldest. Could I predict what job he might have in 10 years or 5 years when he's graduating from college? No. I wouldn't necessarily tell him to go and learn Python right now. Maybe that might not be exactly the right skill to use, learn versus Being able to learn some more deep-level understanding of systems and programming to be able to architect versus going deep into one specific language, for example, to become more of a systems-level architect. He's not interested in computer science, but I'm cautiously optimistic is the way I would say. The thing I worry about a little more, about— I worry more about the impact of, um, you know, irresponsible, like, fiscal policy, uh, you know, congressional spending, you know, like governmental spending, things like that, um, you know, versus, like, uh, uh, you know, the risk of there being a de-dollarization of the global economy, um, you know, things like that.

Um, interesting that you worry about this before AI. Like, to me, AI and the agentic revolution, which is about to kickstart, is gonna put everything in shambles, and I would worry way more about that.

I don't think so. I just don't think so.

But if I have the option as a CEO to hire an agent that's 1,000 times faster than a human, and granted I'm already recruiting 1% humans, where does that leave like the 99%?

Well, first of all, I think, um, we're finding that the costs are still high, right? Like, you generate a lot of tokens it's expensive.

True.

It's not cheap. There's that. It's not as cheap necessarily as we might think. I think too, I think that we've all found that agents are good, but they still require human supervision. I think Aaron Levy from box.net had a great piece about this saying, you're probably better off keeping your people and giving them the tools to be more productive than getting rid of people. And I think that, like, you know, I think that if, if, um, you know, we're also assuming that everyone stays at the same sort of baseline in terms of productivity, which is a mistake, right? As productivity goes up, your competition as a company is going to go up, you know. And so yeah, if you want to be like— I think if you want to be a solopreneur, one person Um, you know, yeah, agents are going to make it great. You're probably not going to have to hire many more people, but your goal was never to hire thousands of people anyway, right?

Um, I think that, you know, but if we stop hiring, right, like, isn't that gonna create a net loss overall?

But there hasn't been a net loss so far yet.

But what if I have an agent that can just reverse engineer the output of my employee and then put an agent right next to them and then compare both?

I think it will be a long time until you will, I think it will be longer than you think for that agent to be good enough to fully trust it. Right now, what are agents doing? They're basically like spamming people on LinkedIn. You know what I mean?

Well, the litmus is that OpenAI and Anthropic haven't deployed agents yet, and these guys move very fast, right? So that would bring some degree of truth to what you're saying.

Well, that's what I'm saying. Like, yeah, like, if these things were as powerful, I think your point is right, that like, OpenAI would be firing people, not hiring people.

Right. And they would be deploying agents out there, productive agents, because you followed their browser, right? Like the Atlas thing. And the hopes was that Atlas was going to learn of what you do and would help agents deploy themselves faster, but that failed miserably.

Yeah. I mean, let's use another example, right? Let's say that we completely replaced all of, um, you know, like trucking with self-driving cars, self-driving trucks, right? For everything. Last line, you know, like deliveries, everything, right? Um, so let's say that the cost of shipping something dropped a lot because of that, arguably. Um, yes. Now, would that put a lot of people, drivers, out of work? Yes. Would it actually make it— would it enable lots and lots of businesses which pay— have to ship things to become more profitable and more productive because their shipping costs would go down, enabling them to grow their businesses more quickly? People buying more because shipping costs are lower, you know. Um, now, like, I'm not advocating jobs because that's what I'm saying, and it's creating other products, components.

But robots could quickly start. I, I think the humanoid revolution will take a way longer time, like 5, 10 years, but I think probably longer than that. But yeah, interesting. But wait, my thesis, my final, my finality is not like bleak. I think most people will become entrepreneurs is my thinking. I, I think humans are gritty, and I do think a lot of them would choose like VR porn over like working on a daily basis if they were given the choice, but I do think that a bunch will adapt and become entrepreneurs and not accept a job.

I think that we'll see— look, solopreneurs are growing more quickly, we're growing very rapidly right now, and our site, so our AI website builder Solo, is aimed at solopreneurs, right? It's like specifically targeting solopreneurs. That's, that's what we're going after, um, for that reason, because we identified that as a really quickly growing, um, category. I think that, um, you know, we'll, we'll be able to enable a lot more people to build, you know, a lot, uh, to build a business with, um, you know, we'll lower the— we'll make it easier to start a business, which will be good, I think, in the long run for the economy. I think that, um, for us, for bigger companies, I think that they'll be able to, you know, have more productive, uh, employees, which means that they'll be able to grow their businesses as well, right? And so, you know, look, I think the question is really there will be dislocation. If you think about globalization in the '80s and the '90s, it resulted in a lot of people in factory jobs, for example, losing— they lost their jobs. And it created dislocation, which we're still dealing with decades later in lots of parts of the US and other parts of the world. And I think that The question I would ask is, from a, you know, uh, countries have industrial policies, um, you know, and smart ones have, have good industrial policies, right? Um, Germany, South Korea, Taiwan, right? You know, Vietnam, for example. Um, they all have industrial policies, right? We should have an industrial policy for how we want to think about people that are dislocated from jobs. Um, you know, it's not just about retraining. It's, it's not just about, uh, you know You know, there are some things that we can— we should be talking about to be able to, you know, soften the blow and to ease the transition if we are going to have people— like, some jobs we're going to need fewer. We're probably going to need fewer truck drivers eventually, for example.

I had Seth Levine on my podcast not so long ago, and he talked about the Neigentic tax. I thought that was pretty smart. Taxing like agents, tokens. What do you think about that one?

You know, um, I, I was at a conference last year where we talked about taxing AI, right? Uh, and, um, you know, it's certainly— there's some interesting policy work that's being done there. I'm not sure I know it well enough to be able to think about like the implications of it, but, you know, I think generally speaking you want to be able to, um, you know, capture tax revenue, right, and do redistribution when it makes sense, and that might be something where it does make sense. Uh, but I don't know enough about the specifics of the policy to be able to like have a really informed opinion there.

Well, I'll follow your work, Peter, at Firefox, eager to see what you build there. Where can people, uh, follow you or learn more about what you do?

Um, I, um, my personal site, roj.as, it's the easiest place to find me.

All right, so it's Peter and Charles for another Founder